{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"**이 노트북은 이미지 분류 분야에서 sota를 달성한 모델을 살펴보고 각 모델이 가진 아이디어를 이해하여**\n\n**새로운 Convolutional Neural Network 모델을 연구할 때 보다 근거 있고 창의적인 실험을 할 수 있도록 하기 위하여 제작되었습니다.**\n\n**데이터의 크기가 작아 빠르게 결과를 볼 수 있는 mnist 데이터를 이용하여 제작하였습니다.**","metadata":{}},{"cell_type":"markdown","source":"# **Import**","metadata":{}},{"cell_type":"markdown","source":"## **- 데이터 처리**","metadata":{}},{"cell_type":"markdown","source":"csv로 되어있는 mnist 데이터를 불러오기 위해 pandas를 import 합니다.\n\n다음으로 2차원 이미지 데이터의 각 픽셀값을 다루기 위해 numpy를 import 합니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 모델 생성 및 검증**","metadata":{}},{"cell_type":"markdown","source":"먼저 검증 데이터 분리를 위한 train_test_split을 import 합니다.\n\n이후 머신러닝 모델 학습을 위해 support vector machine를 import 합니다.\n\n그리고 딥러닝 모델 구현을 위한 tensorflow를 import한 다음 정확도를 측정하기 위해 accuracy_score를 import 합니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 시각화**","metadata":{}},{"cell_type":"markdown","source":"이미지 데이터 시각화를 위해 matplotlib를 import 합니다.\n\n그리고 딥러닝 모델을 시각화하기 위해 plot_model을 import 합니다.","metadata":{}},{"cell_type":"code","source":"import numpy as np\nimport pandas as pd\nimport tensorflow as tf\nfrom sklearn.svm import SVC\nimport matplotlib.pyplot as plt\nfrom sklearn.metrics import accuracy_score\nfrom tensorflow.keras.utils import plot_model\nfrom sklearn.model_selection import train_test_split","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2022-08-14T21:15:45.628470Z","iopub.execute_input":"2022-08-14T21:15:45.629701Z","iopub.status.idle":"2022-08-14T21:15:54.030191Z","shell.execute_reply.started":"2022-08-14T21:15:45.629589Z","shell.execute_reply":"2022-08-14T21:15:54.028973Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **데이터 불러오기 (data load)**","metadata":{}},{"cell_type":"markdown","source":"pandas를 이용하여 학습에 사용할 train.csv 데이터를 불러옵니다.","metadata":{}},{"cell_type":"code","source":"image_data = pd.read_csv(\"/kaggle/input/digit-recognizer/train.csv\")","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:54.032137Z","iopub.execute_input":"2022-08-14T21:15:54.032793Z","iopub.status.idle":"2022-08-14T21:15:57.502021Z","shell.execute_reply.started":"2022-08-14T21:15:54.032753Z","shell.execute_reply":"2022-08-14T21:15:57.500920Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"각 행은 이미지가 나타내는 숫자와 픽셀들의 명도로 이루어져 있습니다.","metadata":{}},{"cell_type":"code","source":"image_data.head()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:57.503381Z","iopub.execute_input":"2022-08-14T21:15:57.503821Z","iopub.status.idle":"2022-08-14T21:15:57.534985Z","shell.execute_reply.started":"2022-08-14T21:15:57.503780Z","shell.execute_reply":"2022-08-14T21:15:57.533151Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"픽셀값을 학습하여 라벨링 되어있는 값을 분류하기 위해 두 값을 분리해줍니다.","metadata":{}},{"cell_type":"code","source":"X = image_data.iloc[:, 1:].values\ny = image_data.iloc[:, 0].values","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:57.538749Z","iopub.execute_input":"2022-08-14T21:15:57.539234Z","iopub.status.idle":"2022-08-14T21:15:57.545134Z","shell.execute_reply.started":"2022-08-14T21:15:57.539196Z","shell.execute_reply":"2022-08-14T21:15:57.544024Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Convolutional Neural Network 등장배경**","metadata":{}},{"cell_type":"markdown","source":"1998년 손글씨로 되어있는 우편번호를 자동으로 인식하기 위한 Gradient-Based Learning Applied to Document Recognition 라는 논문이 등장합니다.\n\n이 논문은 Convolutional Neural Network를 최초로 적용하여 mnist 이미지 데이터 인식문제에서 높은 정확도를 보였습니다.\n\nmnist 이미지 데이터는 손글씨로 쓰여진 0부터 9까지의 숫자를 28*28 이미지로 나타낸 데이터입니다.\n\n먼저 mnist 이미지 데이터를 예시로 시각화하여 살펴보았습니다.","metadata":{}},{"cell_type":"code","source":"X = X.reshape(X.shape[0], 28, 28)\n\nplt.imshow(X[0], cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:57.546499Z","iopub.execute_input":"2022-08-14T21:15:57.546935Z","iopub.status.idle":"2022-08-14T21:15:57.773616Z","shell.execute_reply.started":"2022-08-14T21:15:57.546899Z","shell.execute_reply":"2022-08-14T21:15:57.772530Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Convolutional Neural Network 등장 이전**","metadata":{}},{"cell_type":"markdown","source":"물론 Convolutional Neural Network가 등장하기 전에도 이미지 데이터를 인식하기 위해 다양한 연구들이 진행되었습니다.\n\n기존에는 이미지를 1차원으로 flatten 시킨 후에 사람이 작성한 알고리즘을 활용하여 이미지의 특징을 찾았습니다.(Feature Extractor) \n\n그다음에는 다양한 머신러닝 모델이나 Neural Network를 이용하여 분류(classification)를 진행하였습니다.\n\n그럼 1차원으로 flatten된 이미지를 시각화하여 살펴보았습니다.","metadata":{}},{"cell_type":"code","source":"X1 = X.reshape(X.shape[0], 784)\n\nplt.figure(figsize=(20,5))\nplt.imshow(X1[0][150:350].reshape(1,200), cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:57.774844Z","iopub.execute_input":"2022-08-14T21:15:57.775228Z","iopub.status.idle":"2022-08-14T21:15:57.964275Z","shell.execute_reply.started":"2022-08-14T21:15:57.775194Z","shell.execute_reply":"2022-08-14T21:15:57.963321Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **기존 방식의 문제점**\n\n1차원으로 flatten 시킨 이후 인공지능을 이용해 학습시킬 경우 크게 3가지의 문제가 발생합니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 공간 정보의 손실**\n\n먼저, 이미지는 인접한 픽셀 사이에 높은 상관관계를 가지고 있습니다.\n\n그러나 이미지를 1차원으로 flatten 시키면 공간이 가지고 있는 정보를 잃게 됩니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 이미지 특징의 손실**\n\n또, 기존에는 사람이 작성한 알고리즘을 이용해 이미지의 특징을 학습시켰습니다.\n\n그러나 사람이 이미지의 특징을 놓칠 수 있다는 점이 있습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 연산의 비효율성**\n\n마지막으로 flatten 시킨 모델은 784개의 feature를 가진 input 데이터를 처리하게되는데, 이는 막대한 연산량을 가지게됩니다.","metadata":{}},{"cell_type":"markdown","source":"# **기존 방식 구현**","metadata":{}},{"cell_type":"markdown","source":"기존 방식으로 machine laerning과 neural network를 이용하여 분류를 진행하였습니다","metadata":{}},{"cell_type":"markdown","source":"## **- 데이터셋 분리**","metadata":{}},{"cell_type":"markdown","source":"먼저 모델을 검증시키기 위해 데이터셋을 훈련 데이터셋과 검증 데이터셋으로 나눕니다.\n\nneural network에서 학습시킬 때 다양한 가중치를 곱하며 값이 커져 계산이 복잡해지는 것을 막기 위해 이미지 scaling을 진행해줍니다","metadata":{}},{"cell_type":"code","source":"X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state = 62)\n\nX_test = X_test / 255\nX_train = X_train / 255","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:57.965639Z","iopub.execute_input":"2022-08-14T21:15:57.966211Z","iopub.status.idle":"2022-08-14T21:15:58.555708Z","shell.execute_reply.started":"2022-08-14T21:15:57.966179Z","shell.execute_reply":"2022-08-14T21:15:58.554582Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Machine Learning**","metadata":{}},{"cell_type":"markdown","source":"이제 flatten한 데이터를 머신러닝 모델을 이용해 분류해보았습니다.\n\n784개의 input 데이터를 학습시키기 위하여 많은 feature에서도 효율적인 성능을 보이는 support vector machine을 이용하였습니다.","metadata":{}},{"cell_type":"code","source":"model = SVC()\nmodel.fit(X_train.reshape(X_train.shape[0], 784),y_train)\npred = model.predict(X_test.reshape(X_test.shape[0], 784))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:15:58.556997Z","iopub.execute_input":"2022-08-14T21:15:58.557407Z","iopub.status.idle":"2022-08-14T21:18:59.841786Z","shell.execute_reply.started":"2022-08-14T21:15:58.557372Z","shell.execute_reply":"2022-08-14T21:18:59.840767Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **- 모델 학습 결과**","metadata":{}},{"cell_type":"markdown","source":"약 97.5%의 정확도를 보이는 것을 확인할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"accuracy_score(pred,y_test)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:18:59.843250Z","iopub.execute_input":"2022-08-14T21:18:59.843772Z","iopub.status.idle":"2022-08-14T21:18:59.853799Z","shell.execute_reply.started":"2022-08-14T21:18:59.843729Z","shell.execute_reply":"2022-08-14T21:18:59.852511Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **Neural Network**","metadata":{}},{"cell_type":"markdown","source":"다음으로 Neural Network을 이용하여 분류를 진행해보았습니다\n\n간단한 Neural Network 모델을 구현해보아도 첫 레이어부터 200960개의 parameter를 가지게 됩니다.\n\n심지어 보편적인 이미지는 28*28보다 몇배는 크기때문에 neural network은 활용도가 떨어진다는 문제점이 있습니다.","metadata":{}},{"cell_type":"markdown","source":"처음에는 학습을 진행하는 정도를 결정하는 learning_rate를 설정합니다.","metadata":{}},{"cell_type":"code","source":"learning_rate = 0.00062","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:18:59.856758Z","iopub.execute_input":"2022-08-14T21:18:59.857148Z","iopub.status.idle":"2022-08-14T21:18:59.862333Z","shell.execute_reply.started":"2022-08-14T21:18:59.857115Z","shell.execute_reply":"2022-08-14T21:18:59.861254Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"그리고 Neural Network에서 10개의 class로 분류를 진행하기 위하여 to_categorical을 적용합니다.","metadata":{}},{"cell_type":"code","source":"y_train = tf.keras.utils.to_categorical(y_train, 10)\ny_test = tf.keras.utils.to_categorical(y_test, 10)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:18:59.863618Z","iopub.execute_input":"2022-08-14T21:18:59.864481Z","iopub.status.idle":"2022-08-14T21:18:59.874658Z","shell.execute_reply.started":"2022-08-14T21:18:59.864445Z","shell.execute_reply":"2022-08-14T21:18:59.873661Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이후 784개의 input과 softmax를 이용한 10개의 output를 만드는 모델을 작성합니다.","metadata":{}},{"cell_type":"markdown","source":"이때 tensorflow를 이용하면 두가지 방법으로 모델을 작성할 수 있습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- sequential api**","metadata":{}},{"cell_type":"markdown","source":"sequential하게 Neural Network를 작성해봅시다.\n\nsequential api를 사용하면 keras의 sequential 모델에 layer를 추가하는 방식으로 모델을 작성하게되며 직관적이라는 장점이 있습니다.\n\n먼저 keras의 sequential을 model 변수에 지정했습니다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.Sequential()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:18:59.876193Z","iopub.execute_input":"2022-08-14T21:18:59.876714Z","iopub.status.idle":"2022-08-14T21:18:59.943702Z","shell.execute_reply.started":"2022-08-14T21:18:59.876682Z","shell.execute_reply":"2022-08-14T21:18:59.942607Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- input layer**","metadata":{}},{"cell_type":"markdown","source":"그리고 입력을 받을 layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Input(shape=[28, 28, 1]))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:18:59.945163Z","iopub.execute_input":"2022-08-14T21:18:59.945591Z","iopub.status.idle":"2022-08-14T21:18:59.958922Z","shell.execute_reply.started":"2022-08-14T21:18:59.945555Z","shell.execute_reply":"2022-08-14T21:18:59.958081Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"분류를 위해 flatten layer를 추가시키고 dense layer를 이용하여 분류를 진행했습니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Flatten())\nmodel.add(tf.keras.layers.Dense(units=256, activation='relu'))\nmodel.add(tf.keras.layers.Dense(units=512, activation='relu'))\nmodel.add(tf.keras.layers.Dense(units=512, activation='relu'))\nmodel.add(tf.keras.layers.Dense(units=256, activation='relu'))\nmodel.add(tf.keras.layers.Dense(units=128, activation='relu'))\nmodel.add(tf.keras.layers.Dense(units=10, activation='softmax'))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:18:59.960332Z","iopub.execute_input":"2022-08-14T21:18:59.961400Z","iopub.status.idle":"2022-08-14T21:19:00.074463Z","shell.execute_reply.started":"2022-08-14T21:18:59.961355Z","shell.execute_reply":"2022-08-14T21:19:00.073548Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"모델을 summary를 이용하여 모델을 확인할 수 있습니다.\n\n레이어가 잘 배치된 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.075631Z","iopub.execute_input":"2022-08-14T21:19:00.075937Z","iopub.status.idle":"2022-08-14T21:19:00.081990Z","shell.execute_reply.started":"2022-08-14T21:19:00.075910Z","shell.execute_reply":"2022-08-14T21:19:00.081110Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **- functional api**","metadata":{}},{"cell_type":"markdown","source":"다음으로는 같은 모델을 functional api를 이용하여 작성하여봅시다.\n\nfunctional api를 사용하면 보다 자유로운 모델 구성이 가능하다는 장점이 있습니다.","metadata":{}},{"cell_type":"markdown","source":"먼저 이미지를 input으로 받는 layer를 생성합니다.","metadata":{}},{"cell_type":"code","source":"x = tf.keras.layers.Input(shape=[28, 28, 1])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.083304Z","iopub.execute_input":"2022-08-14T21:19:00.084313Z","iopub.status.idle":"2022-08-14T21:19:00.094994Z","shell.execute_reply.started":"2022-08-14T21:19:00.084277Z","shell.execute_reply":"2022-08-14T21:19:00.094126Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"다음으로 flatten layer를 이전의 input layer인 x와 연결해주어 784개의 1차원 데이터로 변형했습니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Flatten()(x)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.096867Z","iopub.execute_input":"2022-08-14T21:19:00.097677Z","iopub.status.idle":"2022-08-14T21:19:00.110542Z","shell.execute_reply.started":"2022-08-14T21:19:00.097632Z","shell.execute_reply":"2022-08-14T21:19:00.109416Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이후 dense layer들을 이용해 network를 작성합니다","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Dense(units=256, activation='relu')(h)\nh = tf.keras.layers.Dense(units=512, activation='relu')(h)\nh = tf.keras.layers.Dense(units=512, activation='relu')(h)\nh = tf.keras.layers.Dense(units=256, activation='relu')(h)\nh = tf.keras.layers.Dense(units=128, activation='relu')(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.111947Z","iopub.execute_input":"2022-08-14T21:19:00.112391Z","iopub.status.idle":"2022-08-14T21:19:00.165347Z","shell.execute_reply.started":"2022-08-14T21:19:00.112358Z","shell.execute_reply":"2022-08-14T21:19:00.164225Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"마지막으로 softmax를 적용하여 10개의 class로 분류합니다.","metadata":{}},{"cell_type":"code","source":"Y = tf.keras.layers.Dense(units=10, activation='softmax')(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.166905Z","iopub.execute_input":"2022-08-14T21:19:00.167332Z","iopub.status.idle":"2022-08-14T21:19:00.182568Z","shell.execute_reply.started":"2022-08-14T21:19:00.167296Z","shell.execute_reply":"2022-08-14T21:19:00.181476Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"x를 input으로하고 Y를 output으로하는 모델을 keras.models.model을 이용해 생성합니다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.models.Model(x, Y,name = 'dense_model')\nmodel.compile(loss='categorical_crossentropy', optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate), metrics=['accuracy'])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.183943Z","iopub.execute_input":"2022-08-14T21:19:00.184516Z","iopub.status.idle":"2022-08-14T21:19:00.204883Z","shell.execute_reply.started":"2022-08-14T21:19:00.184480Z","shell.execute_reply":"2022-08-14T21:19:00.203971Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"summary 를 통해 모델을 확인하면 sequential한 모델과 같은 760714개의 파라미터로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.206529Z","iopub.execute_input":"2022-08-14T21:19:00.206871Z","iopub.status.idle":"2022-08-14T21:19:00.213330Z","shell.execute_reply.started":"2022-08-14T21:19:00.206841Z","shell.execute_reply":"2022-08-14T21:19:00.212213Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하면 모델의 연결을 시각화하여 볼 수 있습니다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:00.215017Z","iopub.execute_input":"2022-08-14T21:19:00.215496Z","iopub.status.idle":"2022-08-14T21:19:01.448374Z","shell.execute_reply.started":"2022-08-14T21:19:00.215452Z","shell.execute_reply":"2022-08-14T21:19:01.446781Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습**","metadata":{}},{"cell_type":"markdown","source":"fit을 이용하여 X_train과 y_train을 학습시키고 X_test와 y_test로 검증해보았습니다.","metadata":{}},{"cell_type":"code","source":"history = model.fit(X_train, y_train, batch_size=100, epochs=20, validation_data = (X_test,y_test), verbose=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:19:01.452194Z","iopub.execute_input":"2022-08-14T21:19:01.452768Z","iopub.status.idle":"2022-08-14T21:20:22.479532Z","shell.execute_reply.started":"2022-08-14T21:19:01.452704Z","shell.execute_reply":"2022-08-14T21:20:22.478426Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습 결과**","metadata":{}},{"cell_type":"markdown","source":"약 97.6%의 정확도를 보이는 것을 확인할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"max(history.history[\"val_accuracy\"])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:22.481128Z","iopub.execute_input":"2022-08-14T21:20:22.481490Z","iopub.status.idle":"2022-08-14T21:20:22.488362Z","shell.execute_reply.started":"2022-08-14T21:20:22.481459Z","shell.execute_reply":"2022-08-14T21:20:22.487211Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"그러나 학습과정을 살펴보면 불안정한 그래프를 확인할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"history_frame = pd.DataFrame(history.history)\nhistory_frame.loc[:, ['loss', 'val_loss']].plot()\nhistory_frame.loc[:, ['accuracy', 'val_accuracy']].plot();","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:22.489446Z","iopub.execute_input":"2022-08-14T21:20:22.489769Z","iopub.status.idle":"2022-08-14T21:20:22.944857Z","shell.execute_reply.started":"2022-08-14T21:20:22.489740Z","shell.execute_reply":"2022-08-14T21:20:22.943847Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **lenet**\n\nlenet은 기존의 3가지 문제를 해결하기 위해 convolution layer와 pooling layer를 고안하였습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 공간 정보 보존**\n\n먼저, 이미지의 공간이 가지고 있는 정보를 보존하기 위해 2차원 상태에서 convolution layer를 이용해 특징 추출(Feature Extractor)을 진행하였습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 이미지 특징 보존**\n\n다음으로, 사람이 놓칠 수 있는 특징을 고려하기 위하여 모든 feature를 이용하여 Feature Extractor 과정을 진행하였습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 효율적인 연산**\n\n이때, Feature Extractor 과정이 효과적으로 진행된다면 유의미한 특징만 남게되어 적은 연산량으로 분류를 수행할 수 있게됩니다.","metadata":{}},{"cell_type":"markdown","source":"# **convolution layer**\n\n그럼 이제 lenet에 사용된 convolution layer의 개념을 알아봅시다.","metadata":{}},{"cell_type":"markdown","source":"## **- filter**","metadata":{}},{"cell_type":"markdown","source":"먼저, convolution layer에는 이미지에서 특징을 찾아내기 위해 픽셀별로 weight과 bias를 가지는 사각형 모양의 filter가 있습니다.\n\nconvolution layer에서는 filter를 다양하게 생성하여 다음 layer로 값을 전달합니다.","metadata":{}},{"cell_type":"markdown","source":"## **- stride**","metadata":{}},{"cell_type":"markdown","source":"stride는 filter를 얼만큼 이동하며 학습을 진행할지 결정하는 수치입니다.\n\nconvolution layer는 filter를 stride만큼 이동시키며 convolution 연산을 적용하여 결과를 산출합니다.","metadata":{}},{"cell_type":"markdown","source":"## **- convolution**","metadata":{}},{"cell_type":"markdown","source":"이때, convolution 연산은 이미지에서 filter에 해당하는 값에 weight과 bias를 적용하여 합한 하나의 값을 결과의 한 픽셀로 도출합니다.\n\nkaggle 강의에 있는 이미지를 보시면 개념을 더 잘 이해할 수 있습니다.\n\nhttps://www.kaggle.com/code/ryanholbrook/the-sliding-window","metadata":{}},{"cell_type":"markdown","source":"# **pooling layer**","metadata":{}},{"cell_type":"markdown","source":"다음으로는 lenet을 구성하는 또다른 요소인 pooling layer에 대해 알아봅시다.\n\npooling layer는 각 지역을 대표하는 값을 정의하여 출력하는 과정입니다.\n\nfilter의 크기를 정하면 filter가 stride만큼 움직이며 각 filter에 해당하는 값의 평균이나 최대값을 그 filter의 대푯값으로 지정하여 결과를 산출합니다.\n\npooling 과정을 거치면 이전 layer의 출력 이미지를 작게하거나 대푯값을 이용하여 노이즈 등을 무시하고 특징을 강화하여 인식하려는 대상의 왜곡이나 위치의 차이를 극복하는데 도움이 됩니다.\n\npooling에는 3가지 종류(max pooling, average pooling, global average pooling)가 있습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- max pooling**","metadata":{}},{"cell_type":"markdown","source":"max pooling은 각 filter에 해당하는 범위의 최대값만을 다음 layer로 전달합니다.\n\nmax pooling은 확실하게 특징을 잡아낼 수 있다는 장점이 있습니다.\n\n하지만 작은 값이 무시되고 특징의 모양이 조금 달라질 수 있습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- average pooling**","metadata":{}},{"cell_type":"markdown","source":"다음으로는 average pooling이 있습니다.\n\naverage pooling은 각 filter에 해당하는 범위의 평균값을 다음 layer로 전달합니다.\n\naverage pooling은 덜 중요한 요소도 포함할 수 있다는 특징이 있습니다.\n\n그리고 분산을 이용하여 대상의 위치를 보다 쉽게 알 수 있어서 객체 탐지에 이용할 수 있습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- global average pooling**","metadata":{}},{"cell_type":"markdown","source":"마지막으로 global average pooling은 각 채널별로 평균값을 출력합니다.\n\n이는 차원을 빠르게 감소시킬 수 있는 장점을 가집니다.","metadata":{}},{"cell_type":"markdown","source":"# **padding**","metadata":{}},{"cell_type":"markdown","source":"그러나 convolution layer나 pooling layer를 거치며 생기는 문제가 있습니다.\n\n이것은 이미지가 축소되고 이미지의 끝부분에 있는 값이 filter의 중앙에 위치하는 경우가 존재하지 않아서 끝부분에 있는 특징을 놓칠 수 있다는 것입니다.\n\nlenet은 이 문제를 해결하기 위해서 padding 기법을 적용하게 됩니다.\n\npadding 기법은 이미지의 주변에 특정값을 추가하여 이미지의 크기를 늘리고 convolution layer나 pooling layer를 거쳐도 이미지의 끝부분의 값도 고려하면서 출력 데이터의 사이즈도 조절할 수 있습니다.","metadata":{}},{"cell_type":"markdown","source":"이외에도 Convolutional Neural Network는 대상의 위치 변화, 크기 변화, 왜곡에 대응하기 위하여 2가지 아이디어(Local receptive field, Shared weight)을 적용하였습니다.","metadata":{}},{"cell_type":"markdown","source":"# **Local receptive field**","metadata":{}},{"cell_type":"markdown","source":"filter를 이동시키며 filter에 해당하는 특징을 다음 layer에 적용시켜서 Feature Extractor를 진행하기 때문에 인식하려는 대상의 위치가 어디에 있든 대상의 특징을 나타내는 값이 검출된다는 것이 Local receptive field의 특징입니다.","metadata":{}},{"cell_type":"markdown","source":"# **Shared weight**","metadata":{}},{"cell_type":"markdown","source":"한 채널에서 filter의 이동마다 weight을 동일하게 적용시키면 대상의 특징이 어디에 있든 공유하는 filter의 weight에 영향을 미쳐서 filter가 특징을 학습할 수 있습니다. ","metadata":{}},{"cell_type":"markdown","source":"# **모델 구현**","metadata":{}},{"cell_type":"markdown","source":"이제 Convolutional Neural Network를 최초로 이용한 lenet-1부터 lenet-4, lenet-5를 tensorflow를 활용해 구현하며 모델의 구조를 알아봅시다.","metadata":{}},{"cell_type":"markdown","source":"# **lenet-1**","metadata":{}},{"cell_type":"markdown","source":"![image.png](attachment:da078e32-5207-477d-8465-da627dad562c.png)","metadata":{},"attachments":{"da078e32-5207-477d-8465-da627dad562c.png":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAsIAAADqCAIAAAAnELT8AAAgAElEQVR4nOy9V3BcV5rneb3JzJv3pjdIg0wg4S0BkiDovUj5Upnu6p42E9EzMbsTuxG7EROxMfvQERs7szsPOxHT0dM93dXd1dVVXaqSRImk6EAQhCG89x7ITCCR3pubee0+kFJRKhlKTalUqvw9JnDPuXnymP/5vu98B5RlGShTpkyZMmXKlPniQL/pFyhTpkyZMmXK/LZSlhFlypQpU6ZMmS9JWUaUKVOmTJkyZb4kZRlRpkyZMmXKlPmSlGVEmTJlypQpU+ZLUpYRZcqUKVOmTJkvSVlGlClTpkyZMmW+JGUZUaZMmTJlypT5kpRlRJkyZcqUKVPmS1KWEWXKlClTpkyZL0lZRpQpU6ZMmTJlviTIcylFFkvZZDyeEzBKo1PCbCYRywi4WmPQUgQCgWIpHYvGC4BKb9ArwUI8EommCtLjR0EQIdUanV6rwiGJzSQS6SKoYPRaNYGAklDMpZKpvICqaAoV88l4Klfkn74DBMSUtE6vp5UEUhZEHyDx+XQ0EksVOEipMRgNWgUKPvOzIlcqcoIM4wSOItAzP1cGAGSJZzOpeDyZ5WRcrTXoNBSBPGMLypLAFVhOhjCSxMrt/hhZKOXSsWgizQoIpTEY9Az5rO0JSCLHspwEIASJofC3f3KQxVI+nUgkUzkeUdB6g06twOBn7XwiXyqxnIRgBPnMPbZMmQ95PjJCykdXR+/eXkhbDp290qjYHbt/fSSkbTn96gtd9WYKKUZWHrx7awNqffm1l5uw7aHrv3z34ToLAAAAgDCqstV3nLp46Wi9GQzM9vWMeNG6My9d7KykYS4bWhl+MLSSM7YdbdUV1h7c7Z/fiQlPVYxaG7svv/rqiUY7jZe7PwAAACDzmf2Fweu3x70lzN7edepEN6NA4Wd7VuJyEe/meiCHm6rrXWaGfMbnygCALHKZ4OrIYO/wShzVeA53nzzcVvOsk7JYTOyvz2/FEX1Va62NfvbV8luMLBYTO9MDt3tnDySlq+PYyWOH1ST1bD1SKmVC20tbIU5ha6x1GtXf+slBYhPeheHe4XlfCnXWdx4/ebjehsLPpCMknk1419Y2IpC9vqHRqSmP+TJflOdkjeAyBxtzA8ORGrq+2077Vqbuvzcqb+QYi8VM1+m5TGB5qn8MxjrPX6knwmtTA+/3bKgsZiOFAbIg7Hq98RJKKi85E97l8UfzJOA5cfIQQENCMRXYmB8bibsopxMDU5GDfd+2L56MRtIlWKk16WhG1tVkWF4s31L6BIGNb8/3vvWLn9/bJZu6bB0oBsulTCwRiqYKAkrrDEY9rcBgUBY5NpOMhWOpfEnGFIzeqNMogMz+cv/d233LWfOh8zB+uEot55J5EVfrTVolzKWi4UgeoHQGHS4ko9EUBxIIyJdknDFaTEqQzcSDkXimCFFao9moUeG/msRkiS+ko5F4mgMJApZKJR5SaPSMEuQyiUSmBCk1Op2WIlFI5ouZeDQaTeY4GVXQOpNBRytgPheNRJMcpCJRIZ9JcbBab7Lq1CQKyUIxm4yGoolsUUKVtN5gevz1vu5mlyWhkPAvjN5569r7i1nz4ZM1MAoDfC4eTEQSOR4itXqDXqMiEEgS2FwyFo0ls6wIk2qdwailEC66MX7/nWsTEbrxEqYgag1gPhFjIY3RbKJxMRcLBhMlXGs20VAuGg7nBAxDAYGHFTqTQYOK2VQ8HEsXRZzRGwx6NYn+ypgh8Ww6EYtlSiCCwjJf4kGS1mgplE8nYglWVtAGk45WYDAglnKpcDiayBZEgKB0RotJQ2FyIREOhrIySZEol0vnJYIxmgyMCocAoZRPJ8LRRIYVYFKt1en1GhX+PK0oUikbWpvuefvtXw4F6dbuysMICkpsKpIIxzNFCWd0BoNOTaIQKIulfDIRi8TTLAcSKo3eqFHBOf/swM1rj3Zke7eI4m0OksskCrJSazAxJFBMRULxAqw2mhiMTYZCaR7FMYAryoTWYjaQQD4dD0USuRJC6wxGE6N8uj/JYimXikSSeRHCMVAs8SKq1OpoQmZT8WSeRyidXqdR4bDMFzLRaDSeyvMyptLoTUatmgRLqVgolBZxpQITsqksj6mNZpOOImBA4thMMhKOJfMciKsYncGoo0j0GTuyxGWDO3O9t6+/NeJHze2mWhiFpVI2Fo7HExmBoHQmi45WYBAIyEIxEw+HIskcL6FKRm80abFSaHns9i/fH01oj72E0WSVCiikMyWE0hu0SoTLxiPJnIgxBgaXCvFYuiiDMATIEE5rdRoFzOcS0VgizUEkozd/vA/IIpsKBiMpHqOVCJfPFSGlwain4GIyHE0WQKXBaNKrSRjgi7lkLBJNZksSTNI6o0HPkBCfS0WiaRZECATgcywPKRijXscoMEjmi9lENBZLZksSoqB1RqOWVmIw+KydT5b4Yi6TiEST2QIP4Upaq9NpaCWOQIDIFdLxWLIAkIxOzygxGBCK2UQ0nuERSq0EuUw8nmY54enlBlJqjCadspSOhOM5QZIBAABAEFHQOqPZSCtQUCikwpF4FlAYTUatAgUlgc0mIvFMCVHpGQVQSERiqXxJfKpIGFcxBpNR+5Gh/I3m+ciIXweSMgfLk0N9LU0u01H9x1Z5ECJdhy7+4PsvN+uBfGhl9G7PyuLy5vEjZsWnFoczzrqzv6+tPre/PTd878ZkSFl3+tWzhzxOm62yyqh8dqv9txuZDS303r72Xv9KDKqgK5PJoliIrC6M3b/9aOWgoKo9dPLCxVMdHhPBx72Ljx4+uD+6dJCSaWvTsXPnT7eowzMD12/eGlxN6wN5lIJP0OmV4Y28tePiqyfrFNH5hzdurktNZ69crMgM3bk9sM1qFUA2p649/+p3zxpTC8P3rvfNbidQd8vxS5fPH220a4gnxmSplN5f6Xvv7pivQOkUQjKRU9iaDre40Pj61Ox6FHMdPn3xwtF6C5HdWx7p6R0YXfCnRaWl/tjFyxdOtxiLW4Pvv9+/U7LoFYWD3fU0XnXo7GuXT7ZXa/jA2uiD27eGF3xxTmWr6zj1wpUT7TUW6uvezct8IbI9fvfund7RjRINWWPZdCEf807OD/f3TXvziLXt2Onzp4/UWUk2tDbe39s/MrcbKRIGz5FzV84frsiv9N+6frN3IY5uFghKLdVml/rv+xXHX/rOa4dN+d2J628OhMynvv9GJ77Y8/Z7k2GcVsE8YGm7+PK5JjgwNvDg3uhavKhpOHrqwuXuFreRQqEnr5ULr4z13B7ZYmGCEHLxPGJrbD/kVmVWZsbmD4SKtvMvXjjV7qSEyPrEwI17Q3NbIQ7SVrafe+31C8dqicBc7y/fmowxFWZl2rsVFI0Nxy9eOnu4xgCnNyeHHt7tm96OsLjJ037s9IXTnfU27fMyo8iSkNmf6blz4/3BtTTh0kVSqWIpc7A4MdJ7b2wzLmgbDp+8cO54i1uPseGtuf6+vocTa5EcpHW0nbxwsk2fnr136+a9YS9viYGkGmyHA9MDO2LdySuvdruhvfl77/TtkC0vvX7c4B9855eP9lCaAnIFuvHS9185po5ND95/f3B+P6Go6Thx4eqpztoKGv/AKyKwiZ25ezf65iK8mkaKiUxJ42o/XG8RDhYn570ZdW33uQvnDrmo0v7CyJ2egYnVvbyktjWeuPrS+RMtmvTq0LVfjAQIg5nO+9a9Oar62OUXL3U32RSsb37kwfv3xtf2M5DW2Xjk1KVzx1rdRuWztKfEpfbnRvpu9o4s+or6kjWaKhUzodWl6Z6Ho/NbBV3loXMvXzjV4TEr5ZRvYfDu+/fGVvazotJW33X+6rkqePXO+9feurkqGSKcUo2ccgLbkzMhquX8i+fqVdGVoeuDawXzkctHrNxW373RzTQHISip8xy/cLLVyHsn+/uGp1ZTqKm++8qFM0cb7RrFh/5lmQ2v9rz93qAfdlsVmcB+ArF3n+2qQqILDx/NRVBn96WXr3TVG+TQ8vjD+73Di96ERJjqDp+5+MLZRkNpe/LW+8PreYhRShl/JIdWtJ47e/Zkq4sWQutT93v6x+a3EpzS4uk8d/ncic5qE4U/g+qSJbGYCu0sTY6PDo0v74bzKGP1tHQe7T7aWltpVUmp/fn+3pEdsbLrwoWuWpNSLkS3Rm8/mEsyLR0eJDTf2zfui+ekp0rEq7tffu2cJzJ5672+pQwvAQAAgBBprWk/dfXFU4c8Wj6wdO9G76JU/fLrL53xaCE+H9mcuP5gOkw3XTzqArYGb/aMbYYLTxWpMNcde+G1l860u7Tkb8fC9lXJCESj1Qml4Pxw/3i17YRSkj+qJECUUKg1Op0OwEtqFY7CggzIwKc3GYgqtLaaaletmHQS2c0xP8jUd564cLnDpvr2uz2/CLJQYlm2yIkiVMxl0smobz621Hu3fz7AgwCSy+d5lFBQZ+qJRMi3tbXtDwTjyZzXG81wJE22k2yRLZW4EssW8gU2l8xvL4xOpKp1beeOVMPpvfXZwTER8hzpUsbW5x+9PxSzWs06c6tDYOObEz3Xrt+Z9HGyDOUymRKEEcwrJ9zqx9OgJBRTB+vjw71TUaa6gpazkdz82u5GjQkrBHbWdzNrcUFptpqUpkzQv7614w2GIqlsbucgISn1DksXGd5aGLk9EDLbbHqUjcSyOzvxEqCgNa3yTN/1t94fjgokgeckbH1jp7HOU2n++mUEIIl8iS0WS5zAFwu5TDLsX8nPTT8cnN2OCRAUyhVLMqlSHHdDob3NzZ3dvVAonspte+McyjCXzFmWLXJcqSTk8/lCMRfbXZ7sX1brOy/wkp6NeedHH+24bWcv1yv8y+N9d1YwW7XL7mbqC4m9ufUHN+8+WgkXQOkgW+JFhUJJnWi2qmAAAABZKmWDO/ODDx9FAY1Dg+QS2cWVrW23RVEIb2/uBGYPeMLgcOhccsS7vb7l3YvEErnUztYuS+ocVY6a1N7qxIP3l0hXnZsSktHkui+YFhEMP6YPjdy9c/PBQhrGFSoe2d6y1dZVV5o1z88bI0tCiS2wRU4UQDaXSceCWzPehQe9IyshAQQD2UJBxBVKVZcTiAa8W1s7+8FgIpnb3Y0XRBQ+pkvn2RLHcaVCIV9gc4nkxtyjKV6oPHpJsEPJveXJ0TlK0XmuFQ2sTw/cnRCNVXaTtc3D5SOryw/eu/lg9iAHCkiOLXI4qaDpI071E1O/xOdje8sjQ73rWb3bRHLpWGlpc2fVrQXS/u1NH7uVgegKs7pS3N/d2tr1hyKRTGp7ezcLqcyVjjbxYGt26O44Z/BUaYBMLJrd3otzEE69WJuf7Ln17u3xOIyTShYgNze9dTXVNsMzyQhAloRSiWVLHM+V2Gwum4psze0vjwyMrh3kC3IslsrDCE5rLjeSib3tjeUN30E4kcvtev1BFle+1A6ybLHEcXKxkMvnsqlQfHV6YlurbDx1vApOBTdnpiYzVaa2aozdnR94OBziNTZ3XZu1mA6vexf6bvfN+qM5GQ5FCqBMqDXqTpeWeLKey3w2uDI9cGO82NhQqeZj+8G5Pb+3wUpkdlaX/enVkMyYLIZjVHh3c2t9Zz8USWbzuweJAqq36o4yYd/CUP/APmesNJBcPp1e96cLEkFcrJEWB+/fuj8ZyHAYQfHg5mZ1bX2dw6B6BoOEJBRi3ukH19+5M7p5kBZ4QYSTiVh0fzNwELry8tUOC5/c31iYnBVKtkPHBRkAJC4X3V2amQiZVFa9JZ9LJxOxYDAcOgjmANpgsWqVakO2UMzGdhbGHgwuYwarhcEhWSiFAv4Ej1M6fbc2ebAyOzYkCu1nL0oAAIpcNuKdn57wGhRNdSYtm0un4uFAKBGJJAqgUmc0akyEJcdywscXzW8wX5WMgDX2ukoLlYouDI9MGVtYXnqqSWSJ3Z2+/4/J9ZsEJJVKgoQ4j9dVVdB46it6m98ZQNLcfOrUyUczu/NI16vfe+VkRaTv2vCkT6xs6nArC/s+78LMpLuh0X1IZ288cUnhao9HfCvDD2f3fbuB1OHTNe3dhzYSSLH23JVLnXXEqvfTx6UMUZaGM2+8erqj3kmG7r39YGAtb2/uatSLwe2VpelRR3NHi0OtRZ96OZSgnc0nLx+x5OZ7+qb2oqWqC2dfbXNPDg55o7sHsZzo0Nnrj18hqo4k4r7loZ5Ha/6tzUCi1SnKgAwqtM5DF149VgnvT771y4fDo5PHjtvsseBeTFRVdpzs8NiMZlulx6l/tqn3+QKhCoOr/eSxzsU9Nm058/IL3dXA0p2pue2MsrrZo5MiOwfbM7PztdW2DnPN0fOovTmZCKyMPuxdCm5tRs8117V3H5ndEBPW41evnGioSHk/dWMFgrDCUH3k6u9d6aq34OmV6yOTi2HI1XjIhuX83sDc+Jynrr7KpHo6iAAk1KbKrqsnKrnV4TsPd/dSrhcuv9LRvtBzYzHiDUTSXJXV5Dl09g1DUzYb3RjvvX7Xt7bqjeTcAAAAEEZbG069eq5OlVzsv9c7PzM50+Q8xMfiaVZla+hob3bZzBUOd5WZeY7ReSCIqitaT53snlyN7FDHX3ntYgftu//e+GIIdjW02rCUf29vbmbaU1dbUWdyt5y9StcdiQe35wcfLEa9+6njdTUdh9s2cmqs4eKV021V8vjcZ+w2UNredvEPXjrZ4tYBvltDw2O7JWdLu0fNB7Z9C+OTroamBhtFP/2LgJhSW3Xo3LkmVXji3uDSXhKoPXe5s8X5qH8kEvSGUnyNs7K5+yXafSIV25kbutPj29nwBlN1OlkGAEShrz5y5YVOq7A5eOPtvqmRsSNtOioaiaVRg+fI4bZqm8lsd9XYdM8aGQ1iTEXzoaPHptfiCrj9witXj+gDE7ceLYZRa80xuzLvX1+YmRiube+obmacjceu4s4TuXRwdehuz8Tqlv/0iUvHj3cteHnWeeG1iydb1P7Bz6iKZKx1Z1978UJXvVH2D/ZMznkLSldLkx6IrG+tzkxMtTXVWGlc8XRPkGHaUn/y1cv2VP/P3+lZ3DV5Xn/hjzrqR3t6Rvd3vTH2hK2i4egF3NqSiu+vTT4cXvev7fgjjaQMACCq1NoOnb/U5UaCEwM94xvTU2vVFBM6SLACXdVS11TrtFjtLk+lQYlBn+/UkEUuG1yZ7L8/vskaOl//7tFqIwnkQ2vzo4OrE/3DJpfljPNTV25YaWo6VOVq7E4drDy4ee36mtx64dU3uj0GU4WzAvEtQgBmbzn7xg/f6DCjXGxt5Pq1saXlzb1kh+ZT3wfC1dbGk68xrqMHu8vD9+6NeIHqE5cvdTVWVlhsTguF/9Zskb8qGQFgtLX52CF+vXd8ZeiRqAlnBUn71J8lUeQ5rgQhCGaqaT164cUzLU6GzyIoDH8sxAeEQBSB4d+BYOvnAUpZ3G63w6jcRp2NrQ32wu7Bvm97t8SJYgznk/FESUv79yOZJg2GoyqNHiZQIa5WIVKCLUkobbA5K6ymqOj21NXYjamdz1gYMLXF037i3OWTVcrC6uKu37/jzXDAVFYpZeLhGAz7vIFUsU771FSIkKqKqtYTZ87aItLqijeFuGoPnzyjsXLBldi+IAiCJCMgqiIYg5kkSjGtAuZyhSxb5EUZAFDK7Gw7cf6FCx7UT2xMjs0GAoEk0F7T2t4ZXsrx6VSSxBWaHMtxoiTJwNccHgFCmFJb4aq0WYwM4qhqqLVTG8P7e/4dH1IqpBVyNpoqqZGdQCTdpFWplLTBTJJAQqfGxRCX5SHKYHM6zLodvLK2vt5lglY//fVBBNG7ao+ePn+6hcqtDwzs+Lzbu4USnwkipXgkD8UN+3uxXKed+tWAgQnabG/rOnO2NouExqeSYkVt2+mz7iTmm91bBSSBFwEABnGlUgvRDJ7aYTDIzxbybEkkARCCtM7arrMvXG1VRjU57/zSZCSa4pXV9e1Nft4nFLOpRJJUaQrFEi9JMvAZFsUv2KAEbXW7K20mKs64Gho95uhSYN+/vSOWuGIQYeOJLJQz7wVi+Wo7QmCUzoASYCFIKeAoz8k4bTLYrAYzq66urauq0Aexz/IxMzZ35+mL57tMQGh6ete/u+1NcUApQQipSLKAKfy74XSpmlY8VQKmpJ0Nh06e6VZuZ+fXQoK+uuHomU5EnfStZnMCz0sACGMKBWNUqIjCgQZHdkpsni2JkgwAgLrC3XH60gsn7aJT9I4P34qEIlnEXdfe2ppeL3L5VCJJkHSuWOTEZ2xPECYYS4XDaTVqMmRVU73Hkt1IBn17XjCVZYM4nwp5wwnLmj+Wa3CiKM5odAotDSe1CkwOlziRNFRWOm2mlYzD01TjthbCn1ElodbbWo6ePnGshU5MrsQCPv/+AZjPxJRALhKM4FjVbiTd7dF/REZg6oq6zlMXX6jcjw0P9O0i5qYjp18w+0ve+eVVgRNECYRJQqnTmxSknNyhEDCdZYssJ0gygKu15pYjZy5eacZ24fja3MNoKMFCLnddY3NkJSmV0vE0oWSMxSIviIAsA5+jJCSejfsWFzd2JceFS99542q7VYUCXKrBqZKKb43tr23vt+jJT21lnDK7G4xNMB9Uh5dH+6JyZfOR0+c7LSoU5PwHEAgAIAyjKIbjGIChCAxBIAh+5huBCKE2u9w2V13arkxvLW0WwKrmI6fOHXHQX9m6/NXw1b0ujDOu9npLPvJu3/SjtWggraA/UHogRLo6Lv6rP/xOmwmDYaXGaK2wG9RoKYorFDAKsoVCocgJEihwJZYVZZlQ4ARaPtH5JXnc6iAEYypDhY6xVdlUUtI7uTw+sM3ZGus8tEbPqJV5DEU+MbJblmVJkkSuWGTZIi8hH/yIMIoTJIEhHxkoIAghpMZcbdF53L9mFoAgCMcxHENhCAJhFCEIgiAwDENQFIIhQGDj3unx0anNkr69oZLWMhqaFAgMQR7vM0RB5DhekORfdVlYaW09+z1YU7fpC0Uj/s2Ra1Pr/izKvHbEqfrmxJuDEKbQ2RjK5rKo+YPlka213by6qsWt1jKMmkyhGIJAn7TKiZIkSo9N+2xR+MDACYIgRmAEjkIfqwQh1GaHxWB3Whnio18eghGcwIknUxuKYSRB4BiGYwiGgBAICLnA4uKNB8sZbc3ZDqNer2EUJIpgyJPeIAk8z/Oi9KuqUJWpuvbyK2rr2m4gGPJ7p+8vLvvSIkGfajR99c5cEEQI2mw3G9yVFgUf3hhZmJo4gKqaPBWMRktToTyGIPAn7E0lSRZFSeA4lmVLgqz6cN+J4hhBYDAAPB3mBsKoUl+hVevcTp3i450JhmEcR3EMQSAYRDCUIAkCxyAMwRAIAsV8ZH1yZnIlAFu72is0Oo1GRRZQDIGhJx4+ged4QfpQIYAwzthbL7yCG5a39kLRA9/yw4XZjQiLMlc6bMp/4bQHoUq9o1Fvq7WqoezexKMb97ekqs4TrRqdgaZIFPnV7/wUsiyJkiTyPMcW2RIvfPjbwwhM4CiGfKx5IYTU2qq0+mq7TvlrUYEwRpAkgSEQBEEwhBOEgiQwBMExBIVkoZjYW5obH1rJMfVNtYxGq1EpYxgKIzDEA4AsSbzA84IofWjUBDHa3nZWy5gqt/3BYMC/2r80vx4s4tSVDoca+8xRL8tiMRMLRnNFhafCaTNSGAwCAK42WJ1uq3oumkwmsyz6WSV8Ntz+/MOfJbZuk5DMCZIMO4831jk0BJj58kX+tvAVqh4QJvXVdadP7W6uLt/e9edqmz8cuCDCWKvajp0+48B+Fc0vihil0TAYt7wy9vAeEbeRbHBlci2Qh90MQ5UTQ3xhQBBWMGa71eKMCNUtnbUmJYwSjL2h0UnzB1Pjo+PrcgZFEToZjedFyYDCj5czgU0HdzbXNupJlQShSDERXJ181JsA96ZWfemi/detfiCuNtstZnuFwt3c2WxVwTCiMNW01ds1X+icnVCI763Ojo+sy24FKmkyoXRBAGEMfTz98unA+kR/j0P2kqHp+d2S1mZ3aOBcLJbO8oSC1tJsHGDD3u1NXyTLSQDwG5UREErSWpPdYjmAqNrGRoeGgEna6mm0qdIrE8OjS7yFUwAmNhjPcJISReHH07KUjfnXV1ebLE4BgVA4FdqZGX4Aa0JjszvhPFH/sTpACEYpjdFhqXBmJU9bY5VBgSBKvaO5vurDEMtng88GV2fHH06WaiQbk+cOovmirIdRFIEAQJbE2PbyUM8tRUSfW55ajwJ0u1lHivlEKl8CcRVNqxQQmwwdbPqC8YLw/MwRHwGCURVjdljMzhziaWmv0pMIrtRXNtdYyfzGysjoZEjB4bCAHiRSBQmBURiGIBCQ8sngzvratrsRBkAEycb2FiYGqIPCytTGXopnPt6RQRAmaaPdXOGsYKpa2uotKhjCaUtdc72V+YIdObK1MT0wEtQ2aKkSEQxncyIKoigCgSAIAEn/xljvbV3RDe8ML+zzqpoKkxooxBOZgoSpaLqYxYVcZHdjZz+S47+McxzCVBqL0WI3pzFHTbNLr0BhROVo6XRreP/d8ZHedZJnrGomEkzkOdgC4wgMAbLI5cP+zaXNXZIREVjk0v616aEHWWxvcmkvnmV+rRIQIRmdwVxhMePqak+1jSERTGGp66i3afAv4NuShGIysDkxNjZbtEEkasqG4nlOVKIwCkNFEGDT0eDMSK8R3icj00v7HGq06lUwl81kWBAmGbWmgIO50M7mti+a5Z7tuN6TqI0PdlfAY80kSpIswwAEwQiCohAkffQhGIYwFP4kbfrRsmEYwTAMg2HS4mw6fOGVi22VDJYNoASMoOJH/hMCUQxFUfi35CTG5/HVGk9gQlfdefzMwsLc2u7K5zQYCKEKg7P+SPPcwcO5Gz+evyEDiAyhanNNd1N9lUGFfTsa/OsERFWmmqNdR7djw2trE8E1WFNR025pJSg1RWtpDI1Oj98M+80Un2R5XCxxnASp1bQCym6NDd7V25jLZkh0wTsAACAASURBVJOzip4dG7720w0Lg2b2ErzC9gn14Fr3ke6TKwcPFjenHq7LlNHZeLoBVyqJL9S7YJQkNUaiNDPTfy+wYaLleA4iS1ypxIuSDAAQlw4t9f9y8348keVwz7nuo80WNLkwP3Tr7ujmQTxbEgAE07d6XCb6N+5ThDCVqbK5q3MlPLC0NT/iRUmjp4OpUagYVKdViunJkfvxHZMGyhRAXBQ4jgMwFa1RlQ5mHt41myt+0GFxVlVBG/Pv/mNk1qqSIsmSaPm1SkCYYCpqjh5p9ycntpYn9yFM56g/7GzHlcSz5gl5DIypNDoLlh4Zu/V2wG4gC0UQ4PlikRMJAJBlIR/dHLsVGc0mUhxuPnTlSLtTVQqNjz3o6Z/2RVMFQQZJQ43bYdV/ZQemQJhgKuqPHjnkT01vLY3tw6jO2XjY2UFSaoVGp4bkqZmh6/tbBrKQLokaocSJkFbFUGJidqTvjsGmOa61Vrrxxbmet/5h1aKGUuFcSf3r8WsQrDBWHek6uhEeWl2fDq/DjKWq1diIq8gvdoQYQhUUYybY5bkHN0NrJkrICJJaKBWfBM3JXMo31/uLxUIynhOphovHu2r0QGx6evBWz+hOOJkTJAChKtpdDhOFfZmODKJqS1P74RPb8bGd9dG9VYI2uA9bO1QqFUUbDFp4cObBW8kduwZkJRiTOZaXUCVNk9z+Qu97vTbqUo3ZZtaMPpi4/ZOdSR3GRlIF7BPc+yhldDUf7lg/GFjxzo/uoEqmsv1cvUJJ4F9gaQRBCCGUNENCydmJ2xG/hYGyvASAIvd40IMCm9lbGLy+01fIZkFNXXfn4Voznt+bHLndN77ijeQLvMATlsOVTpMa/1y3NwjChFpj0RPYind3aydQq3cwGFBKhfd2NvbSecyj1WooZZwkACCfz+fZEi+KMldkC6wMQ6QCxz6nBszecvb3/+T3j1YQEEjSOpOlQk8rIK6IkyoEhdOFbL5QEkRC5NhCiRdhBalUYN+SzfHzkREgpjJXNXXn0laXkaKU9tq2k5LQYNcSCKY01xy98ur3itikWF1jUsEwZvC0dl+w2Kr05MfssjDBWOtPvvg6TOsH5jYOYpxCbavrONx9uqvVpX3yzyCq1DsajhzRKast5exInwSI0famzjMyUm+lcZx2tJ9+tQRR94aX9pKixVVV63ZYdHqaOnzxhbRALkVAQ1W1SQVmWcJhUCpUek/LkQv7mdmkUkcxhqrqky/mWXJ0LYNZ7RXK5vqAZGqy6SgGrW09ejpvd2pJCAQAUGFqPPHaDwClsn9uK1jQO6vqPJUVDPnh5AshBGOtPdTJl2wOnZJU6h2NbYcprNpKE4RSb/ccOsQwlRazzcScuxwV8bmgZK7yOJBCPE9U6hQIlAAAVG2tajnUoC+Ft7NE1aHzr19qr9Qqc5W1Tc2+ICuBsRJqcLccu3D5mMf4G+kXIIzTppr2zkza5DJQSoZuPH7xVQnTDExvJXnS5a6qdlTY7CR9/MU4S075eI2jtpLiMkXEYqYwkrE3dJ09m4B2EYOaZCzOk5feSAkDswegzeOxIZ7qKFbrNqiUuKO+8zSmqDVRKAhAiELvOnzuKo8qesdXQjnE5vbUuCoMvwrNAiGcsrgbj7IZU6WBIpWgvabzGAJ6bFoSJyRzTdshHHaZ9SaH/tRL8Qw+sl3QeFqbjB1N4YLORMKyBIAQpKv0HOpsJRO+kGhsOH75hRNNNo2EV9XWNexFeADLoxp7Y+eJs93Ndua5Sn0QJrWVLZ2nKKXHrCaV+qrD517jUaZ3bDWUg2yu6prKCpPeSOLHLl/Jg/RGhrB6XAwsZEWmQqvWWqhDp06HuVXWolHSppojZ15Nc+qJnZLWXmknG2rjoKPWTKuYitqO00WxqUKNQQAAoWpr65mrRZi4Ozzvj0kml6e22mmmn3ISQahSb2/s7FSh1TZGqcBMrtb2gkNTaaJwAjK56jqLeZvLWWVzomweHFqJ47a6eqvUnsyp7DT+eH1lbO62zg4649szUdXHrrx0rt2hRVKuuvomb7AkwSmQMntau8+d7azSE8/cnrBC46xr61bhtVaGVOo0h06/KkL4wNj8VgI3VNbVVjsNlEpXdfzyK5E8ORMA7W0tNdriQRC06ZUEQ7WeOnshTx2gWjWtc7qOvXA1B9BraUhTaaWh5rykrneatHrO1Xq0kDFVmigcBkBEbW08cYEDUXp0ei0gUBZPTbXDzJBPLYwgSlkaOroLQp2NwSFc42jqPI3gtRYKgwlNhaftKEbUOyprGOJyvITM7PNMVZ2DgYs52WxWogALAKRaV6VrarISqTSLVrSdO3e202NXsHB1bZ0/kmblRAEzOFqOv3D+cLWR/HwLHISQWkdjvatyeWjq9pvK3F6dRQXkg+sLYyM+SdteX2036pC4jlHw61sTfXeI1JpGjm1MzO+lQQfNqD8vjwdMGRwN7V3Hqp5e2CQIVTEmvRra3BzrvcXE7Qoutja+cJCGHBoNTX5LMhWA8vM4VSILxXQ8EskKOK0zquBCKhbNPU72osRgQCikQqFQXCAMJouZAnORYCjKEUaL2Uh9fN6RJYErZJKJSDyVL0oopqR1Op1OrfgwP7AslrJPpZ9SfCHr7e8GQiEeDIayIGOxmDUkJAulbCoWiibyJZlU6wwGA63EIKmUT8cj0QwLEjStQCS2KKJKWqchgWImFo2mSqBSYzQyJFBMx6KxNA8rVSpULOZ4WK3/IP2USBpMRp3ycfiayOUz8WA4likIBKXRm8xa1a/8UE/STyWyEqEz6dQIn4nG00VIqdNrKbiUTsSyHKrS6GkS5DKxaCxZBBRqWgXxhYKAqGk0Mv6jv/jrayH7D/74hy/WqDkZ/cT0UxBBaQ1Gk0b1m/F/yZLAZqORWJrHtWaDnsIAkSukErFIPF2SEEqj1+u0FAEJbDoeiyQKMq5iKFQslniI1Og0KoTPxiKRWE5WaoxmPQVzqUg4Gi8ASppWQcVsXiQ+SD8VKcCM0WyiCRgEAFni2EwqHo6lCjykpLV6g5ZW/OoA/ZP0UzkBV+uMNC5kE5F4AVAyRoMaE/LxSJIFlTqDRokKuXg4Es2KOKPTknw+m5dIg1Heeuu//ef/9xFy+c/+/Z91m+TiJ6WfKvAQrmK0er1WrXiWg/tfpEG5fCwYjrMwYzYbGQKS+CcZ0wo8qKR1eoOeJlFQLGaTsUgix8EKRo2DAluSCTWjUcFcOhaJZHiM0psNakzMx6PRWF4klJQK4QsFAaGN5sfppzIyZbBYNI/jAmWhlEvHQ5FEtigr1Bq9Uc88nWn/cfqpaKoIKnVGjRIqpiOJLI+q9TpGARZS8XheJNQarQrh0rFoNMMhSoZRyqVCQcQ0eihw7+//y3+6l+v4/r/7d+ecKC/gn5B+qgSiCrXWYNB9oQz/EpeLR2OJIqTW6vU0iQJiMZ+KRGKJTBEm1TqjUadW4ojM51Phg3CsAKh0egbnc5kSROmMGkLIxkPhREFWGExGLQWyqXg4luFAnFJiIMfxAPEk/VQ8K2CUwahVEwgIALJYKqQT0VginZdQBWMw6ZiPZEx6kn4qLavMFqMW5eLBUCQPMWazkUZKqVg0XoQorUFLiLlEJJIqSJhaQ2EAVyyBBCGGJt7/67+979V2/t6/utxpUcKfmH5KgAhKazDpGIp4lkxNssRlwxvDPTeu3RnZ8Mc5jhdhDCcZvaXx2MUrL105Uq0TAnODN39xrWdqI1jgZRFEYUxpre+++sb3Xzpeb6VQEOCD0z/77//1v43L5/7o3/9vrz0JsRz4y//n//qLNfu//g//5/9+6SMyApC4TGCu99rP3r47sZkQBUFGEITQmGu7r77x+sunG8wKGACEtHfi3b//+3cXwUNv/OmfvvzbF2L5fGREmTJfGWJ6vfcv/+t/fyfi+qP/6X/+16eqvpypt8wXRopO/eT/+0//ZRh/6X/9j//Ha03lLMn/QuTk8rW//s//971C1x//h//4e0es5LfEMf5VUIwu9779F//jvt944s/+lx++0Gr+1BMUXwxZEorpiG9ldnLi0cTKbjiHMpbq5vbOI4dbaiotDIFIxVTYtzw9NDI+sepNZCCNyd1y7FjXsbY6u16FwSAACMmdobs3b2/KTWeuvnzErSEQUIgt33n37TsB3blXv/tKq/mjE5QslrKRnYWJ0aGBqXV/rETSFk/r4a7jXW0em0GFggAAiIXI1nhf34QfcHWdO9dZrf+1oN5vOL9lqqfM7x4gSpnrO09cTBtqTKqy9enrA8Q1rpbuF3C01UaXI5OeA5jaVt91kSt5XDry60/Y/lsFjNOWqo4T+Uq6puLjJ4/+JYAQQjLm6o7Tpuqm7nSO5SGMpGiGodUkjsAgAMAEbXIfOq1zt55K5zgBxAgVo9Wo1YoP03zDlKXt3HfsXYBSo3t82AOA6cpjL/5RTQnV6LW/ZkECYZwyVnecNbqbT2UKRRHGSIrWaBiKxD4IRoVwxt529hV3ESBpzW8+tOuLU7ZGlPmmI3H5RCKRFTBao2EUn5kCoMzzRCym47EYC1JavY7Cf8s2SN9AxFI2GY/lJZLW6cvXr30msljKZxLJvICQjJZWfZHTH2W+fp6bjJCFQuJgd2Njey8hUyZXbV1lhV6FwwCXTxxsrK5v78cElamytrHBYVR/qrqUhGIm7N9a3vAli7il0lNd7TSqCVjm88mAd2djey8pIFp7dY3badUon/Ximt9RZLGY8K+ubsQgS11jnZ3GHn8siRybjkeCezveQCSZlzAFY3a6nQ6biVESKAQ8DnQJ72xtbPhiMq6vrKl12Y1q4sPgFCEX2Vrd8BcUlXW1boPy065jkiUuF93bWlrbjeZhk6PaU+s20R+EQUlcLr63tRtIQXqX22XTfKv2ZrLAJoPbm94ITzlqPJVGJQIAACCLXDGXih7s+Xf9wVS+BChoY4WrymEz6T96B5NQiO5tLvvzCktVg9ugQiBAKqWDuwtL6/6YqKtw1zZWV+hUn+3YkfhCMuDb3U+BWpvLZdGSiCywqbB/a2PLF+MJvbP2gx/1K26LfzmyJLCJfe/GTgoxVdbXWCgUkAQ2deDbXN3cTwuk1eWprnLoqQ8WGlkosflkZN/v390LpVkOUmpMNle1o8Koo55sKKXHc9LG9n5MoIyumsb6z5iTJLGUi+3vbq7vhrKC0lJZ5fE4DOpv6nXaIpsI7Kz6M7DOUes2MTgCSlw24lte2dgNsZTRWdtU6zCqP9juypJQzCbjoX3f7l4gluERTG2wV1a67BYtpfhIDgaplIn4Nr2hHGaqcjstNPHZA1bIBjdXlvxF2l3f5DYovk2ju8wnAf/5n//5cyhGLMS2Z/ve/dlPfvLPb90amloOFGCVwWxkkOzu5IO3//7v/+nNt28+nFzcSsOM0WYzfKJ7WxaLqf3l/hu/+PGPfvL27Qdja4FUSaE36mmM9S8MXvv5z37801/efjgy50vxuNZkenJnXZlPRBbZyPrwtR//3U8GDjBHfZNbi4OyxOej3vXp4em5td2D0EEoGI6lcrl8IZdOJSJpDsCUFInwKd/C8I1//qef/OTNmwNji96MjGlNJi31OHZKyO3PPXjzR/94fTFNu2trLepPjGySJT4XWn90661/+Jsf//Jmz/CKP5YntAaTQatAQEDiMoGVkfd++tbtRwHI6HT/ZvJXf0XIfCG2NXnvzZ+9c2+N1birq6xqVOILyf2VxZnxudUdX+AgGApHY+l8gS1kY9FEnAUIhUKJoxAo8YVUcH3y/rWf/t2t5aTK0VxrplAp45t/cO2f/urvfvr2zf75dX8CofQWi+nXopM/qF/ki+ng+uzDa2+/e28qAOkq3TY9ISb9S0M33vynn/zslzcfji/4MzKhMxm0FPk8L+V87sgSz6YONqfuvPPWP1+bTRLW2gYbDRXj3oW+9978xx/99NrdhxMboayg1Bv1GppEZD4X8y/Oz0wurO/uBQ6CoUgsni4UivlsJBqPFyGFSkliIJfYGb//7t/96Cdvvn3j4dTSVhJijBU2I/VJPVnictHNiftv/+zH//Dmez39U5sHWUCtNxp16s9ZSL9+JL6YiWzN9t9488c3ZwKQsbbaqiWBfGBl6Oabf/MP//Tmu71TS9thWaGzVlhoHJLFYia0NTM2Nr28fRAOBQ9C0UQ2U2ALuXQymsgJiEKt+uDUhVRK+RcG3/nJtZ6pKG51Vtq0is8asHx2b/7uz370V9eXC/rq1jqz8puWHUEWuWx0e3VxbivGYSpahcOyyOXj3tX54eHJhY1IEcCVKvJXl5XKQjGfCvs3F2cmRsYmpmdWdgKJvAxjGI6jTyeT4HPBrcmRxZ0kSH++tUniC8n99dWFlUBGwijqt+Yyz0/kucRGyHw2tDQ2cHdoPYFVNNSJscDi/R5K77Rr3ZmJh30PlpOYq+0wkvNtj966Z/fUGcX89uJGgWloba5hSoH1pbWgqK+qcxChmaH7fTN+0FDdAOdi3on7vQarkWqBl8dGxxYOAGN1A1GK7a2M9NurKh1mjeJrv4LptwRZZOM74/dvvXv30Yr63LE8JwGAJLCxvdXhRzMrfo7RK3AIQZQmm4HGionA+opX9AcjrAw3V0B70/0PBxdCkrW6FsqFVicekEa7q0KvJlSokD1YHe65defBSLRGn2IFScwFVxfnVtKquta2RoMYXF9Y9LGaqqYaTWpu+N7dsU1eW9ViZmOBmfs9enOF3a6xk2I2tDXdd/dOz6OwsqslXxK+RV41WSjEfUuDPXdvDy7yDY5sUZBlkU0G1mfHRtYOMhClhWQQJLU2mxYtxUK+Jd8utB/PCtCJZgcthBfGhwYGhyeGhudiNuOJgiABcim2PnT32q3JOOns7ASS/sX77yj0JocFIvyrGweSvqG51o7nfEtrvgRqaax1akp7M6P994cGh2c2SnRHS54TJCEfX59+1DO4dCDpPbVgOrQx+OCBxm636tVm1Tf2yJks5ELLj/ruDDwanlrc2dfojxV4UeJykc3Jod6BpRBmrq4DMtHN0Z4+o9VqMSqp/N7q9OjIRrSIUlpAAiGF3uHUQPlQcNvr3UECybyMnGjUsttzD+8+Wo7j7kOdcCaw1d971+KpNYv5raXNAt3Q2uJhuP2lpfV9QV9XW4HtzQ4ODi/EyYo6G1yI70w/7K9wVNrNtB77TTfQU8h8LrwyPTo4MDQ+NDwd1CCNOU6U5VJye6zv+u3xA9nU3mHN+td737qhtXhqrfVKNro5Njw8sVGgdbQaFUFSbdJbGDC+51vybSoNwbQAom01ZgUCSKV0YG3iwd17veMZI3Y4XywkA2Hv9n6esFbXuLRyeGd1OcTrK2ubXCYKg4TMwfxgz3t3Ho4l6yrTxW/a6JYlLhfbW5/ou353eAVoePUPdHazCijEd2b63nn73feHtgqI7ejlV777vctd9VY1BolcNrqzury6G8wUcrl0OlvgBUSVZfOZ8J7G5K5t8Nh1CgSSSrnEwerA7Xd/9o7XfPaHJo9d8+leQFnii+nw1uLorfcGFsKaE9/7ntHy2crsm85zkREimzzY2vAn0OqTr790sYafuvbmtfVUNJUrFbKxRF4wt7/0J987rdr8+V+9OZ5IZAps2jd39+cT3OHcn363Lvvw3Z8PRBwv/MDMqHxrW0FW3/md77/eQW7fv3FjPBeNZXI5LJMuwpr681evHDPFHl17sJTNZvPPmLXsdxBZyIVWBgYGp/xFSEHDTz4rxvaWHs0u+zmtQ4eGFocHx9YTEG3RKSRRpqpam/WljH9hDU0j+xubYcnc8cp3L7Ziu3f++e5CJpXMFAVJLCa8k0Mjo2txFlPCj/M3y0I+sHT/FwOp5sy/pTr4wXd/fGvXeOkP7VbJv7bqSypbXv7h75/WHfRff6s3FImmWY5ni4GV6dGxJV8WQEjy2xXhK3HZ4M7c0NjsWlSCiMdpjAU2GdxYGF05YAmtGc1sjDwanQ/IWqOBkPKAuqqz3S0n91ZXdvRUFRDwbvvjEsnQGlUUggAAAGQxE1xdWNkR7Bf+9N/8oEWe/Pnf/nggHoxmcvnM+ui9W+vk8ddfO2UK9r59b5WreclUoYMzezuBWBZQmnREEpYAAACkYiays+kLSxWHX7l6pQVbv3vtnYVcLJEtCtJnfZ3fLLLEZ8M7275AHjcatJnY40+FYjq8u7YTFS3Hvv+dqw3Q8p2bdxdzsUQ2n45GNudG16OSSmcAYitDQ+PLUcRoNiDFNKRvONbuECI7S0s2Q6sik0ykQHPLlT/942PK5et/+bezyUi2UEx7Z3p/Pswf/i76ncZc/ztvDQYtV/7EajDm0okcYmi+9AevdxLbt9/uWUslUuyXSi/5FSJzqcjeri/CwUq1hgojkAwAgCzlIpsra9sF49Hv/+EPj6k33/2bv7ieCEYyrFDIrk2Mjq9l6EqXNrcy2tM74+cVZoeBYHNwRWNbO8oHF1fXKR3ToJXi3sXJscm1QB5CCQIGAFkqpfxzA9dncvZTVy55hLk7t/ujulPfMVc7jIpifHt8eGBsI1FCqM+/cPPrRhaLcf/SUN+DwcGR6cW9TIUxWxQkmc9HdmaGBsdXEyp7pYnL78486nc67BU6JSOGNxYnhqa8WYnUUaIMwyqL3awWksG9TT8Pbh2kSyDeVa/hdycG7vY/GpqemF0VOtryvCTLYjHhX1tY3uf1ntZGhyK/t7i8FYFMTY1VBv5gZvDhvUfDo1O7eaSxOc9908TWF+W5zOMgTGqrWo+/6NE1dze5EO8yoTUYdDaDRq1T1da6J8OB5dG+Ehrd5zU1LTWVRmOForXBM3tzeegm6eUXttLK1sZGt5nhsw1HLmkIz4nWajoaVqvUJpXJZNAbCHuVw+jf8M4Pl4jiQRY31Vbanu16+d9FxNzB0lDPwErR3HwShca9AAAAMs/G9rdmN+OCxqmDIqNzG0HZUFWv4KPelT1W09lZ2eRSZvY2glGlVVPbfc5KVR1p86hCMVKj0WFmg0aB8MmdiZGBiRBmbznEyYsiCgAAAKnM9S0t9RO/2Bi5eSsMzK5E8YbzLdUWGirWdF5AQNfJQ7WWXJZSKY1mk1mnkDO+ldlH0/681tNyiApn8d9sUz1XJC4b3pwdHV2OkpXNh1HvAfI4TiKwvbEd5gmbVk6uLq6GS9pqjxrM+NZ8SXXzEUddCx3b8vv9YY/NZmg4cq7yUHJRjMz5s086N6K2t595tUp/8liTU7G3iqkNFpXNZtDpiYb6qpn5ifkHt/O67GqYNx6v9zh0apxwHzqpcXpWpu9HxoKPy4BxxtnQ9YKZrDncVk2F9xUanQ6rMGqUXyw749cMiCj1NV3nmeZMYPC9+GbgyYck42g+dtmmqutucRP7PpqizQqjkQKy+5sb3risrsS58PziekI21XpUYmJn1ZcxmM9UNrYi3uU9rzfsOWSsrK01hQ/WR/vyiH9P0LrqGxwmYwXZ0lQ1cm/54R3cV1zcTqs6LjZ4LFpYcNdUGkP+9ZH+DJk+ENV2j9uhV37D9C+IKHSe9pOm+uQaWtjcDz7xGSMqa1P3VaOy/UR7tSkZRCm9hXLYtWQpMjW27M8o6lupxOTw7HpO767XINn91fVQqar+osfjgrwTweB2MF4hJRanHy2EeGNtS6s6xmMAAOO0yV3rNswOb4wO8musfyvHNHc1OI0UwsXXp/r65yIK19GOArcgf9O6l8SX0gc+30FCUBmM6vTjnajMF+KB3XVvBnefev31U47U7M0bk4H1HX+4Uc9GFidndjKI3qJiA/MDA5MbKdxmY6RUEatoO9vFcN6tOYXZ2AF7t3d3M5jZbLQoQo/rkh+bhO+8twi3f+eHF+yRobffm87XXDZXWsncwe5+LAeqLXo6Vvjmxyd9Ps9lOMCkvrrriq1DBqTcweyDyZkw5mw71Oo2qhnO6TIqbt+9+ePbIi8rq8/8odthphVqbfPZS8fX/uH9h2+t0I6WM2+c766zaAig7YKhjpfEYnR9YmLaX9A2dnU02Q0aIWvWKHNrvfdvxFjS4D7+2imjSVvOPPWJiLmDlYcPJjdA+5mLLdCjwKwPeDx8gv7tPV4wUGJ8Z2sfqOj+4WuvtCn3H73/0ztrqM3uspk4b0xIwLi5sev4CRQQ2cj64MT8Hqht6GyqseB539Sj0ZWksvL0Ud02610/AEEAAP5/9t7zOY4zz/NMn1neWxQKBe8BkqD3ZIukpFYbzXSre6ane2Zvd2fmbjfuLuIi7u/Y27m4uIndm+7pnZmeNqNuqSWSkuid6EASIEB4VAGoKqC895WZ9+InpkpAgQThCIDP5wWjkJnMeqoq88nv87M4pajpOv3uyWf/7aOr//JUbus49ifnjnfaNXK84+T3Go4IYik+/ejulxMxVVvfvm4zFpx49OUzf8nStdfGj95Nju6Yn1AUiumFqcFbD6biuua9nUrf1XAwgYl8PjHv88/EKblFiPndvqJlz7e/985BS2rgD//y6ZOCq7bGYZMXZhciqXyR0LiabQ4xNelTsASexjAMw3BSXXfwOz/eJRBkKTR06er9caFh36HDHVaNkm7qO3xs0v2vl2+eH9K0HDh7+sSuBqOKwxUNXfp8RBae4p7X7iU5vbPvLUs3Lwi54PDNhw/nREvHrl3NdvVWzr7ACUpla+kx1iY9hYdSDColM9TvPWfrLvPlzMLQ3YcDC2VLV/euJm05ODU/l+L0qkLINx0Q6w69//2ze7She7/+5YVR2uWsqbGJ4bGFZDLHs1Z7g40+f/GP/3ApWy5om0/8eWOjWS3X6HYdO3dy8ufnb/zuGVvXferbZw+02lR0Vm91mMnoZ+evfJblVKa2Uz962/GNupZbApzRWBuUZjHtid6RPQ9FwAlVzZ6zf9LB4yQfm7hx7dZQwbn31LFuMxZ8/HQ8nlG1cZnZu8MJVfd3f/LTo/bc8Pmf/8vVeUtNvHbJfwAAIABJREFUg8NuzISxuUI87J+IDt8dD2M1PX261MCd+7MEhmEEq3V2Hzg+Pf/hJzevDAmGzpNvHe9rr1GJsYl7Nx8MplQ9p/cYRyOPRl7UK/S1gJOM1t5+/Ky5NzDxeSQ0n8QwDBNL+VQ0GswR8ub61rZWZyzxWDMYjMQi4eCsf2LSnWGdtUR6ZvCxJy+v3VXLFnzjA+OFeqe5ubcrc3tg7pkn0Nnl2H3iB71C8v5vZ28G4b0IktXXtva02Z9cGrryccmcm54vmI50ddSb1DJZfd+31K6g+/EfP705ttW+pNWwPqqaoBi5AkuHph9e+eKTm7NU/Z63Tu1qNiuE+OzExHyKazj2zl4zmZjzp8YHnwzvazY3q+31jS6mcP7ZNN78raYWp0lOEzjGyjE+MvPs/uVPbo4k9d3nThzsdiiFxKR3diFB2npPtahpPhHlQzOTk3OtDq1MjSoRLYJPeB5c++Li7TFlm+5WKN4/MucjH39x6SpRXxj3lgoyTkj6vFFe39ezq7OlNp/W6LTaGqPdblQyPqyMsyQnV6rkdMrz7OalL65N5K09x0/ub7PSif7+25fv9kdUMaZIeYbG/dH5q59fcWnfPtZlszQ0N8mLfxgcVzqON7W7LAqawDFOocDi3uH+yx9deRJQtJ85fbjbLM7dvH/z1p2xsimXm05PDY/7KOHLugaX+nhXnVG+xRZ4r4TIF5L+0TvXr19/kLTEsIQ4NzQ+nU3ePW/I64oz3rKyjk8mItGCtqOpq7PVxXg8Ko3BSDtsZhWTFvEyzlAURdMMQ4nUN2zBOEnLVUo8OvP4s48/vvis3HjkvfcOtZrlBIHJdbZal0nIhCYD8j2Whga7niNxHCNphuC/EfmFESQjU+BiZKb/5tVPrk4VzJ1nT/Z11Gi2uD0PJyiGIYTK/kU4TpAMJ8fKoekndy+d/3K6aNn19rF97UZscjASiGCcohCLhJO8aX9LV3tLnciPKDRGo77WZlaS8yJRJliKT857xmfTXOPBd4/qyVjIl5wdGBjb22ppUltcLQ76D38cmdC3fKupxWmU43w6EZxxB7OKxoPv7dHShUhUCIxMTO1qsrSZZVtp7sFJiiYxjP5Gu06coGVKgoj7nl49/9Enj5M1fe9+91iXlY3dn3Fn84KTTnun5hJUW8+unnYX45lXK/XGWkdtjU4mpooiUcrHPU/Hb925O0Na80kuOjY8EZZTt511DuWRFr2z3qgUAnMh1mCwOR1GFZ33Dd+//sXtwaRW5JL45PDsfCJz+dJll/bskXarYkvoLoLidLXNaos9QIRvUM+HBC2MRZwgqa8gCIwvFxIh90LAn2Vqhczc+FyMbn3rh+9+q1fmPv/LfMZjbHQ4asxROeUPpnO4rrnD0YTHh0fYr2cxnJIZXLtPnpwY/9eL184PGZoOnD1xal+DSSHDZHUdekvUnZl+lQYkW5n1mrvL6cD4nQsX/njbR9Tve+/bx/c1W5SsGAlMDzzxZFR7fvof/+SYdvajv/uvv3j86In3zNF6YsEz5QlkCA1XCM+Ojc0dbtFbOTEX8fRf/uTjG2MZe9/b59462l6jYYrB4Oz4iCembD31w/ePm2M3/+33n09PjHtCe1vsSEYshk8HpiZnR54MZofGviSEYqFQZmIfyQ3kqUYqRwoKBiunczwmYzminAzOTo55Ylm2SafE86lsoqDTqo1KMTb58PKnF667S7Y9p945eaDDrqVjc765ubGRwYnIvfskVi4WSwLuL2lsTR29LUrBMznlz5A6ZTk6O/psJtxutMuwQsI7cO2PH37xJGLcde7tcyd7HLL0dGjePT795Mvp9INbuFAqlXlmHuPMzc2tLvv2lhGYUMpGFybHPINfjpQe3ifFcrFUFmdzqUK4xSlXW1txrJzPF0ssRVNiLjI7MTYZShHtGg1TiEbzBUHv0Kpl1S9msRh1f/nJv/3u03Gm5dh3f3R2f4tZTuKYUMrHF2amZmI5QUamFtwTU/O76zQmrvoNUc6Gph988dnH16cLtt6z75441GHXbtWkxZcglDLBiXtffPrpPU/Zuf+ds6cOtlqVBV8uW8zyJIeJpVyhXFZQNCVmQp7xselojmvSqMlsKFIsYSabGo8NPn0ymdIc+uCvf3xQ4/78v/6XDx89Gjl39kAdFvSMTwdTglqeC7vHxmYPNqlUyYWZ4VF/3rTnB//Td3bJPJ//00c3RsfGPAd2t2wtGbEMYik+++Dzjz/841DR3vfuB28f7rQpqJw/k8piolKOFxcSGUEg5RSWT/inJib8WarBrOfEdCCSVHIqHZX1eMamHj2czd6l4IZlg6TC1tpcr7IszPgCWYIWS9FZj9sXbjfKI3MzvvGnI+744AMSKxULJWE6QNhqWrp6my2KraFYcZygaAKjF3c6r4KQT4ZTsWicU5syQf98kqzZ1dbVUm/PRTmVxmJzNtoMHB7gMZ5kGYZmGIbCFgeD4CSntdXW19C5z91zio7vNrrsBhmB4xhJEeTOae+JrVemRjkdePblF7/+14/uLdANZez6R7NPa5t79vW1CBhNZAMjX174dXpCGX86OI/pWvQKKj03ePmzm8NC3fHv7c5OT9/89FJzs+VsMz336MqHv/7thfFy3S7qrhCf7a9r393lwHESy4Xd/Zf/UJhV56aeupOKXoVcRm+DrPdNh1A595798X/SHyqIfD489fDB4zmy8dixg/sbsamoP45hMpVJJQ4OX/7Xv5+9xPsG70+lNfv3FyK+2UxS0La4agxlf//lj37z+6vTmLUbp84HJp42dPZ2WJ19Z372145QKl9O+wcfPh5JyNtOHD3UZsKCI5c+vfIoaz/2w97C+Mitjy+2ttq+260MPr3+h1/9+veDGXsvbSCz/kf2+haXtuX4D/9KvzeUKaaCU0PPJoNU/d6De9vrDNs+1pJkVPaOk9//qbJtoQgdz4dn8pY9xw73GrD5+RJBshqtgog/u/X7X4QeUsFnDx9GVHs6ComZ6UyAVzpb6+16WbXrWSxGJ+988i//3z9f8it75NrrH4cGNPU9Bw7sbiJn+m/deLCgaD58RpHxuu9+fr253nysdXHDOwzDxHI2PPHw6r/95vfXpss13eU75/0Tg42dfQf2t9XqtlcVJFEoZUITDy5/+NsPr84Q9bup28Wwu7++vdVa5HEGE0haptTKhKHha7/9794bgvfpg/6U6eiBfGxqIhjCdQ1t9VZdYoIhisHRuxd+nR9VxEYH5/KGZrWCzsw9vvnZ5XHRefR7fbnp8XufftHaaDpioUiSTwRGbl34bfIpGx0fm00bahQqblusX8RS3PPw0u/+8VfnJ7DGfdztC7GnN+u69++yUxwtYnyJVptNnHD39u//oTzARUaePJkW25qFdGg2Pp/mXPWdnXbCUvoxbjscyRYS8+ODI56YrGXfgV11yuTYw+t3JrPm3pOOnN83dOtKY6P5kKn14Ht/RTSFM8Vyyjtw/+5IStN1/Nih5i1eOAKnGIVapaP52Wg4HImp47FYrkRoFDJWzPFFkcUxPp8r5HiaIIlS0jf9bNyfxI1GvQpLBIOiQNZZ9UqawCtajz9H5PNxv2d8JpIlFEwmPDU25d3r0mwL/fmKrMv0XUx4Rx7euHHv6dB4KDsxfvcSQci7zv4lbWs60Xr4zOGpud/c/Kcr/1bW2JsPv//u6b2m7NPPP/3scdJx+id/+bbdd+Gf//sXX35ysc5YMHmuXr95Z3AyUnBPPb1FyM2Nh//0b/6Xn51r3Htk/+jM7y989IsvSrSp9ci7p/b3tVrV27Bo6IZDaRuPfc958D1B5Avx0Yv/9/+VuIUd+t73vv0t3dzV4PiX8Rxm7NrXPTnxh9ufTutrazRqWTHnuf9oQLavraWztd2lyQzdv3317uOnI/7c0MSjawShazvxwd/+73/9/cPf6zv5PVEUi/4Hv/y7v8/MmN/9/rfPtlBjn1z49F7EdOTHf/WDlsjnv/y7D+9+9Gmjg6oPXL129eaTqVDe4xm5/0dOX9P3nf/wn//mp2//u+PfFsRy0jtw8Z9/fX6A2H3q7Km9zWYl/bq/uLWBk5zWufvdP+86J4hCNjR667f/7V+vRetO/eD7R1W+m589WCBpfX3XLvfEh1cvT07ZnAZOi2cCz/ofynd3Njbv7m2u0VVN+BLywdHrly9dfjw6lxHn5x5foliF9dRP/zeNSijcuPzQQ7Wc+eF7e+RTn/zq04FrV667HKZTbUv6QpZS85OPb9/88sngmC85OvHwOoHL2k588LeGBqdVu61khCgUU4HJh1dv3L47NBErTI0/vkYobW0nfvQff3raabSaggmSq+/o6Zme+Oj25+MWR61GrhUSvqcPH9A9rQ3NXbubrDqZpvfwmdPTc7+5/D+u/VtZa2/pe//dUz3G9Ojn568O5GtP/+zPz9YuXPzHX1269sWlBscPDjQdOto3NffhF7+5lyypXLuPvn3y4K4mo3wbzD1CITx158bVK/3DnnDeMztwnSBw1aG/+D/+z//0XmObdmwqnFWeOHh89+QvPv3sdyM1rbUqNRWff/So38S0NTfuam9vd2qZhrr2g+8JYinuefjxP/7m8qTy0Knjvfrcww8fTSQ0ve++c6I2fe+3f7h098YXDc6fnDvyJz0nMAzDir7b//Bfwmlv/Xvffe/tHrtySy8ScEZhsDsbzPjj/gs//7sRYyHgDlDN51z1NaqShxLzBGPWazX8+MBn//T/jGizs/13J8pNZ4hC8NmwN640t/fVmxTV4vREPhdxP7x85Y4bbzp+Tp30zly/fLWxznKm4/W0IN5I1idTg1KY2w69/VfG3mTxqxQyytLc1+nQmaymcz/S6p2Hnoz7Sjpnx96jRzpruUTJ0fP2B/vbjh3b3aBolImEYyymMKvk2vo9b/0Z2/Ttr2QdpdDXdfU12W21Cv2f/Ht9fc+RsUCJtrbu3reru8WuRSqiGjhBMSyFYZjI4LU9b73/sybM2e3Qa5R09+72yWtjvnDLkdM/+18bj8wLujqnkY5OTs9nFbVtjTZFLjh414eLms7TP/zbzm/lvzohZ6rv2dVgVMo4jsYxDGNMDQff+VN5TNHVbFYwOYW1460fdDUdOba7TZeV/XnJMBjgrEpWTXWf+OFfO85+9VOSnLqmY1+LVSNjORLDBMJSv/vUu/I23N5Vq+O2dA2klYITJMOSGIaJlKG248h7f2rLaVtbnE6lsss7G34aTog9h3/0H5z7A0Wt3WVhUrNuX5a2Nrc5VWXfSH8g0tjdWmuU0zhrbD323r8zZK17XBqaxGldw6H3/kq953l6M0GrGnfvceiUiaYDZ6za1sMHO6yEU4ar22fKZt1XCoKSG5r2nP6RepfM2WpRMhRmaOr71k8ULe/keTgLZazv2eXSbwMNgVOsofHAuR/KY6rORgNL0gpj0/5zf6HqSH11AK0y1ff0dbRZC1h8/vZkKFPbffIv/qbhaEQ017oMeGR6eqGkcLS11cpy04P3/PbW3sbucz/+G71zz8C4r6R3du49crizlouPOXpPf3Cw6eix3Q2KlOwntGMorjapFLq6vd/+c0Vty56n7lhJ09C5a09fp8us2KJzD6OtP3DmJ/KYqq3FIqeJnKau760/+8+tuSL//ADn7oMNNU7FsX3PFm55JiKnT/zof7Z1edKymnaXOuOemFrAbC1N9VY6Pf7ofrC2tavRqpRhGItbm/adeU+/h3V21mjZWO2u49/bbenc292oK5oZuXEkprCrOZZhOZbAMYwxNx9+9y+USd3uRsPL2mu/fghGZWved/zQ2Nzv/vjpH5K4bd87Pzh6oKe1QSxPjnlCBUrVsOdAt/s3n17+dMrZYFeo5dHZif5rD/nelqbu7j31JnkVO6LI56NTD69/cWOabDz5wV8c1kxc+qd/unPt/NX6evNbnZYdZpFYp2LYoiiI4jdOheM4joMHShQFURDFrzeJoiCKGPbV/q/+wvCv91SchVh0kopNiJchCoKIYRhOELhQyoQm7l/+5NZwSNl69EDf7maHQa1gSaGUz8QDc6MjA3eHg5i27dTJfd0NhkoXe8Uv+dVZK34+URREAcNwnCDwij0VP+vXpyGqnAXfmb/l118RJuTCUwM3L14bCuL23fv6ettcVp2Co3G+kE1G/VPDA48ej2d1zQdOnt7TaFbQOPxqIoZ99XWJgvDNO+vr+0a6h8Svbr6Kb/Mb95goiot+jiU/6lZmJZ8F57OBsf7rF2+OJVhn3/493S11Fq2cpTC+kEmEvRNPn/Q/mSxZuo6cPrWrTs+S0vX3ojmp8i+x8uity6KRL3Px8Gnf4GcffnZzDG89evDwwVanSa/mSLFczKXCvunxgTvDMzHSdezYsUMd9q8shZX39vNrD8crvzkc//qb+eYwthhCKR2Y6v9yYCqn6zqyb5dLx4jlfCIwNTz48LE7gRvb9/Xt6nQaZKXgwO3z5x/4ONee3TV4zDuXYGobHercwtizBczgau+xUZGFUEHh2t3balVRRH5h8MvrNz1404ETx9uMRNz95MH94YS2Y9/BXjsVmeq/+2Qqq+s8vH9Xg1FG8Lmod/TBk/EAUbOrr7fdptqydeBWAGrN9cYg8oXkwvjQ43ujvlxRJLOZZCZXJnCSlikUCoplGYurp7uru7a6kx6xavhiKuR+Nvh4ZDaWLwqldCZdKPK4QHMKtUJO0HJtXcfunnaXSYGKu68FkS8kFyaHBp6M+ZKlIl9IZTKlEo8LtEypUchwSqFv7O7raa3VoVRxDBP5XNQzcu/GwHSyRHGlbDyRLQkiTpJypVomk2Fae2vn7r2N5qptC94UxHJ6YejO9YtXp4o6175juzuaaoxKjiYxoZBJRLyj95/0P5jBmnefeudYd416R9hUVwmSEW8WAl/MxoN+z9To0zH3fCSLEYzSUNPQ1t7aWGc3LGrIg1g/+GI2HZ2fnRgbHZ72xlJ5DJPrra7mrrbmertFq2CQdFsf+EImGfHPjI2Njkz7EpkihimM9oaWnramOptZI9/yFvZNRBT5QjIU8E2PPBud8ATSJYxi5WZXe0d7R4PDqFFu01SedYXPx/xj/YNPpwJZrFTMp1KZAi8IAq1Qa5WyEq00NO062t1Uo2HfYA2BIRmBQCAQCMSylAvpZNjnmRwbHfcEEvmSgClNtY2tPa0NDotRtTNCu9YGkhEIBAKBQCBWCTKlIhAIBAKBWCVbOqMXgUC8CQiCwPP8y4/bAAiCIEkUEYRArB4kIxAIxGsmkUhEo1FBeA2Ny/V6vV6vx7dgXiICsU1AMgKBQLxmMpnM/Pz8azFIUBSl1+s3/30RiB0Dio1AIBAIBAKxSpCMQCAQCAQCsUqQjEAgEAgEArFKkIxAIBAIBAKxSpCMQCAQCAQCsUqQjEAgEAgEArFKkIxAIBAIBAKxSl6hbkSpVEomk/l8fuNGg3gpcrlcpVJR1CoLfgiCUCwWS6XS+o7qjYVlWYZhXvcodiyiKBYKhQ29XMPhME3TS8tPkSSpUqkUCsXGvfW6UC6XC4XCGit34TjOsixN0+s1KsQbxSs8jVKp1NDQkM/n27jRIF6Ky+Xq7u5WqVSr++/lcjkcDsdisfUd1ZsJSZJms9loNL7ugexYBEGIRqORSGTj3mJhYWF8fHypjOA4rru7u7m5eePeel3I5XKBQCCXy63lJDRNm81mVIYLsTpeQUYUCgW/3z8xMbFxo0G8FJlM1tbWtur/LghCOp2ORqPrOKQ3Foqi1Gr16x7FTkYUxWw2u6EyYjmUSqXT6dz8931VisViMplMpVJrOQnLshqNZr2GhHjTQLERCAQCgUAgVgmSEQgEAoFAIFYJkhEIBAKBQCBWCZIRCAQCgUAgVgmSEQgEAoFAIFYJkhEIBAKBQCBWySqrGCF2HqVSKZPJoMpUK4ckSVEUq+basSxrNBpROujGQVGUXC5nWXaDzs9xHM/z4XB4ubfe+mXH8vl8JpNZSWUqmqYFQYjH40t3yWQyo9GoVCo3YICIHQKSEYivyOfzCwsLa0xAf6PAcXxubq5q7T+9Xr93714kIzYOhmHMZrNOp9ug8xMEUSwWPR7P0l0KhcJut299GZFMJv1+/0oWBjiOz87OVq2Na7FY9u7di2QE4gUgGYH4Cp7nYfnyugeyQygUCq97CDsZgiA4jtvQYtWCIFS9HQiC4Hl+4953vSiXy9lstlgsruUkDMOs8QyIHQ+KjUAgEAgEArFKkIxAIBAIBAKxSpCMQCAQCAQCsUqQjEAgEAgEArFKkIxAIBAIBAKxSpCMQCAQCAQCsUpQwifi5cjlcrVaXTWtHFEVpVKZy+V8Pt/SXSzLqtXqrV91YPuSz+eTyeTGpSmyLBuPx6vmmup0OofDIZfLN+it1wUcx9VqtUqlWsnBWq02nU5XvZJlMplKpapaNwXxRoEeDIiXo1QqHQ4Hx3GveyDbBhzHM5lM1eJFOp2OZVkkIzaOXC43Pz+/cYXUcBwnCALH8aW7WlpatFrt1pcRIHdWcjBBEKlUqmr9DIPBwHEckhEIJCMQLwfHcZIkkTXilRBFsVwuL92+LSoXbWtEUeR5vuqXv9EUi0VRFDf/fV8VgiBWfjsLglC1orYgCNviwyI2GhQbgUAgEAgEYpUgGYFAIBAIBGKVIDM1AoFAbFfkcrnNZjMYDEt35fP52dnZzR/SJlAulwuFwtr9gxRFMQyD3LVrBH19CAQCsV2RyWQymazqrmAwSJLkJo9ncygUCh6Px+fzrbEBnlwur6urQ/HjawTJCAQCgUBsJyiKoigqGo16PJ61KAmWZYPBYLlcdrlcSEmsGiQjEGtCFMWqUdyI5SiXy8VisWpVA4IgSJKsmkkIbK9vG3/O6x7I10Cu5oaeXxCEqsb2jX7rdWSFKRhwJVdN+CRJckMNISzLulwueD0zM1MoFFaXM1IoFGZmZuASraurW86ug3gxSEYg1kQikYhGoyjva+UsLCz4fL6qdSMcDkdjYyPLslX/oyiKsVgsHo9v8ADXDZlMptPpttQiT6VS6fX6jXvCKZXKcDicy+WW7lKr1Xq9fusriVKpFI1G0+n0S48MBAJzc3NVZURdXV1jY+OGxhysr5KA13V1dRzHbSnhuy1AMgKxJpLJpNfr3UZL5K3Mnj17amtrXyAjEonE3NzcJo9q1ej1eoVCsaVkhFKptNvtG1r7KxqNRqPRpdtramq0Wu22kBGRSCQYDK7lJKIoulyujQ5dXC8lkc/nK20SSEm8KkhGIBAIBGJbso5KQqo5i5TEq4JkBAKBQCC2KxuhJJxOp0wmQ0pihSAZgUAgXjMkSbIsuzQyEWJOX8uQdgAMw6hUqqWeBRzHl3OcbVMkJSGKIiiJ1Z1HUhKiKELEJVISKwHJCAQC8ZpRq9VOp3PpIrJQKKzRSf8mU1NTc/jw4aq9RQqFwhorLmw1ltokVneeRd4NpCRWApIRCATiNaNQKKr23c7n81W3I1aC0Wg0Go1LtwuCMDMzs/MKXEpKAsfxtdSTkCIucRxH3o2VgGQEAoFAIHYClTaJtSiJXC6HIi5XDpIRCMROQBTFYrG4pbqQQ53BqoW2OI6Ty+VbLfuxVCqVSqUNOnkikQiHw1VzIBUKxbYofESSJE3TK4lWwXE8l8stvRpxHKdpehPqScBTf+1KgiRJiqJekIaNwJCMQCB2BqVSKRQKJZPJ1z2Qr2EYZnZ2tmp5ovr6+o6Oji317ITqXuFweIPOHwgEpqenly5qKYrq6OhoaWnZoPddRziOM5vNcrn8pUdSFOX1epcKDoIgTCaTXq/fmAF+BcuydXV1EGqzdiUBusdmsyElsRxIRiAQO4FyuZxMJrdLQCLLslvtwSmKYjqd3vwvkGEYs9m81b6NqjAMo9VqNRrNSg6ORCJLN1IUpVAoNlpGYEsiLvP5/OrOk8lkpqamwGyGlMRyIBmBQCAQiJ0G2CSw5xGXq1MSoiimUqmJiQn4EymJqiAZgUAgEIgdCMdxoCQwDFsXJYHjuM1m29Bi6tsRJCMQCMQWhSAIrVZbW1u7dBfDMFsqtGIbAeWn1Gp11b0EQVQNZ9mmrLuSwDDMarUim0QlSEYgEIgtCkVRTU1NJpNp6a5SqZRMJqv20kS8GBzHdTrdciIsFot5vd5NHtKGwnEcFDcD78bqrplKJSGK4kY3eNteIBmBQCC2KARBLFdDKZPJrPqRgJDJZMvJCEEQdt4DcpFNYu1KAnk3KkEyAoFAIBA7GRzH111JYBiGlASAZAQCscOhaVomk21ozZ9XhaKoZDJZtd0DhmEymYzjuE0e0gsgSZLjuA1yh1MUBSUrlu4iCEImk22LB1WxWMzlcoIgvPgwgiAoiqpayIFhGJ1Op1QqN2aAXykJp9MJf65dSUBdCqQkMCQjEIgdD8dxVqt14yboVcCy7Pz8fNV6iDiOW61Wq9W6+aNaDoqiNq5oEo7jPM9LpZcrgR9uWzyl0un0/Px81Yqli/D5fFUFmVar3b1794ZepTiOy2SydVQS2HPvxk4KSl0FSEYgEDscqPmzXGT+6yKTyVTdjuP4JpQneiXAKrBxX6AoilXLj5bL5Y0rzr2+FIvFVCq16ipPGIbl8/lNiHSpVBI4jrvd7lUriXQ6XendeJOVBJIRCAQCgXhTWGSTWLuSkLwbb6ySQDICgUAgEG8Q664k4E+73b6lIpA2jTfxMyMQiO0ORVEajaZq+2ae59/M2XxdYBjG4XDACnvprpX05doWrKOSyGQyb7iSeOM+MAKB2AHQNL1c2KMgCJOTk5s/pJ2BXC7v6OhobGxcuiudTodCoWw2u/mj2gg2QkngOG6321fSS30ngWQEAoHYfhAEsVwGJsuy0JIRsQpIklSpVCqVaukuiqLi8fjmD2njqFQSoiiuJXcjk8mMj4/Dn2+akkAyAoFAIBBvKOuYBZrNZicmJkiShObvVT1uOxIkIxADad5TAAAgAElEQVSIN5dcLpdKpXief90D+Rocx4vFYiQSWbqLYRibzbal0kEh0XEl9RJWAcMwuVzO5/Mt3aVSqWw2m0Kh2Ij3XV+USmVV28YiVCpVPp9fWFhYuothGJVKtXF5EOuoJDKZzNTUFMdxvb29BoPhDVESSEYgti6RSGR+fr5qzbtKCIIwm81Wq/WNMiSuC6lUyuv1vvQb3mTm5uaqxqmp1eqDBw9uKRmRy+UWFhaqVn1YO1DYoOpV7XQ65XL5tpARy/VoXQRBENAnpeoZGIbZ0HTKdVQSiURiZGQEx/Genp43REkgGYHYuiwsLNy9ezeRSLz4MJIk7XZ7bW0twzAKhUKhUNA0TdM0SZI0TVMURVEUUhhVEQShWCxuNRmx3HgIgthq5ZhEUSyVSpv/BWYymS1lQ3oBFEWtsI64IAhVv8lSqfTSMttrZ72UhCAIsVjs2bNnGIa9IUoCyQjE1qVcLmcymVQq9eLDcByHdYwUHcYwDMMwFEXRz2FZVi6Xy+VyWNZI8oKiKIIgdvx9jkAgXsrSiMvV1eV805QEkhGIbQ+sCGGdmslkgsEgXgGGYTiO0zStUCiUSiXLspUKA5ZKoDCkXRIkSe7s+x+BQFQCSqKyF+iqlUQ8Hn9DlASSEYgdhSAIVe2f+Xwest4lYQH/4jgOLScUCgXHcWCrAJEBSkKhUKhUKpZlWZat9JK8gUVmEIg3gaVdxVenJHief0OUBJoKEdsGnU5nNBpFUSwWi/F4PJVKVa21VxVRFEVRrKowcBzPZrPhcHiRwoB/5XK5Wq2WyWQsy0ryAiwZcrkczBscx4HCgL3IhvF60el0PT09Lpdr6S6CIFCUzKphWdZsNldtUVYsFr1e7+YPaYNYL5sEKInh4WFsRysJJCMQ2waHw9HX10eSZKFQGBwcHBoaAhlBEATkAUJHxEKhUCwWk8nkCiOkQGEst7dYLCYSiUptgT03Y8jlcpVKxXEcx3GVXhKoGSyXy2GXFJ+BFMbmYDAYtFpt1d80EAjMzc1tUH7mjgcal1fdlUgkdp4+k8lkLpcLx3GIk1hdIG2lTaK3t1ev1++8GQDJCMS2QS6XWywWjuNEUaxMpidJsqGhYdeuXRAhIcmITCZTKpXK5TLP8/ACdmWz2XQ6vcKY/xeIjFKpJCkMrEJkkCQpk8nAS7JIRixSGFIoKPKSrCPgqKq6a+c96jYTKdjolXZta8C7AZPAzMzM6pSEFHEJWaA7T0mgmQuxnYAn+qKbECyQBoNBOgAQBOEFMqJQKMBGoFgsQupjOp3OZrPg/lgqICq3wOulx/A8Dzpm0SAxDCMIQiaTLScjgsFgNBq1WCwajQZSTuBgmUwGbyQIQtVvALEFUSqVWq12g9ShRqOJxWLLpUHCW29xzQTL9JfmYWEYFggE5ufnq2aNWq1Wl8u1woTS1cFxXH19PeikVdskQEkMDw+LorjzvBtIRiC2DV6v99atW1CFZmZmpvL5XSwW0+k0hmE4jpMkSVEU3KVgBoBjFt23IDJ4npdkBIgMkBFg2ABBAOns+XweShZW+kqkMSyVFIvkBfwpCEIqlVpu6hwYGLhy5YpOp9NoNEqlUq1WcxwH1gutVmswGHK5XLFY5DgOFAZN0wzDQMIqz/NIZGwplEqlw+HgOG6Dzh+LxWKxWNVdNptNqVRufRkRiUT8fv9aTrJnzx6bzbahMgLDMJZlwbuBYdjalQSGYWCT2DGdX5CMQGwbfD5f1cLAPM9PTU1VlREASZIcxy3q2AQbK0vjEQShUqlMJhNFUWDAgH9xHBcEAWSEJDJ4nq80cuTz+Vwul8lkyuUyOFMrR7ic3WIR8Xh8udZHOp3OYDAwDANmDKVSCTICXqvVaoIgUqlUPp+HkE+ovgWfl39OZQbsy75sBALxNSzLgncDW5uSkOIkuru7DQbDzlASSEYgtj08z8/Nzc3NzS13gCQjlmqLRYIDykgs3VhpmgbnAk3TMplMEASe50FGyGQypVKJYRhBEHAMWAjA4FEoFAqFwqqL8S239CQIAiI9QSRBWyCWZaWiWxzH6fX6aDRqNBqhfDLU4CIIgiCIYrEITp/KMhsIBGIpYJOA1xAnsfJMMYnKOImurq6doSSQjEDsfHiez2QymUxmvU6I4zgYBiorYEKRK5IkJf9FJauYcVaCIAjpdBosMVUHCcW1wEsiBX4yDAOip1wup9NpaLrNsixYOCqLaixiIz4CArFdWC8lIdkkOjs7jUbjdlcSSEYgdgKwtl7qSnhxMueqEUURrAvrfuZ1RBpkMplc2jiRYRilUglmCYixACUhZZTIZDKw1shkMsmZAkkQi6w1lUDEyfp2QIAhrcupCIKgKKrqJSEIwnafzV8jFEVBHE/VXZs/no1jXZQEz/NStezOzk6DwbDFA1lezI76gRE7DHiGvTTLH8dxtVotl8vL5XJlzGOpVMrlcksTO6vmXywVHC/euEHWhU2jWCxGo9Gqu6AGBqgKeDxIfzIM43Q66+rqlgtqSyQSExMTVdt8rw6KopxOZ21tLU3TPM+v8WtXKpW1tbVVm1rxPB8Oh9dy8jcZq9V68ODBqsIaqrls/pA2jnVUEtALFJTE9lWxSEYgti4vmJsqwXFcqVRyHFf5mBFFsVwu5/P5crlcebAgCBANULmR5/l8Pl8qlRbli0LuRuVGiIQoFAqLzlA1U2OR2pD+XLodex5y8eJPujlA6krV4l1yuZwgCLvdvpyMSKfTY2NjU1NT6zUYiqK8Xm9DQ4PBYKitrZ2fn6+trQVlA23VXulsUPW86q5yuTwxMbEeQ34TMRqNRqOx6q6ZmZlFeVU7gPVSEtFodGRkhCTJzs5OrVa7TZUEkhGIrYvBYIBqEOuIFPlYuRESOxetmaCJ9lJtARUmFp1hkTFfiqwEpI3lchn+lQwncCRU+BZFEbwM0mmlWhHwAv67uATpSGlv5RnW0XaC4ziYJdblbCuhXC7Pzc2Fw2GNRqPX6+fm5p4+fdrc3Gyz2YxGo0ajkcvlq5MUrxG4Njbu/NlsNpVKVdXfm/zzrQ4IFl7JD4rj+AsatUPd2PUe3VdUKom11LiMRqOjo6MMw7S2tkLK1boOczNAMgLxZgFRFEu3V3XrrpxF6kQURcixlB780kY4UjKcSM/+crlss9laWloqKw2AVQAESi6XC4VCyWQSgh5AyhSLRbCLlMvlZDKZTqfh5NJIQLLgOL5UnUhaR9q4yCIibZGKcREEsShLdhMQBAEiZP1+//j4+L179xwOh9FotFqtzc3NTU1NVqtVkhRQzmuLz8XxeDwSiWzcAj0YDHo8nqpP0NbW1paWlg163/WCZVmDwbCc3WjRkfPz8wzDVN27EeuQRe8OSkJcQ43LcrkcCoWePXtGEERzc/N2VBJIRiAQ68By6uSV2LNnz8mTJ1UqlbQFfCigRfL5vMfj8fv9YLoAM4nkoMnn816vNxQKSeYN6QwMw+j1eiiFKW3M5XL5fD4Wi8XjcVAScNrKrFQwscBGMJ9A/c2VfyIIs5DGuYoHJxQCwXEcBlAsFqXkXpZlrVZrTU2NXq8HBdbQ0ACSQqvVbmVJkclkFhYWNt+HBVFEW19G0DSt1+v1ev1KDl4uEAcadW6ojMAqlMRLa1xCtBbkii9SeKVSaX5+Hl5vRyWBZAQCsXUhSVIyk4BBYrl5KpvN0jRddQGn0+lcLpdGo5G2SGEf8/Pzs7OzEMQK01ylawbMJ1JBT1EUKYoym80rH79arXY4HARBxGIxqcUJnHCFkgJaoJEkmcvlRFEkSRJsMPARwO+OYZhMJrNarTabTafT2e321tZWl8tVKSkgCQXlrCLWnZV4N3ieT6VSwWCwkM9bLRajyYR/UyiUSiUpnWrbKQkkIxCINw5o7SGTyXK5XDweX10T5JWg0Wg6Ojq0Wm0qlYICoJlMJhqNxmIxcMdAVEplis1SYC9FUdCZPZ/Px+NxQRBIkgRRAgLL7Xa73W4Mw+RyudVqtVgsOp3O4XC0trbW1dVZLBaj0ajT6ZCkQKw7L6iWLYpiPp+PRCIej8fj8fDFYltTE0OSSp1ukU2iWCwuLCzASbaXkkAyAoFAbBQ0TavVaniog2IoFArQVSSbzUqqIplMSh1MFoWvCoKQzWal0l4QUofjuFarlclkqVQKmqxCNQjw/mSz2enp6enpaQzDlEql1Wo1mUxarba2tra1tdXpdFosFpPJpNPpoBjXDitsgHgtLK2WDea9RCLh8/ncbrfX600kEgSGCeWySJLNbW06nW6RUAADIbxuamrSaDTbQkmg+weBQGw4Ut0quVyu0+kqoy4qVUUikYhGo9FodFHJUYiKKBaLiUQCex5yAd4NMFQoFAqoVSqKIvQ0gQk9nU5PTk5OTk5iGKZSqaRgTKfT2dbWBlYKs9ms0+mg4pbVal06eNi1Kd/TDoRhmKq+tnw+v8MEXGXE5fT0dDQaDQaDk5OTMzMzsVhMsvn5QyHa7TaYzVVVAigJuHq3i5LYUb8iAoHYFkjlt1UqlcFgAEFQKBRyuVwymXz8+PHo6Kh0JJTRlCI9MQwrl8uVPcxwHAcbBuSSkCQJ7pJFbwp6BYpDgI0EJEVjY2NnZ6fVasVx3G63S44P6SEHJcM3zvWzs9FoNFUzKTKZjLTy3jGwLFtbWxuJRIaHh4eHhz0ej8/nW1SrHi7+RandlYB3A15vC+8GkhEIBOJ1ArkYEKuh1WpNJlNllzWKoiBYMplMQrIrjuPgDUmn0xAcCjGbcDxBECvpYAJnGx8fxzBMr9fb7XaNRgOSor29HYpSgJVCLpeLouj1epGMWB3Q6X7pdqiwvvnj2ThEUczlch6PZ3Bw8OHDh/39/YlEYlE1GpIk9Xp9Q0ODyWR6gTjYXnESSEYgEIgtxKLUWYqi7HZ7T09PPp8XRbGmpkalUoVCIa/XOzMzE4lE8vl8IpGIxWIQ1yZVuVg54EaBt9bpdDabTa1Wa7XapqamtrY2i8WiVqshwxa6mm3lCR3xuiiVSsFgcGho6MKFCzdu3HC73ZUGMwDq7dbX1zudzpe6ySq9G1tcSSAZgUAgti7Qp1SlUqlUKo1Gc+jQocbGxnQ6HYvFAoFAKBSKRqMej2d8fDwcDudyuWw2G4/HIWbzVd9LEIRIJAJ1CEiSfPDggdlsVqvVKpUKjBN6vV6r1ep0OpVKBd1Q1/fDSubu9T2tBKxuIb5kk996vYCalVW7oiwCx/FAIFB1F0VRGo1mJeWtVoIgCKlUamxs7NatW9euXbt///5y70vTtNVqbWho0Gq1K0kUqoy43MpKAskIBAKxPYAamtC+oa6urlwuFwqFbDYLkiISicRiMSjg6PF4IE4znU4nEgko7vlK78XzfCgUCoVCGIaRJKlWqzUajUwm02g0tbW1NTU1Op0OIu1TqVS5XIYskjV+QJIkN7rwIoZhUGljETRNWyyWFVZ8eo1ks9mFhYUVOpgWFhaq+k0UCkVvb29jY+Pax5PL5bxe76NHjz7//PM7d+7Mzc0tig6WIAhCrVZDVO/KBei2qCeBZAQCgdh+QAooTdNKpdJsNjc3N0ND10wmE4/HQ88JBAIej8fr9SaTyUwmEwqFYrHYKiRFLBaLxWIYhlEUNT4+rlarIZLDYDDAMBQKhWSiWLWkgFpbOp1uFf935cAHWQTHcRv9vutCsViEn3IlB1f9pBiGabXatWuIcrkcDocHBwdv3rx5+/btp0+fvrjAOcMwdrvd6XS+qhVkkU1Co9FstZInSEYgEIhtj5T6odVqa2pqoIh4NptNp9PRaBTUQygUgu6jkF+aTCbD4TCEba78jcrlshRLQdO0lNOhVCqNRqNer1epVEqlUqFQqNXqUqnEsuxWm/QRa0QQhGQyOTExcfv27WvXrg0MDCwsLLy4ySeUf21ubn5xZOVyQLVsOH9LS4tard5SFxWSEQgEYqcBRcTlcjm4P8ChnsvlQFIkEolIJLKwsDA5OTk7OwstzaBkxQqXuQAUF4LXFEXNz89zHAdWCp1OB33UbDabVqtFda52DODF6O/vv3Tp0pdffjk3N5dOp1+sRAmCMJvNx44d6+7uhuryq3jfSu/GVlMS6LJelnw+Hw6Hs9nsep2QIAgI0dqCzi3E1gcs3ss5zhmG2foNoF8LEKQJcZpms7mlpaVYLObz+Uwmk0gkwuFwKBQKBoMLCwtut9vv96fT6WQyGY1G4/H4yns2lstlqEtBEEQwGGRZdnx8/MmTJ3a73W63Q8QDNPhQKBQvLv79hkBRlMViqdonbLmKVa8XaMU5ODh448aNO3fuDA8PR6NRcJBBO73l9ATHcbt3737//fftdvv4+Ljf74dE5VdlyyoJJCOWJZFIDA4Oer3e9TohRVEOh6O2thZS5BmGoWkaSvutS3wWYmcDFaCXm16j0WjV0DnEIiBOk+M4yf2Ry+WkFI9wOByJRAKBADT9CofD4BYJBAIrNFRAL1ZIQw0GgzMzM9AWVa1W22w2m81mMBjK5XKxWITtNE2/mesKlmVbWlocDsfSXdCE4pUsQxsKeDHGx8dv3bp1/fr1wcHBQCAAGcgsy5rNZo1GEwqFFpWZAkiSdDgcJ0+ePHz4MEz7GIatXUlQFNXa2lq1IMfmg2TEspRKpWg06vf71+uE0OfQ7XYrlUqVSgVeVUiyYlkW5hSkLRAvAC6Yqrvy+fy65x++CZAkqVQqlUolhmFOp1Nyf6RSqWg0Gg6Ho9Go1+udmJgAQ0U8Ho9EItFodCXJAjzPp9NpeLqQJOnxeOCuZxhGqVTqdDqYCtRqNdz+28zrIZYL2Wy2KFCcXM7R5CtOVgRBQMmvpbugkdsWkRH5fH5ubu7hw4eXLl26e/cu5GJAbRKz2dzR0aHRaGZmZqpqCAzDZDLZ3r17T5w4YTabcRxvamqCWX2NSmJ4eJhl2fr6+q1Qpn1bXbXbHEgvTqfT1HNIkoRiOzCtqFQqWCeBDZbjOLlcDrHfAPQfet2fA4HYmSxyf7hcLqjALTX7AEPF7OzszMxMMBiEIM1AIJBOp19a86pSUsAtD2sGuVwOvk6lUqnX60mShIBNlmUXdYDcYgjFpP/ZowcPx+KcvXXP7s5Gm4ajdpRZBXIxBgYGrl+/fufOnZGRkWg0CuXYIdejr6/PYDD09/fPzs5ms1mYnMFCAN5wmqabmpreeuuttrY2+DVVKlVjYyMcuRYl4fV6CYLAcby+vv61FwNFMmKzgbZvi6JsoHERmDclbQEdB5RKJcdxsAylaZrneblc7nA4FAqFXC5HoVsIxAZRWaLbZrOBJyKfz4P7A6I1R0dH79696/F4crlcPp+HNfRLE0olxweGYZCeyjAMRVFqtXp2dtbpdBoMBpPJBAWvlEolwzCgQlb31FkJNE2XSqVwOLx0l1wuN5vNYLCpgE8Hph5c++z39/y0vnVuIfX2mf3ttTrZ61MSMB+uxEMkl8uLxWIwGFy6C3J3KYpKpVLj4+M3b968du3a06dPA4FAoVAQBAG6Zhw+fPjw4cMYhl25cmVgYCAej4MygJqnUH4KIivPnDlz7Ngx6dsjCEKlUjU1NcGfq1YShUJhdnYWXr92JYEePytFLpdrtVqKoiSzJ1xV63LyqtqCIIhwOEySpCQsCIIYHx8fHByEFgBQshducpPJBOX2wIax5ZcyiDcCyIur+nBaDkEQKosA8jyfSCTm5+ehRfj8/Lxer6coCtIsSZKs7KG1cUhlKhQKhcFgqKmpgQ6lDQ0NMplsYmICSlZAeYlsNpvP54vFYiaTeamqgC5iICkSiUQoFBoaGoIqDjU1NdDdw2KxKBSKTCZTLBYpitoIXyeO4zMzM1UnDZvNtn///iUyQigXMrGsSJucTj0ZfPrwBieTn+1rMisp4vW4YtVqdU1NzUouBoIg8vl81VgiSNxNJBIPHjyQcjGy2SxM9SaTqa+v7+TJk/v3708mkx9++OGNGzdCoRDsdblcLpfL6/VKpoje3t6zZ8/W1tZW/mTrqCRmZmZAvjQ0NLzGCGskI1aKTqfr7u7WarWS6zSTyZRKpXK5XC6XQQRA+HepVFoX1wO0PVy0MZlMer1eWLuAAYMkSYvF4nA4jEajTqeDMjgKhQI6IKvVaqVSCYsqFMmP2GQikUh/f/8rPeZFUawMOyiXy263O5VK4TiuUql8Pp/NZgNfALj5IGyNZVm9Xq/RaCiKYhgG9koFnkEEsCy7XsGMUpkKvV5vMBiSyaQgCJDLB/MDPEh4no9Go5DBEY/HY7FYLpd7weQAvc4zmQyO41COE6KmzGazwWAAf4dWq4VYClgqbEL4lFwur5ajSLJytVZrsdP2w4fqxMmBJ6NPHtbXmLX1Oo56LToCLoYVXmyCICzNwhMEIR6PDw0NPX78+Pbt26Ojo5IXQ6PRNDU1HT169MyZM83NzbOzsx999NEXX3wRDAZBQzQ2Np46dcrv9/t8vnK5TBCEw+E4ffp0b2/v0ol3vZREsVgEmwR4N17XDI9kxErhOA4CrUVRhAVEqYJCoVAsFqGmTTableRFNBpdrpLa6oCexYs2er3eoaEhmFVhroGGRg6HA9yuoC1UKpVOp4PRgqoARwkK5ERsEEttbK8Kz/NQ6BrDMJIk4TqHpqA4jksvZDKZ1WqFmpKQFsEwDHglQD2wLCuTySCg0mw2y+VySKBVKBTgUoQnEBz5qsGqYCkEG4lCoYD2YHq93mKxQCIoFL/y+/2QgBCLxRYWFpYLysMqzJNSngi0BKNpWqPRQFMPlUpVKSk23bNJyvW2Bpt25EkkUmzvabGN+MafPpvZ01ajYSlyu80o0Jxzbm4OGnxPTEyEQiHJi+FwOA4fPnzmzJm+vj6TyfTs2bNf/epXFy9eDAQCkob44IMPGIa5fft2LpfDMIxhmP37958+fXq5DO11VxIYhr0uJYFkxEqBaWJR6xpRFKErMbyQViRg8CwWi0+ePJFkBEmSkBokKQ9YtYDaXQvwXpVbcBz3+XyDg4MgKSDkguM4i8Wi0WhgsgMTBQR2aTQasB5DwD8E76xxVIh1QSik4olMiVJoNUp2203Pa0YURVDk8OcLGha43W4w+EsBRtjzB7wkODAMgzxPlUoF5g2VSgWPZ4VCAbY9MG9QFGUymQwGA9j85HI56BIQ35JoWDoSeCOSJMH9wbIsz/NgsAQrZiqVcrvdw8PD4XC4UCgUCgWoYLFcjQpRFKUbHMfxRCLh9/vhdgZJoVQq1Wq1TqertFKsy5e/PEIhnUzlSFt9bc3Il6MPnur3Wa16xUIqFs8Uywb59rpOy+VyMBgcHR3t7+8fGhoKBoOSw9poNPb19Z0+ffr48eMtLS0Mwzx+/PiXv/zlhQsXKjXET3/6087Ozt/97ncejwfDMJIkm5ubz50719ra+gIDWKWSEEVxfn5+jUpCFMXX4t1AMmJNSHNTJeJzBEFwu93SdpqmGxsb29vbQUOABySZTObzebBegPjI5XLpdHrldW+qUjn1VI7W6/WCqpCAXkcw35VKJUEQwCasVCqlHNRisbheUSCIV0NIzw/euPjFvVm2ft+Jk4e7nHqu+hOC4zjw0y/dRZLkjs8FhaDFlRzp9/snJyfhtgW1ASJDegGvKYoC6yMsHlQqFTyeZTIZx3HwlXIcl0gkfD4fTdMqlYqiKIhMgr1QDBvDMJqmQbJDp1BBEODmghCKXC6XTCYjkUgqlQKjJrQFqToDVIoqyB/x+/1wF+t0Oo1GA+mjGy8p+Exoun/AnaCUrlaH/8uJh0PsnpZOl8G6Lh4NmqYNBkPVsEGe530+35rf4Ssge25ycvLRo0eDg4Nzc3OpVApiWdRqdXNz89GjR8+ePdvb2ws/XH9//y9+8YtKDdHU1PSzn/3szJkzN2/efPDgQT6fx3HcaDS+8847J06ceGk2JvTrqrRJrM56V2mT2HwlgWTE+gMzEfZ8JVS5XalUglsEbBiCIEhxFfDUl7LLwEohaQvwmEAvwVUPbDltEQqFpNUbbOE4DpJEwCZsMBiePXsGgZySvACFserBIFZEOT59/9rFP1yaUddNzcfi3/3OO/sbDNWUhFwu7+joqFoTMJlMQs2DjR8uplQqW1tbjUbjcgfAgxBmajDd8zwP4QLSxSmKopQbue5AvamXHobj+Pz8PDhQKkWGZKiD12D0tlqtGo0GjBYQ5w8+iFKpBEmkYCaBmGiSJKHqNuyCit1QskIQBJlMxvP8wsICBFVA89JUKrV0hJWSArQI3MUgKSCDVK1Wg60RJMX61bnCCT41Ozk2QTad6tu9P5W89NQ9U9vQ3dJo166DjoCaTlWDSNarPgqE4Ph8vsePH/f3909NTcXjcaguyjBMTU3NkSNHzp07t2/fPofDIZPJCoVCf3//z3/+8/Pnz0vxEKAhfvCDH0xMTJw/f97r9YqiSJJkR8+ub519u6amZiU2Xbgwmpub4c/tqCSQjNg8YN6BO7lyiQCmC+kFOEdggigWi2C6KBQKUKNXJpOBLoa2Q6lUamFhwe/3v2rTwsp3X+rAhpQ2ySAMizOdTme1WtVqNThEYGKCMBGIcZOKWyCHyLoh8qVcgTQ17z+w25iZu/nFZYVKca7XJl8SDC8F/S09R7FY3LS0HY1G09vb+wLbFVzbcLkKggB2ONgF1zmGYaIoJpPJZDKJfVN2QIChXC4H72E2m4UQ+nK5HI/Hg8HgSvTBCql6X1SFIAgQ4ljFygFuAbivpXWFVquFNBNQTvBfoHoE9ryYhFarValUFovFYrHADQ79w6QZIJPJLF1LVEqKfD6fSqWgrgDHcVJRCrBSgBNkzRFRpEylM7Dk2HysQHftOXYgOH9lfNy9sK+rXpuOZ3lSrlax9KozNiBapequdZleeJ4Ph8PPnj178ODB4OBgMBgsl8tw0ep0ur6+vrNnz544caKlpUWlUpEkmc/nl9MQP/rRjwiCuHTpUn9//1crNJwUFPaiqqZMUCvUO+Bf275KAv3ksrYAACAASURBVMmI1480yyyiUltIBozGxsY9e/YwDAPhnMlk0ufzQWE1iLQAowUsQKX2sq+KFEZauTGdToMFtdL8q1QqIdgCQuIBSEaVtAU4lZG2WA2UQl9j1tpIR/epdwyejz988ODG47Y6Y7t+i9qBQM28+JjKVeai19KfIKaxbz4glUplb29vXV0dyIh0Og0yAgrOLiwsZLNZnufzqeD0yNAzb0ZuddVb1WQ2NDsbSAoKrV7JYCVcFMrlciAQCAQCqxbflVTNqKpKKpXy+XzSjVD1BThEQExI5WSgYIzkgln0vxat2uEbg9eFQiGVSs3NzUlWCr1er1arwS6i0WigKMWr35s4o7E21ZufznjGJgNNh2rb61QTT2Y8/lAbmRkYmskYu0/tb7Eq6a12z4Oha3Jy8uHDh0+ePIHGWnClyWQyh8Nx6NCh73//+4cOHYKwGBzH8/n8w4cPF2mIxsZG0BAWi+Xjjz++fv065H3gON7aVGtkitdu9mv0xv2NenZlX0GlkoA4iW2kJJCM2LpIttPKjTqdrra2VqfTLYrrBA0B2WKQXeb1er1eL/QLgI3QcMjv968ueUSa2SvJZrOhUEhSQmC6gBq3LMtK5b0hHAwczJLdAmmLl0Oqatrb7TcvTjyaOPzDth57/4WJwYHZ/U160/YNdqj80Vd4AcCTUqvVNjU1dXd3w0YQ1tJrWOILAp+eH7rw63/53YNUw/Fvf/dIs6bgvvfZlQd+xrV3b5dDJSPFcrk8Pz8/Pz/P87wgCLlcTnKmgBzHnnvNIaKzUCj4fL5Xqn5Rlap30FJwHE8mk5JVQzJsSNHcK3/HyvVAsVhMp9OgY6AKDiR6VIZTrLytOc7qXD3dXbPe60+ufpK2U9Oe2RBpiEYjzOzdy3cjdbLOrgazgt4KsZaiKCYSiWQyCalz0DJtcnIymUzCNQOhjvX19b29vbt27WIYZnJycnp6GsOwUqk0Ozt78eLFK1euLNUQdXV1IyMjn3322eTkJHzJrta2n/37n+yyslfvPLxy2WQxHm/RrlTxg5KQ/JLbSEkgGbHNkALRF22Hjk1iBWDsleQFmC6gGLvH44F6fBDhBSFdkUhkFe1Mq06LxWIR5lxJW1AUBSYKyW4BjmQwsVZqizezTVF1+HRwIZGVNezvtk5cun3hS83x+npzNJkIxAuiacst9DaSymCjl2U2imxOpZKpdXp1c3NrT3enmXWqsuHITa+6tuXwqS6bksYrnuggwSVniiQjeJ4HDwKGYfl83uv1hkIh7HksJwQdS/8XXmcyGQhpyufzwWAwEoms7jEg3b8vPexVTytJilKplEwmwUrBsiyEUBiNRpfLZbfbVxTzhNMqe+vhY/FE4sLlz38/MT6bMu7dQ/CZSDASTmONDENjpWKRx3CapYVMPJbnCYV2uQDhDUUQhGg06na7Z2dnh4eH3W53IpGQInJYlrXb7U1NTS6Xy2QypVKpJ0+ewPUGcvPRo0ejo6OpVAqOBw3x4x//uL6+PplMXr169d69exDei3Oqg99657sf/IWTS5cjv/50fODhVJerz7pyyyHYd1taWuRyeTgcXkuyNBiGN6d5ApIRO4SqpgtI7Kx0juRyuV27dvl8PvBDS7aKWCwG1otUKiUFdUJx31XEdS6dAaE2F1SflQbJsiwYbyVtIZPJpHpZkpfkza1swSc8dy/1B2VNHQcPuf9w8dotw3dOHT2l0tpkb+TXsVIIgqRImZJjNAo5y1AUq61p62z3Y0mxXMZwkqYXPcqUSmVVP0vlC8kMABUnJRmRzWYLhQL4ESQZkcvloLcnpHQWi0WI3RMEAe47CGyqlCDw5+Z8P1iF/wgS1DOZzMLCgkajgeo4KzwJTiktrfvfJmmWYz6n9ET9gb56dflJNMXjNWaNii1FJwYmF8qmru6awsS12xNZy553j7YY2dewTgDPVygUcrvd0WgUPjtN03q93uVyNTU12e12mUwGMbMwfYGGePz48fj4eCaTqdQQf/Znf+ZyuXAcv3fv3kcffTQ3N/eVW8TW2rLnuLPGIRMybb31V6amxkb88R6L+VUkP9gkWltbq4ZLvxKw5lzjSVYCkhHbHLGY8E8M9I/Ol7XNPV2tdSYFvfi6qVQY4GqF9DPsmxNlPB6fmpoKhUJLZQSU94FlFvhNoJ+yFG//9XCWF79L7bHZbDaXy0kBHGC0gHaLlTJCLpeD6VXykkih79iKreLbEoLCE9ODT4p4+0/e+u4J/88vPHrk7vjL9ztaDKgc6fLgFKc0aLVGtkiJAjwvNLU9J09ZUoJ8uXTEV/KzVGbxLY1LkF5At3GoYL1IRvj9fnA4LpIRJElqNBqGYWAZCpIln8+DXonH49Bbslwuw3kW6fXl7r6XLkmle/MV069wglHbWg9+x1jXeyLGywwOQ+b+9RxPqiwmjRLPugev/eF2qoUyv6P23Lt8I9qgObCv6bXICBzHaZq2Wq16vT4ajYIXo7a2tqWlxel0Qhxl5fHlctnv90saAs5QaYegKMrtdl+8ePHx48dgylKRpL2hw9rYyZEkgctM9V37W7NBLJ3MCebFwvXlbC+jLJIR2xqxGHPf+/Sf//4Xn0/ktV1vfe9HH3znZI9Tw6z0EqwM14IiPEsb2IuiqNfrrVYrRBtJMiIUCuVyuUgkIlXRyefzyWRSytNbtJir/gG+KUGKxWI0Go1Go5XHQFL+Ihkhl8shQKwyuhO0xdIYtO0KpbI01sr7R+bmsspvn/jOtzz/76ej9wf8fU2aUjBaxFiNUcutwPksZSpW3bUB437NEDKN3WkwuKfjqVgqXzLIWILV2JvVcK2JgiBgOE6szydf9AVW/gmV36qmzxAEIaUGVMIwTF1dncFgWCoj4vG42+0Oh8OLZATUtpIW0FCEBhbfhUJBqrdbKBSWWsgr71D8edWsV70kcJJVm11dZheGYWJirFygZDKVSa2gS5mI3+tZwGpKhXQ4EItlCZZjlyxyXhVpnEt3vfjRS5KkyWRqbm5Op9MsyzY3N9fX15tMpqXpo9C6ZZGG0Ov1Z86cef/990FDZDKZGzdu3Lx5ExzBBEG0W+Rmq6ZMsiKGYQSltLWd+rYqWpJp3oBn7BvwEXcyhcjU4xv/f3vv1d3Gme571ls5J6CQAYI5k2IQRWXLstp2u8POs86sWWvuzufYX+FczNWsNWtmnTOzzzl79+7o3t1225IlS1akxCDmBBJEzhmoKtRclI2WbdmyRSoQqt8VBUBAoVD11r+e8H8+vr1ZZjsGXEj0/h9/iwD4F5fHvczRlTaZ3e08z3/DSsUwjHq9vr29vbW1VavVWjKilRZptaqazSNm7Bd6ogMF+l6F0cKsw/+GtjA36cm6CjOD8+RMVLNJpKUtjmH0glR6Robtq188vPPwxC+Ge/vdyN7+6k44KTcffv4oSQ9dvjzTKT6z1pIkSZfLJUnSt59qNBr7+/svYtNfJRjr7A561sJrB/uhaKcdExBgAJwkkGY9lziIpKuY6PI4ZJZAXuTBwDCM1+t9akIwn8+bdoffxnRt+bbzUrFYxHH824Ycpqowqy9NedGSEY1G49sywmxkNRWGGR0xczSNRsNMahzqO6MEI7AiB6MIpJWy0UTNIDxuCSmvxosa8DkECoEMAzrMXrfb7dPT00+t4npmQYmpHhiGIQjC6/WSJPnt1cDUEI8ePfqGhhgbGztx4oTT6TT7dVdXVz/++OPNzU1N0wAAw8ND/3Cha09HI6FEftZnQwGMc0pHvwJBEGQ0m7oBAfiopOvrhyUjjjN6Mb65vh0BwfP/+J//9/Pu7IPf/+b+1T85fJ4PJlwvPH1urndmwyfHca3HzSXsSSVhyghz1MiTRlumxVCj0fi2sHhmfZlZxflk8bwpd0wZ0YpPmFtoWmaZ/oOmsGhNZHhSYbx+wJRj4MLF0YN/vfXH/7e26yyE4tGqmM2l49mFW1e3RHj41HSn+NW6bBhNA3rad3nqZckkk8m04yRYjHP1DPXubn++dv2alnGRAKFsPaPDrubO53/89z98cYB3TF9+/93z41025sXNtWYYxix8/jahUOioDjnTbaL1zyfPxKfyDRlhnonmKYlhmMPhOMzGAFIKDHbYoluh7Y3lSnx5r9TkHA5Kz0bi2WQE3Xm8sCSO9fgVjnhuSwm73f5d5mYHBwe7u7vfU8hlNpHxPP9dZv8tDbG2tmZqCBiGZVkeHR0dGRmRJMkMeKTT6WvXrt27d88sZ1EU5ec//8Xf/e3MjT/em197vBLuPunGGg0DRkmaMKq5RPggWYEFt9dl48lXNf70hWLJiOOMWiukc2VMDo6OjAwO+mBb/SDyf99ZurUyPegMvqoyvKf6IJnKoDXJrOXXWSwWzTBsS3OYBRmlUslcDlrhipbv53d9rhkdqdfrT2oLGIYpimoNEHkybsEwjKktzBkKrUkirRTAa6EtMKFz5vL72dL//O2nv/k8FE6xJ6YopFZIJss14OFoHKjVUk0HCI7DjVI2X4dJXpbo5zf+aRNg2t4zNjYR2f/D9d/fDceKhmf6Z/8JO4Xc++yzT+7vEO5y/SoOcPJvLox4efyN2lem2n5R/rMI5x0/NZsqffrgT//Xn9YfLlU73rOLSGU5lkln8+juo6t/KMVnzl8+P+RiXs0U0O9JfHw7DmFqiJGRkeHhYUmSTMHdaDTu3Lnz4Ycf7u3tmWYek5OTV65c8Q8PnU1nQ79f/NOvsPwwlS8i9q7xk13w9o3f/+rD22EkOHPpypUL411Orv16rCwZcZxpamrDwARRcYgsBsOEo29iwDF/P7R+kDnT4f2BvicvhVZG8xu3xWYTWsvz21QSrSjFkw+adr+m3dA3+lqf9A/4Bs1m0yzmePJBczNa8QkzRGFOgDQfbD3Vmj8Cvj5z4QXuqW8CENY79s4vEYYRP7q5VvaePtVvg/YTJQjnHU4OraVC84/CDd7bG0AiCw/WS8LA7JkTAfHNujY+BZR19M5efEer13+XyhZLBInB9XwymgH2sZ/+4oMhNLK2vjT3MOCxDzt/SH2JxQ8Dpuw9Zy4aOH7z6se7Sxwuu2RKy0dzBtl56tLlM0pidfHubc7lfHfcSb5Oe73RaMRisfn5+fX1dbO6q6UhRkZGZFk29YdhGKFQ6C9/+cv8/LzZ5BkIBN59993R0VGMlDpOnjm9G/mfH/3b//FxttCwTfzsfxMgbv6za5/c2kQDGnmdwBGcuDzeIb5OS/NRYMmIH4rZx/VdAwafilkz+OQ/zWl+h9mMfD6fSqW+DNzVssWahmEIopXzqaSOAZVQejzcdulgZy+GCU8JVtdqtUwmk8vlnvrOpVLpqZtn+lk9dVaQ2Vt/+MFdpsueOcvDbrcbX3kXPhm9MDfPFBatwgvTVutHNVibLf5P1RZmxML84xvawvyDJEmz5OIlDrsCuOAde+tv3f2nklVMtIvq2mquhtI+u0SBYnTr3tUvss5JkrCFl+duJ5149/hoQHxZ2/Yag1C24InL/6AEhy7EqpS3L0DFrgPG7gtMnrp41hnBUr9emp8PTfXIDhwBAEYsMXEkAIx1D57/eaA3qNC/34ADIlzZiucRpXv6/KV3/Acg8j8WFx9sT/UrXhK8zFKlVlzzScsy829zwufi4uLa2lpLQ9hstuHh4eHh4Sc1RLFYvHnz5s2bN82XEQRx+vTpS5cuiaIIAIzLnbM//WndqP/6jzfKOsbSqFbMxjKQOPzOB39zSkmHNleX5n0u+7Sfba+DzZIRP5RMJjM3N/ddyc6nYhhGJBJp/VNVVdM67TCbsbi4eOPGjS/v6fVSbHV5I9GM/X+lnU94DAGGXs3HoqnGndWlT+30U2SEpmnFYvGpHermJfmpWoGmaXOA4befMgcfvxyTE+grs0JzEIOpJ1o2Mofku7SFqSpamFUXvb29raF8LwWAkIKrU3BBUFMt7jSbTYRmOIHE9FImFUtk6kKjXipmkqVG00eT+KuJF7+GwBhrD45fCI5DEARp2bok2uUaTWAIKbm8Nm1uaTUUTQf0Rq2BCF6/ZIUljgiAYIyjb+b9D4Iawxn7d3KZMhIgSILkZUVQy5tbu5FUP01oGkxyAkv+gBzckyVTT8Ygn4xK5vN5s767lQZtOYzlcrlcLmfefrTmuZhlp+b9YTgcDoVCrS4zQRAGBwef1BAQBGmatrq6evXq1a2tLXNOzcjIyPvvv9/V1fVVgRHG+0fe+Se5c/TcXoXydneJmfsAU7r8U+cuXFAS9wu/Xwgtb8f6XZ0sAgEAI22SfLRkxA8ln8/n8/nDvIOmaaFQKBQKHdUm/ZWVuWtH/6YWUGv+05MPIgjCsuzLlRF/BQCEZAU7zxShRq1aLKfTeRWVeR5Vi5m8CvE4ARu6qqOvhQ3xawXC2DxdslDIZ9LZcoABeL1YLiUz8f3YvQdrDd+lv78w6GBw/MXVXL5RAIx1D40rhlHc3tERoEF6tVSs1uq1ei2f1+OhpdvrG2tx1D91fmrQZ6NQ0PzKYdOUBWaPSesOwbzYm00opqVNtVo1W2FbJduxWCyRSNTrdbN1xZzDDj0x6c2MbpraolWqZeqJSqXSihwDAERRNDtvn6ylSCaTjx49unv3rqk2nE7ne++9d+7cua/fW6KUHBg5GxiBIKhZz2qCU5QaFI4RFCfbFKq8GN/djfbQtKaitKzYuR/cnP86Y8kIC4vjBIAx3tkx0vN4LrP+6H65vLyZ0ogOmWrW9mPpdLSws7q4qOBD3T4Hb91cfw2ACd7hPmfoTmj+4UODPVgJh1NsoJZPbMzd+GNY1ux4fay3q9NnZ9tiaT9SWgGA1iNmXPDJCWrmhb/1GvOaDRn1MiL4B9yFYmju89s5sLRysB+myIfXN5evf7HRUB6sLU4NdztZrFIqlkqllld3sVjM5/OtS7uu6y190LLNMDtaW7FJUzeY82BbTl/P8WVNHzxz1HvrQU3TFhYWHjx4EAqFNE3DcXxiYuLSpUsul+s7yzYBSkqOYCeXyKZjqYKL0lWomCgmDvY3m+mNHVUZPXNmPCBT2LHv3rBkhIXFsQJGaSU4ce5E7trtuT/PbW/u5u1neAFVC9lMtVLR0qHHd41quXr61ES3g35WesOcTH1cej7NmSzP/d8BLnZMjI4ko598/u93wtsra43O93hMy8UKmXQNjqzd/TQWCs9eujTVZXvBox/MYZvfrigy7ape0Iead/A/5JVmO3QrGECS5O7ubstz03yNWatkek6YTaRmTrB14c9kMul0utlsGmopGT/Yjy7trH3Oo7nQVqSM78fWtHQslq1tRveX79hsAoWYzvymjDCjgKVS6TBDJZ6bb5vZmCa/ZvGEuQ8DgcDly5dHRka+7/cCMM45gsOBpRu787fuanJyZSuSQIVq9mBj7ua1AzquIsbJoa6ARxGoY52ItGTEd4KiKMdxNpvtVW/I1zCL/r617htNtVbK54sNQPEih2vFbKEKUZLEYs1qsawBguZZ4vuXRlVVzbDht58y/fieKrrNwOAhvtDxw2wifXWfDxBC8A2fegshCOKzSiZPOESRNqr7RYP1jU2dm/HpBztrC4uyUxG65WeUw7vd7tnZ2R845/qVg+O40+k8xBsgtKN3+nSpkE7826pK+gampntEfSlbxj0DJ8/OehJ37929Ttjd9rNBAQHwi7tH7OjoIEny2zU95tyv57uHfiahUGh1dfWZ1dAwDNM0jaJoK85P0/QXX3xBEEQrIwB9tVy0fL5NA00zGGB+L7P2+Ts+7isDfAiqlgrZ+Ks3QDM9fFumMjzPP1kKVq/Xw+Hw/v6+mSWhafrUqVMXLlyQZfl7a0QBgvOe/vHhUPzjW7+7G9mPpnT3mRlKq+7E89l09WB9/kYpG584eXqi28Ud4+YqS0Z8JzzPj4+Pd3V1veoN+RqBQGBwcJBl2a89amjVzMHa/MOVFOofnRzkktc/vLmD9b33wYyrunn7fthwDp4+1S1/b7A2mUyurKyYYwy/gSAIiqI8tdfcdMV+aSWWrwMAgMNdzA4PjNL24Ph5m9vtdniWKrINa6TTxSbp6p2YOd1VuZ/7dG53dzPW5xcJ/HvL4SVJeqq7ZdsCU7bOiXf+UXQPXcxAtt5hV/Xzz4oqExydmj0X3E8+nr+9+mhruo9vkjDFS8+M5jwnDofjqUZPjUYjFAo9WZd9hCQSiYWFBfN2/3swfV8QBHmqf/axxmw7N+dvPVl4YeoGt9utKIppXsfzvCAI5rO6rqdSqd3d3Uwmo+s6hmGjo6Pvvvtud3f3s9u1YIy2d0yemclno/H9PdrVf2KkV0EP5nXW2Tdycsxb3Q/N3cc5m8z3KzTy2vrgPQNLRnwnNE0Hg8FXvRXfZGxs7OLFi7Isf/3hZqOYPtjZDhcRe0ens/qotpln6DM//6df9lQXOP1amB07/9MrgzxiABj9jiKyUCgkCMJTPXqdTmcwGHzqLfje3t7u7u7hGz4tfiwAwVlbx9i5y46SpuV2NpO5UpVBMYLiYJJC8nup8H40zFUhHWFE2cZTb7wl1VeglOQbPOPuNyAA6wc3EqkKIPrdEkegKApVSsVkKr6zHK9qYu/0SOANM6eCIKjZbB7trNEnzejMsgMYhhq1Wl1tQjAwtIaq6s0XcB9imtian26a05hd5WZPR7lcNhUVgiA2m62vr6+np0eWZYqiWmP/TMrlcigUMqciQxDkdDrfeeedM2fOPNMw1ARGKVtw/J2/k4Ijb6UhweNms/ObZSD4BiZnT3kiWnJrfn9jI9wlwzJDMwyNH8OKJktGtAcwzimdY0onBEGQ0YhLbrsUbwIIoJRoU9j6cmpnJxIXs+WiTjs6/PYXdZdl8TIBMM7Inm5eraQ3IhhOknCzUa2Uqs1GqVLPNlOh5Vt7id0k8I/PnJ0aDCg8cQxXqBcDgM20YMMwYNru5XxOgdSqhVSuWtExpFmKrtzdJ6vuDi9vs5ZI6KuiDTOzad7Tt1xfYRg25wa3Lr2t0Vlm/aPpP20mQarVKkEQBAZlkvF0UQMoopdzhWKloR9KRqAo+mRLtrmpsiw7HA673S7LcjKZTCQSMAxXKpW9vb1cLmeGIkwNYfZ22my2bxcJqaoaiUS2t7dzuVyz2SRJ8sSJExcvXnS73T+4ogjAKC37+qc9vQZo1jPbN4rFBiAFSWRpHGk2tWKpED/YXsmHKHtnX2/ARh2PSqUnsM6RNgTlHN199vW1VCic6vI1GnohkYeisRCdnH+YoAffeudsv4ujCOv29PgDAIxgOMU7uybOn60tx7XY2sNKeW07nK9KjXx0e+/hfLgZLek4gOGpoYDCHnrCYnuBicHp81cckN/JgspuMl5QYdlhE9liI5UvhdNlzZCAbkBvkjkVgiAMw5Akafq3QhBkztrwer1mzQSKoqbBPIZhCIKY43YJgmjNAzOzBqVSySyTNN+2UChEo9FsNlsul1AEKhcrtZoOYLjZ0H5sKIIgCJqmTVM4DMMwDBMEweFwKIpis9lM3WBOADan91EUNTc3Nzc3l8/nV1dX0+m0aYZraoihoaGhoaGnagizsjIUCsXjcTO/EwgELl26NDw8/KNnmIEvlSuAIISkZRflcnCYUS0Wi1W9icCN1N5WKJdSaYdToujjdpJaMqL9AICQO8b7XQdzi9evNj2VpceheLNfK6X2Hj/4fEmNGgRcHOnr6+5wcMcxgGbxTQBKyf6xc2/zysLtL+59dnVhca/MDgadLptWtrl0xIbVY+uPlwSBY7uVVzTL4DUF5XzjZ97p0RGSUhPZeKYOGMVtkzidhYxMuVwtFdOlSFbFeJdPYdsomAMAMC+xEAQ1m03Ti9Z8iqKovr6+QCBgqgQIgux2++TkZE9Pj5mbaDabpiW86dlgNmu03NtyuVy1Wk0kEvv7+2Zzhzkop1qt5vP5bDb7owpIMQxrKQbT+Q3DMFmWPR6P0+l0OBzmt5Akied5cywfy7IMw+A43pIFjUZjY2Oj0Wjs7OwsLy8nk0lzFoYZh/guDQF95ZAdDofNLA/LsidPnjx37pzdbv/+oeTfDYzRts7RmYui7vSzcD2Uzxc0XBbsAplPF6KZTKZc13QCMSAYeT2G+vwgLBnRhgCEtHWNTY0mPvz483/9ZGM/pjrPnZfgejhVLlWRWnZ/8U4xnmtA58a6jmEAzeLbAASjJHewTy8nI6G1dV6xD4wM9HXJxVoiUq2SDAWgciwWSxa9MsNa5/wTAISSFMowDL2gaRrC2hWnQ2CpIo3p6Uomube8ef+T5bLrxE9+dmbQyRyXZf2ZIAjS1dU1NDQEQVCtVltcXFxfXzefIgjC5/MNDg6aV0pd151OZ39/v9PpLJVKhULB9KQ3GzEymczBwUEsFqtUKqZiMDMXuVwum83+6JIpAKMYTmIoguA0z/McLQhSoCPQ4fMoimJOEmZZlud5SZI4juM4rmUs+8xrrqZp5rAeU0PIsvw9uQwTsxLTzNpAENTX1/f22293d3cfYrAZQCnJPzAh+XUUN4o7uXy+CtOsTZHkBkFrpWalUkrHcoWCRssOxcaRx8MMzVpS2hIY51xDM6fL5Xy5kDck18zMiJeIL9ZJe+/YmZkhEHq8OD/PudxuMcBaOqI9ACgpegZmr+C27tkSzNtEvJ7OYDQOp5NpQ/EFPDSBGnrTgKC2uRgeGQBAMC54RyanRbJD4UmKZFlVLW4/vh1au/p51HGx51J71RADAOx2e29vLwRB5XJ5b2+v9ZRZYplMJs1OjUajEQ6HY7EYgiAHBwfxeNwcm2fGGCqVimlB/cwGkG9gNkeYRiAkgTXr1XyuCtGi4uDwRrWqS90T40MdIoYJnSPjJwYDNp41AxJmF8lzfF+KogKBQCaTCYVCFEUNDQ19v4aAIAjDMI/H02g0aJoulUpTU1NTU1OtDo7nBCA4xckkBOmlktaEcVLiZJvE82mcMhqNQnx7YeXx/K4emHr7rZk+J3csMs+WXCQYZgAAIABJREFUjGhTAMYo3TNX/sHdO5NSKbuDLy1tFVRK6RwaHxuuquG5tdjudiI/7KFpxGyQJwjC5XI99c3MPOjL/QIWPxqA4KwSGJI8/c2mWogs3FndTxusTdKS2XIR4iSZx+rJ/Z2yhgmyLFu9G08CYFLuOPk20Q8oicFUkqbr2dUHm7vNTJ4dujjWrZBaIZsxEJKhyRfnl21akojiUyarGYZxhOegYRjZbNZsy6rVasVisfVUvV7f2toKh8NmL4Ou6+ZscU3TzFDEj1UMEARxHCcIgjk115x4R9N0IBDw+/2CIEg8VY3vPLy7VbX3jo3I2aUHjw6Y/rGpIbm6sZ7LlECTlDhZZMjnLOwBAEiS1N/f39HREQwGNzY2aJr+nlzGk+i6Pjg4mEwmTRnh8XieOlrox28TBMEYY/cNTE2KRKdTZBEGp/Bydm/x9vzGvbVKBzmoNo9NF70lI9oWAGOMzdcvefqgplqK3i/mqwbutskcjVSamlouF7PZZDwK8wwvigwGi6I4Pj7e39//7bcql8v5fL7NmsjbEwCjGI5CEMayAkMgdR2S7F5I208mE7lStaSG7t26vV5WBqfPnhrpcgltlO4/LABj5ECHCAEI0ookyTRTB48Xq97hcz8/P9svVLYffPF4r0x5BsdGBjocHPlCSkzMhP03XWEgCIIgkiSfjBkcEl3Xt7a2UqkU9NXkqtZTtVptf3/frJd8jnfGcdw0XWj1TWAY5vf7/X6/LMuiKJpZCYZhJEn68mU4KB2syNjnmw17R4eNyYQ3KzjDkvXc1urSmpGtQqCSGx8d7vbaaOw5djuCIN3d3YqimN8ul8thGPajzFvNacOCIBw2FPEkMM66uk+ctZWalChgBQZDm/GNR4vhdBV2TQ72BWRMK2TSMEYxDIm93kO8LBnR5phefACGCVZ0B0DAI5BQvVTIVzRdhqrJncchlekYGRv2CSRJflc0IpFIlMtlS0YcI2BC8HZ1D8RzS9F0LB4JR1UjNDzWGcQQvRrfWco2EBjFzox02BkrJtECICgCQZBh4LzkdvMcJ8pDs29fGHEb0c//8rt/u7Zek/tOxkrq26fH+5wM0lRVVYMQDMOOah+asfenGrRUKpUjj0Zks9lvP2VOo/iB78MwjCzLDMOYzRoEQYii6PV6fT6f2SjB8zxN0zabjed5U1iY0YivezM0KbR3ZDpTehjaW14KR7Ma08VQUC1VaqCkzIHC3sr9ShPD6YkehfrxwheGYVmWW147ZsXG89ZIHiEAITibg5EMAECtxhCQnlrf2MyJY++ePjXVxRQ2736+ctBg/UMnRvsDCv9ilOuRYMmINwGAEoJ/ePYduswFRFSN57K5BoLzEqnlI4treo31dXkEzCqSaCdgnHMHhyerlVs3rj2Kxku0otdrdYN29s++JSf3w5GNpTm7LLBdlonINwEIzrsGz731LjrADJ+f7BS0jbnV1f1Uk+91Cc3kzqNHsiCxfZKeCO/tFxC719/h4PB2V2NmmEEURbMd1Iw0OByOYDBot9vNmkeWZVttlqZiMHsrnnXfD2OMPTAyoULIw/tfrFSrsIMksEa1olKunqkz067c+tzCw3sEQcBjXW4bT+OHuTl/DQREi6/aQCGMFmXRLjOiy37i5NnZAZu2f+0vf/jNrZAud4bSJfXtM+NdCoMa9ZraNBDiNQtPWDLiTQDAGGPvGDypqAAxqpH1XKYEkT7FKXLFuJrJ5zPlmqrRTQOCYRhBXp+TzOIQAATnnIGBk00DhtlAAfN0UOm1u2ubKVhyu1g4v7b7aDnQ4xMQkSQohsKP/ZjBowMghNQxcfEXgTrGyg5K3S1XK4bcNTHzNz8fQUPLC3vb65texwAaejx3bV3vmSVsEisT7bP7zGIIiqLMGRMcxzkcDqfT6fF4vF6vGVowExPmH092Y37DAvKHfiKMszZv/xTFCQTMCNuajVbL6bJK2V3Brg77zs6t2Mqdz4rFQmZ8bOzESJdbbjPxi1KSf3D67Z/TWW7o5FiQayyENtYOqrjS6+O06PrCI5tNZLrlZmJ9M5xDPUM9Phv9+pyxlox4QwAIRnICCenVOgTjNGsj7Ios8A2C1nWjXC4kD7KpRBmXPYH2apB/owEIybu6xy8qXdMNrZFZ//S3i7fuxNhAvkso7+7kCDIR21tPr9cQpXd8wCcf7ymDRwmAMVpwdHAGBAGg5hGU4mkH7Qn29PXhWCa0tRTZDUXsUmI/tLMJiYNVrY36OHAcN0WDmY8gSdLv9586daq3t9eUDmaYwfR9OsI7ewBjFO/oGDqJMo6ueL4YWtyoA9rNCwTcqJYzVbVBNSvJ7Uf3ajpBnhsP2trqcAUYY++avCD3NlBBlJjaVrHYQKTeU5d+8XaXtvZwIbS+tt/hIsHBo5sfr4IxWlJEypIRFq8IGKNl39DsGUq1+2WOyhEMqmrF2PbC9tLdpYJ94t0PJOlZs0Atjg8AxijeRvGQXtYPMIQVld7BsV4p92i5WoWAVs/uPLyzEKP3yo2zE30Bp2C1b7T4asQnijGi7FHoFNJoGIjIUgRaCGciBwktmy2qkEwxJHYcrmh2u31wcPCZXg4URQWDQbfb3Rp32dHRMTk52dvb++L9kABKiJ4gJdlzUSgXjlbrfgcLGnu5co1y906ennUVF+d3F5Z93UGXSLWXAwqMMbJCSQYEANxUEZzkFLvU4Q929+NQZOlgYy0UG7ERkf2dlRDpKTVeqzaOtvohLJ4NQAjROzDNeVWUoeBCnGCJSiS8eHszsrBacJ0d0+uldLyBYBTHsZRlctk+wATv6Orq3KkUeYFjagRNk7xMgHIksrm+DioMXMsnp2ZPjQbtNHps3PNeCjDOKk5/JxtLxHZ2Y5iRz+QziXomkwLZfBkwflFkiOMgIwKBgM1me+bLWmbYrUeeO1XxPAAYIygMQ5HBifOct0rZOZDI5qsQpfg7uzvt+2t3F7f3EslcwcsYCMBoisDaZpUCADZ3MkxyiiI5ogWo2TAwUeAgtRbZT8b8xEG2XMPsLENatREWrxAAoyQj4LQBAaNWpTAUSu8shvaLBuGZGO5V9PCDDz/azmPukZMnJwcDdoZ4nQ5Xi+cGoIzSOz6SrM9v7C9vP97ONtggjzfzhXINc/T6XFh+9e4dnRSlc70y2TYL81EAUFJ0BQe6N2Lztz/81Vxtf3k1hg/1qsVCoaQSTk6gILWUL6IESbzWc2oYhmEY5lVvxQ8Dxmibt5t36FCzuLNTLDcQxilLNNps1Eq1OqrptWxoYTNeYnrG+jqc7Ou8258HgAsuv9e+k4pHw5E4Ui0U4tmils+kQDKvYazNJtDI66T1LRnxJvKV6kUxioTh4u5+KM6f+OnpkyNObfvqJx9eXS2x3oNcWYMunR4JyiRk6BqBMxTDot/SwJlM5vhYv7/ZIJQc6JsGCL1w/+ZjmMR4GwGqsYpOuYemz1z0pf7jN7dvfXJdpLWxbr/Hbo0DbQFQUnD3Dk8k45Hf/+nT++t5pu+SYqf0dLGO0oLA6Lmt+eWoLnUP9He/sOFnNE0Hg8EfPRHqKPguK4sXCoBRnEAhSFUZ3u7r6JI9Lh5RD3KFqor7KAoq7s7d+mxF7cuXLpwa6nTLx3LA9neCkHJgqLcj+tni7d/t3spv3l0tOM6q9UIxX2nSbolD6qVCTmU45vX41j9CRlAU1dnZSdP0i9sai2fi8/nMATlHAULyNsXnd3nz9t5T52b77JX5z/eSTbHzxFinrCc35lecNpHxYZXIbnS/Zh/o8/lk8uuFPbFY7Ig2xuJFAxBScHUM8YLANVF5R1UIPZWr65Ti9XiddhVrJNYW4ziS3h0Ymzk9O9kpW5GoLwEoLXqGTp5vGjDv7C3yPdMnOozNaMmgXAKPq+mVe7cfFNw67/bZ2Rfk9spx3Ojo6MDAwIt5++/DHMP98j/X/HDa3jl5HvNVSI8IRxZzJRXhZY5qlhPR7cerqRSoFnLJmTOz031uDtMrpboOYQx77ONpMC76B8cmU/EP//jnq3fXksz4mF/AK/ulusEpPF7avbOc1F3DZ8aCr0Ps8EfIiFd4HFu0wHH86E5pmOBdfSev/D01CvsGJzrl6qO6hvDBiTPvfTDB7M/fWdtd2w36BD61/vCLe8VO2hkMuCjya8fMy0uaWhwBAMEpTukYvfAW118o50L7dR2mBJEjtGqxUChDtMcugvTC3esNWrGd7RGfewRRuwEQgnN0Trxl75osayiB1g5uLVYagBIEBm6kE5FYRhd1yGg2NbUJIciXBvNHCIIgLz8k8BoAUFLy97BuHYIre2v5WgNmZZFF6plcrgJ4h8/NNcLL926RksiNOLSD+fntLNU9O9ZlZ1791fUwAISUfUOnLxkIzrhGytLghQmP+sXtqoH57AzIb9366EGpl+zr88vkqy+I/xEywiy9eXGbYvHSAQjOu/unBH8dImiG0qIkLSg24HE43QEHFp/fim7vR4fcaix2sJNQpZr2WpUHWzwnAKOVQI+oVhIg0jdoNDwuCa7upHMV3DN48q0r3Ynf/vrT+/eWZkY7BfvxXouPFABjFC+TnAQBLR/aLleqECmKAtEs5/LFGupgBBapJ/f2o3nE5vE6BQxq6hCC43j71AC+EgCM4QQGQU2dFhRPT48YcDF6aTtXQRy9kxfenWa2v7i6tjG/0x9kwP6DG/+xTp73Bjtsr1f1wHMAYJx1dJ24ZO+crmkIxel7HxXqDUxUZELLxKMHKaQTQhBDbagAAPSVNgtZtRFvNgDGaRajGAgCADRYm11yinlIr6gQxtBAayQj6Vi8eZCt1FCJ59or//gGA2CMIDDM1TN7heqrozYZHNxNFFTUHvD53fY6UU3G9kLJYq2E1FUDZUUetzzJvgQAAEEAI3lnR88gzfslTIuUynWYEwSRgsp7C59/tqZ3nbowZsvv7UXr0uCJkV43j7/6O8ZjD0zbeiZPo74K4WDLDzKZok702ZzegAvZvb68sL4TT3uwcCS8k3ZOqbViLo8iJCeQx/rABTBKcjLBGhBk1A+qmaIGaJfCwuW9ZFEDHYrA4lpme20vbdgHBgM8oqpqEybIl968YskIC+irrATK2L0+lxzaiu3t7DFkOhtNlxBHLq2n8g2EkiSBwV7YbEOLlw+MMY7ObpsBwVpkLZ1vwJTdIbKGWirXyhoNw83SwdL1+QTRM3t5wstYv/xfQSkpcOLc254CEAUjstMEgJV4gUW1YiK8tb0Ho8pifeGLTxbS3CTpD3o4qGZAGMOQVlziMMCE6OsZcurNRuxeNpNIFulKQ2vCKIEb5Womlk2lQCRbb1KSTNXD927uloShS9M9Mnnc97m5PsOU6OsemFQUDwflUmUVZR0OgYEq2/PXfveg1mcIVzz5uQfbFbH/7OlhD/tSdaslIyxawDjv7u/riezcfPi7/3azGF7eqEiTvXolXyyphCJwaKOYy6gMx5FWVKI9ADCKwhAEGbQt0DMyWelwSWg1Es82DCbgELBSdOWz/7iVDxQp5NxIp895zO/tjhAYZ+zeoOQygJbJUiTHUgyFIWo1ny6VG7iLhKuJ/Z31A2hwAkWNcnR5eTOFdYxP9nmsBphDAGCMIDFDrVQz2VK5UjZyyVgqIxQyxUS2wZRzmWqyoFMuxcaTxtre0joM9Q92H38ZAUEQBAFMDMy885NgBRGF0q0aTJKiU2BwtZjc217dgdjw+u3V2//133eU0/8wMNErNesqhJIs+3LCiJaMsPgrAGUcXcNTJ5ORX/3qxq2FGD0yFLSz+l6lrtE2Fq/u3/ksWhb7z00PeK3lsJ3AxK5Tl3/ZWcNkycgsxXIazjkcPFRYicTW1w7S2J9Bam/i7KWLk50iqlXLJRVCKZZ7ozMdAEZQGIEghJU9LsW5H8vHd3fr4XCm2GSHXV53ISxB+e3Yzk44GXTmwxublZKtq8PBWefNYTEalUIqr5F2h2QkHn/0m8XYwoOdcueA0cin8zWIVWQORciuzh6GcLwG5YdHBUAoW0e32ISg6g7L0iJHkBjarKRi8bIKu22Mlphf394u2s7hGKjE1ucWY4ZrZOZEh4i/+APOkhEWTwJQRumaeuuXJO8c2Mjg/olJnzq3VlVRRaKR8v7Da7cjDtDd3+XmLcPsNgKglD04IBkQ0JJ72WwFsF5FJhupaLyosV0jI71yZfPubZxT5DMeLbL6xeMU4hk+Ox4QXsIS9ZqDkLbOwbHxRPLO3K/mD3bXkpD3nCiyeBFgFAI1y7lorOC2+buGYOl1moJwjGlq9Vq1gctdY1PDQmn5s48+m48Ipy4GBSS/UNJgj0vA86H7jw+y8sRpO91GUhcABEURCAKyv9/LR6Ohvf3dZng9UjZYVyDgrS9QXH1tb29756DXX83uLqeKVHAgYMkIi5cPgHHW0TP1tn/0gg4QtLL7yfV6DeFtEmUUtjLpouHCcAzS1IZhAAx/QU3yFi8dgKAoBEEGJXl6B0dcAQcPlR7HSk2l/9RP3v+5N/anf7m6u7AUGuCpg42bf1mABsn+fq+AWwsIQkq+sZPnIAi++pfwmm6wkshToFaqG1LfzNSFMz1sYX17N00N9YqsVRtxBMCE5BmamKYathMzftArkvbBimPs0qiS+kOhBmjFwRrZ+Vsf3SqNeK9casfjExU6pk7PxD/+7Pp//y/Z7bn5UuB9p8Lj+xqEcUizlstEUx5PsL/PISn0yzng2nEvWxwSYPraExBk1POVbEk1SFFm0XokXVQNUeF5Us9uLWxH63L/sNNhnxgfUwFGk4RVf3lIPB4Pjr9aqwaM75h47+8dFVTi6luRbLlGBxVZlJ0gYNfXI9tb0X5fOpFO5NgurVHJZwwMe9OzGxAEcNbRdfKKMxhw2zz39wkPj6oHJVVnfAMnxvt9tWufPLgREtnR8ROvekvbApiUO2cui0MNiLGLVFegZ6KmoyRVXfvXXEOnFJeM15LxTKpMYThcy6VqBspKItFOhyjKeoYvXTEw7NonHy1WEVJw2TlQzWR1unP2wk/OjQi13YcHacjb4RRfTmGIJSMsvg+YloMDo7MiGZTQynJVBYxiE1i4Fn984/c30v536Z91VUuxWJodvHJmpktAIeivjR8Wz8Er33kwIXeN8Aak55bvRHPJZLVaqao6jMBIPVvMxDMpOpWpNnGPRNZiK1e383Rg5MyYX3jTlQSCEZyj+8RbP7GnGrSTjD4uNjTcLgg01EjGc/kSZMcIDECQYRiv/jc+7sA473TwBgQBAEEETrIQ1KzmSpmygfOKkzMKC/FiE/W4eSix+Ke7B3D36Q9mg8JxGJ/2Q4FJOTjx3n/qHulWqF9v0W4Bb+SSJUC5BiZPDrmr9z669fkqc7LvrZMvxyzSkhEW3wPAeN/kW1eCZQ1Dc/c0lCQkhWcJvZKJRkLRMhPburt37//8fx5zp8iTb0/V8oWyBmhJYqy4xDEGIBgGQWopHc8U6nojnzoIRRQ4HksmcmK1ks8m0w2Ulp082Uzl97ZShjTY733TZQQEmXVFPSNSl65XIykAozhFkwSoFtOZgkpwNokxCrGdnVyDtDndDqtO4pAA6Gu7DyCcc/DEFNTwuSh1L5qroZzLSavxB9f/8GnGXxSpt8c7vU6Jbh8tARCcFH1jF/8G6mvyClzeTZWbqNcmMUglnokVKw2ZInHEML4Uri9UuloywuJ7ASituCg7ZFQNQWREXsUQSC/l44lKE5PcMlqaO4hHSzLJEGgjtX331lrZfuLiuUFHe3RZvbk0a/lEoti09wbl8tq1//ogsfVgK+c4z+rVbKpiUE6bxDC0MTTU33Q5KEs1fglAUQxCUSA6OjuVjcVsIhIK69FsuUl3cFhx89qf/vzne3G27/R7P3vn7IhfsLo2jgyAy91nP2CG6oAwQnczDYP1uyWkshHLh/dijbnPfl/Y65+6cGmm3/VyHRVeLAAXA1On3E1Dzy0bEEqSNIXBjUI2XaghnCjRUPFgaydTQ0SX2ymxL+5ws2SExTMBAECAEH19AfveWnhv67GeW94raGSPxx/AdkQH/CgcWtvYHR8z6omd/SKXnBqwZMQxp6mpDR1Xgn0XTw/XH/77n24uZPip2V4fpT8oQhgvC0gpvLq0nmd7u0W+rRLPhwdgvG/izEy++riQih4UDnIqJNuxyu6Da/9x7V6c8Fe1PyMAIz84P+ikrJDEUQEwzuVjoWZlZ62gYpzN5WT0bDRZQe3909MjSm759s1PMVH+YNxFtZGQgGCMICBDZ+2+wR6xkA9t7XjoeKrSRP0SWti6+dGHH89F8d7Zt35y+fRoUHlB0RhLRlj8MBDWPTw9kyx9ev/qf/tkfWGh7HlLllk0W2+iEkMiajkaLwc8PX287Ocxa2087sCMZ/TUpWbJMTZxWvKzuPI4ww2emnBWHqSrMOWVuWZ2+9HtGykvPjwzGWw2m+DIh1EdYxBa6Zm6RPiyufj69UcoxTlsmBqPZjX79C9/8csxsPlgd+7WYodX7reGnx0hAAAIxgTv2PQpGu60Y7WVWEGTu6cvvfOea7+68duVh0u7pwddXupVb+hRAxDa2X/hypnG9f3s9k42lq4jKMs2og8/v/qXO3uYW8MQDMdw4uyIT3gRi7MlIyx+IDBp6zp1HsIw/NrHeyssIjptHFzbzNRR/8zln1855dUe317crfnePitS1gXluAMz7pGLv/DWIIxXRN+V/6X/vA7BzeLaJ9lSA+EEkdDLyVSlBmE4VM/FIhBKCzaRtrIbXwFQRgl22lUZLcfidMnul/E4wrqDQ7NnLp0TfLV/+dXK4sPd8a4xG2xAAEYsDXZEAFzuPvczZkRFsMb6p1kVcA6PzFIETqG1ciGdLdU0FdZ0AyAYhrbPbgeEFJy89IEYzWR2H6wZlCDbGb20kgXK8Lvv/WRKzO+E1hYXPU551MMcfTDGkhEWPxiAMc6+Mz/19QTt7IdrWqdC6cVkoYlKXaMTA1700R/u3f7CmBz72cV2Chq+sQCMdzp482+cZnGo2SjlGg0VYQSbnYVrkXQFJl0SqydXb62EdMfI6dkRr4Aj7bM0HxYAAC4EJi+8312DsGY8bLc7CzSJI6Tk7bQ15jaW1qNv+7VKUUVFl6e9WhJfJV9lN4obi1WDkhxuB0dolVKmVC1BkF7L7C4urh9UGF/fUJ/fRmPtcrgChBR9g5wjz6jZZL2TDIpUFqYc3b7p8xcvClG0dG11d2N7sNMRYAEA4Eijh5aMsPhRAASjlZ7pn7wfqBM2Sp3LVJqIIIksVo2nY4WK7mBpHNbUuqYDBEPbSO5bAARjXD1js+fygo/WSrlcA6UFgWwk1+59djXKRVXEODkY9DllxiodbAFQ2uahDcOow91BaW05k0jna04YQerpQvkgGQnF7tzcNrrP/fJCn52yRoofEQAACMalwMTpM3aiy8aAciQTL2iQnTJS659++pvfP0gpJ9/55S/fOz8ckKg22ukAwXnP0IV3nWW1WUstO20yzBI4QQuSKOPpjcRWKNpJMg0d4ySbzGBHpVwtGWHxYwEo4+ofsRtQs7KN4SRF4gjUrObSiWwFcDYbA+V2HofCWUPyd3b6FI60It3tAUBwwTd85n1XudHIbd3KV3TczXNoIxePZRPR4tbCHTifGpqeOTXW+aJKuY4rAABC6hgYcOzeX3t0361Sm+uR/bRtqJqJhu598mlqrSZT9dGuYIdXthJDRwQg7d1n3+eLTYwm9Vguna0BimO09PbDxZ1kQ/LomfXlJZ5AT3Q6+HYSvjBKCU4v31TLIN3hJbOVZCJbtoGmppcSpcTB/uNqbiNiBMZOnhn2ieTR3Ogh//zP/3z4d7F4wwAwgiAIAkP1Qnw/lq7zHqkZnrs5n5MnZkbonY/++7/828cP93Mqxks2ibda5NsEABCMZDiaAKXw+spGEvEN9nlA7NFCxPCOzp7wGam97ZTOOlyedmrQPxoAStAcqCRX527d+Ozq9burRXni4mSgtvLF9eU8RpO1VCxZx2XFLlDWrjsaAEKwHMvSmFE6WH5wd6Uo9PUH8OjSRs179md/904XdLC69Hi/ivF2u8i0l/EJAAAGMKRXEzv78URBVbPbjx6thHXFKzajC9c+W4xVAEMTBEFSJH745dmSERbPDUAIRuQRrVCDUKO8O/8wZHSeHrMf3Pjtb+6Em6yA10plg5adTplpmwykBQAAQM2mDsGs3e8VofTW/GbNOXHp55f6ydz+4808ZvN1d9jaKVR8NMA4I4kMUj5YW1jcrcnjl372Th+8+sXtbajv8rtnvdr2o82EJnb0uC0vr6MDAABBhlYrZJIVyt3VoYDYehjyTV+8ctLb3Ju/9vmjfUjp6+t0se02HgjACI4TWCW+/eiLz65/fmdxryYER/ptzf3llc2kDkONfK7QwHhZ4mn8kMuzldSwOAQI4+yfeYfqLFXi85uE6OQ9IlzdyuhC76W//afL3sLqfGhpfs3rlnqsxrb2AaCk1DFySvBXG5XYwmqxDpGCLHEMgUKGViiXC5VKuYwCzYBxiiIxKxb1FQhl751973+1+Uc3c6i7b9RTuJooNeXOE7Onzzn2ko9+vfR4df+t8QBrLctHCIBRxj0w+x6dh3C0Ed4RhTIGIAQjGZaBDaNe0w2oHY9QgOC8o3vq1EwuGYns7VHOwbH+oARvFVXa3TMxPWQrRDfvPaQEWRa65cMV+FrHq8WhACjr6GSUhqClzxU8sF9hKwghdzhHpqdnA0U49Ycb+xtrB4N+RgAQgFEUteYJtAEAwWnBQTHlVBZGCV4hHA4Og2qlYrGmNZt6Kb41N7d5UKG8Q2NDPV6ZsoJRXwJQ2hYcO+cfMSAAGqFr8ZwOc06XxOIogoJquZwvVFUDasvL2qsDoJTo7mcdzXohWnF6qP1cdH8vipcLNcNgFUHi6fa8DgIYZx3dJ9/9R3vX5F4OsXkVIj73qIYIgcHp06eknVuRj7fbrqmqAAAGaUlEQVR3dyLZ0Q6ZONRtXnvuPouXCgCAkDqn3+IHNAgqb7kcUh6DYQSjBclOFTdSm7vRYS5fqkCs4vHJVta8XQAITts6R6bPCU2vl4Fr6Xy+qGI8rKdWvrj/508fl4XBS1UDu3Ciy8kTlpBoAWAEgSCoqekQQssexaNwmJovpIu1KoZiiJXReAEAGEWNRqVarAGWBZHowtWPdutbq4ma0GMTGbJtW9QBgrNKx4jNP2RARjX1eK9aaBCMaJNYCkcMXS0Vy+W6qh/yUywZYXEkAIyxOWmjqRbKXUExtJfYj2b8slrXyslcOhrZQnMPH+fk4TNXTvU42KOo6rF49QCMtgVHTtqCGkpA5d1crlCBaTuqpfc3dpM1whPAG4mNpQXCGOoNOEUGt8olngQQtuDJCxe9WMDOQJW9TLKkIz6Jpy0P2BeD0VTL2fBWOAeRoliOLt5eWwuXHedEO9/uRa3AFK46gAyY4GQH47CxaDNbLOXrOgMfQbOGJSMsjgwAAIzS9uDggBx9PH/z07qUnV+JJhB7vRjfWZ67voHFGzRUGu7t6fTZWauOrA0ACEHzBAVBerGkahBCCDzHQI1kHfMMTb/37oScW7t/8+pBtHjp8skBr2Wx9CQAF/2Tb3H9TYTAtf1sJldDWNkuMu1W6/e6AGAYwzG9VKhAnqGxINcsxrIJmbfJDPlmHJYAZ13dI+dVBfG6aFDbzhdLOq6wHIUfNhhjyQiLowTAKKt0TU2Ek//xySf/Y3MnnNW952QB1HdTlWKZKia2Ht7NxYvq+ZPDHZIV6G4TAATBGCN7+06M8bjDCVIJllM8no6eHttB9OYfb6+k6M6RwW63YP3iXwNgtE2hDUNrpDRNRxnB5pT4N+Sa9vIBMM45Ogd7NqMr+ViOJEiOlYFdUYQ2rYz4FjBGK52DM0pQx7BmfjOXLzcQmhctGWHx2gEQgveOTJ0tF/LFfMFwBGZPDLmw3fka5uiePD3dre4vLS8+FBxuB+9mrMOvXYAJzt0zeV4u6nAzvhJxSAUc1iEEJwkMh3UIMmBgFdc+HQDBCCn7+0/MSHyHjbFyGi8KgBC8u29iJle98dnNTxdXQgV5ctbreIMG1JrBQxqC1EJOb8IYK0k2iSUO7WZpreMWRw1ACN49cPo9wTt4Pq2zio3KLswXIcrVPXpirDurbS/eD4fCqfKgJSPaCIBSguLmZLVe1BOSzG/EU3s7IaiSzGsG4RQFnrTqYb4DACO00jN5mS3DrM0aa/cCAQgpuvqmZzS1mC9VkGbH2Eino+0cI34AMM7aOgbHKg4mqLCHl1HWOm7xAgAIwSnBIVvAgIx6fidWydVRymW3cSSSa2pqpVyt1tXmq95Ki6MFwDCsq9ViqaYjGFqOLH+eWkptriTq3mFeoIk2r2M7DAAhOaWDskEAtG3PwOsBgDFa9o+c/5kcPJFSKYff/0YW7CAE7+6fIp1lwNvYw8e/LBlh8cIAMAygJgxggrG7vLxP4XCjXCxmq7qOWFOI2hS9korFotkmJ/Jqbm9xbTeu+4ZEkbdGqzwDgKCWhHgZABilRGcXrwShIx50eYwAKCXZCcFmAAAf/sS0ZITFiwWglC04fOGyUBf9PKJu5/IFFZVZniGtY6/9AAhO0ijcrKiQPTAQlBuZXDGlKKLI4FYwwuL14cseyDcZAMNHtQespdzixQIQnHMER1mnBiN6aS+fK9QBycsiQ7zhZ3F7AhO8u7On7yD9KJrYrZSyVZjk7HaJIzHr57awaE+sQKPFiwbAKE7xAsdSGAzBOMXb7IrMUZZvRFsC45yna+zkWJ9U3bn/xb2VOOBtToUjrAYEC4s2xYpGWLw0YJyxdQxPnuYhv1ew2uPbFIDgnCs4NKtrWkWjXLpvZqjLyb2JdWwWFm8GwDCMV70NFm8KRlOtFIv5sk6ynGj57LQzht6o5JKpdFElRLti4+lD96ZbWFi8nlgywuKlYhiGYRgQALDlRtT2fPVbW85TFhZtjCUjLCwsLCwsLJ4TK9JoYWFhYWFh8ZxYMsLCwsLCwsLiObFkhIWFhYWFhcVzYskICwsLCwsLi+fEkhEWFhYWFhYWz4klIywsLCwsLCyeE0tGWFhYWFhYWDwnloywsLCwsLCweE4sGWFhYWFhYWHxnFgywsLCwsLCwuI5sWSEhYWFhYWFxXNiyQgLCwsLCwuL5+T/B18I264VfuL7AAAAAElFTkSuQmCC"}}},{"cell_type":"markdown","source":"## **- sequential api**","metadata":{}},{"cell_type":"markdown","source":"lenet-1를 sequential하게 작성해봅시다.","metadata":{}},{"cell_type":"markdown","source":"먼저 keras의 sequential을 model로 지정해줍니다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.Sequential()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:22.946432Z","iopub.execute_input":"2022-08-14T21:20:22.947085Z","iopub.status.idle":"2022-08-14T21:20:22.954203Z","shell.execute_reply.started":"2022-08-14T21:20:22.947048Z","shell.execute_reply":"2022-08-14T21:20:22.953310Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- input layer**","metadata":{}},{"cell_type":"markdown","source":"그리고 입력을 받을 layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Input(shape=[28, 28, 1]))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:22.955477Z","iopub.execute_input":"2022-08-14T21:20:22.955831Z","iopub.status.idle":"2022-08-14T21:20:22.966663Z","shell.execute_reply.started":"2022-08-14T21:20:22.955802Z","shell.execute_reply":"2022-08-14T21:20:22.965439Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 특징 추출**","metadata":{}},{"cell_type":"markdown","source":"lenet-1의 형태를 참고하여 convolution layer와 pooling layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Conv2D(filters=4, kernel_size=(5, 5), padding='valid', activation='tanh'))\nmodel.add(tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2))\nmodel.add(tf.keras.layers.Conv2D(filters=12, kernel_size=(5, 5), padding='valid', activation='tanh'))\nmodel.add(tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:22.971960Z","iopub.execute_input":"2022-08-14T21:20:22.972650Z","iopub.status.idle":"2022-08-14T21:20:23.011533Z","shell.execute_reply.started":"2022-08-14T21:20:22.972603Z","shell.execute_reply":"2022-08-14T21:20:23.010748Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"특징을 추출하였다면 분류를 위해 flatten layer를 추가해줍니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Flatten())","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.013317Z","iopub.execute_input":"2022-08-14T21:20:23.014073Z","iopub.status.idle":"2022-08-14T21:20:23.024346Z","shell.execute_reply.started":"2022-08-14T21:20:23.014006Z","shell.execute_reply":"2022-08-14T21:20:23.023144Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"마지막으로 flatten된 데이터를 softmax를 이용하여 10개의 class로 분류하도록 작성합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Dense(units=10, activation='softmax'))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.026280Z","iopub.execute_input":"2022-08-14T21:20:23.026720Z","iopub.status.idle":"2022-08-14T21:20:23.042712Z","shell.execute_reply.started":"2022-08-14T21:20:23.026687Z","shell.execute_reply":"2022-08-14T21:20:23.041426Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"그리고 model을 summary해서 살펴보면 3246개의 파라미터로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.044661Z","iopub.execute_input":"2022-08-14T21:20:23.045442Z","iopub.status.idle":"2022-08-14T21:20:23.052726Z","shell.execute_reply.started":"2022-08-14T21:20:23.045395Z","shell.execute_reply":"2022-08-14T21:20:23.051519Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하면 모델의 연결을 시각화하여 볼 수 있습니다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.054380Z","iopub.execute_input":"2022-08-14T21:20:23.054825Z","iopub.status.idle":"2022-08-14T21:20:23.176740Z","shell.execute_reply.started":"2022-08-14T21:20:23.054784Z","shell.execute_reply":"2022-08-14T21:20:23.175200Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **- functional api**","metadata":{}},{"cell_type":"markdown","source":"다음으로 각 layer마다 filter의 개수 등을 자세히 살펴보며 functional api를 사용하여 lenet-1을 작성해봅시다.","metadata":{}},{"cell_type":"markdown","source":"### **-- input layer**","metadata":{}},{"cell_type":"markdown","source":"먼저 28*28 이미지를 입력으로 받습니다.","metadata":{}},{"cell_type":"code","source":"x = tf.keras.layers.Input(shape=[28, 28, 1])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.178804Z","iopub.execute_input":"2022-08-14T21:20:23.179307Z","iopub.status.idle":"2022-08-14T21:20:23.187833Z","shell.execute_reply.started":"2022-08-14T21:20:23.179267Z","shell.execute_reply":"2022-08-14T21:20:23.186733Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 특징추출**","metadata":{}},{"cell_type":"markdown","source":"input layer를 통해 입력된 데이터는 5*5 크기를 가진 4개의 filter로 구성된 convolutoin layer를 거칩니다. \n\n이때, active function은 tanh를 사용합니다.\n\npadding은 사용하지 않을 것이기 때문에 valid로 설정해줍니다.\n\n마지막으로 convolution layer를 이전의 input layer인 x와 연결해줍니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Conv2D(filters=4, kernel_size=(5, 5), padding='valid', activation='tanh')(x)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.189405Z","iopub.execute_input":"2022-08-14T21:20:23.189928Z","iopub.status.idle":"2022-08-14T21:20:23.215120Z","shell.execute_reply.started":"2022-08-14T21:20:23.189891Z","shell.execute_reply":"2022-08-14T21:20:23.214113Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이후 2*2 크기의 filter로 averagepooling을 적용합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2)(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.216450Z","iopub.execute_input":"2022-08-14T21:20:23.217519Z","iopub.status.idle":"2022-08-14T21:20:23.226389Z","shell.execute_reply.started":"2022-08-14T21:20:23.217472Z","shell.execute_reply":"2022-08-14T21:20:23.225149Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"다음으로 5*5 크기를 가진 12개의 filter로 구성된 convolutoin layer와 average pooling을 한번 더 거쳐서 Feature Extractor를 완료합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Conv2D(filters=12, kernel_size=(5, 5), padding='valid', activation='tanh')(h)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid',strides=2)(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.227627Z","iopub.execute_input":"2022-08-14T21:20:23.228284Z","iopub.status.idle":"2022-08-14T21:20:23.255431Z","shell.execute_reply.started":"2022-08-14T21:20:23.228234Z","shell.execute_reply":"2022-08-14T21:20:23.254645Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"마지막으로 flatten을 이용해 데이터를 1차원으로 만든 다음 softmax를 적용한 dense layer로 분류를 진행합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Flatten()(h)\nY = tf.keras.layers.Dense(units=10, activation='softmax')(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.256630Z","iopub.execute_input":"2022-08-14T21:20:23.257071Z","iopub.status.idle":"2022-08-14T21:20:23.276214Z","shell.execute_reply.started":"2022-08-14T21:20:23.257025Z","shell.execute_reply":"2022-08-14T21:20:23.274923Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 생성**","metadata":{}},{"cell_type":"markdown","source":"functional api를 사용하면 keras의 Model을 이용해 x를 입력으로 받고 Y를 출력해주는 모델을 생성해야 합니다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.models.Model(x, Y,name = 'lenet-1')\nmodel.compile(loss='categorical_crossentropy', optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate), metrics=['accuracy'])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.277775Z","iopub.execute_input":"2022-08-14T21:20:23.278448Z","iopub.status.idle":"2022-08-14T21:20:23.294358Z","shell.execute_reply.started":"2022-08-14T21:20:23.278408Z","shell.execute_reply":"2022-08-14T21:20:23.293447Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"summary를 통해 모델을 확인하면 sequential한 모델과 똑같이 3246개의 파라미터로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.295389Z","iopub.execute_input":"2022-08-14T21:20:23.296099Z","iopub.status.idle":"2022-08-14T21:20:23.302648Z","shell.execute_reply.started":"2022-08-14T21:20:23.296020Z","shell.execute_reply":"2022-08-14T21:20:23.301478Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하여 모델의 연결을 시각화하여 살펴봅시다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.304070Z","iopub.execute_input":"2022-08-14T21:20:23.304718Z","iopub.status.idle":"2022-08-14T21:20:23.423419Z","shell.execute_reply.started":"2022-08-14T21:20:23.304674Z","shell.execute_reply":"2022-08-14T21:20:23.422157Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **-- 모델 학습**","metadata":{}},{"cell_type":"markdown","source":"fit을 이용하여 X_train과 y_train을 학습시키고 X_test와 y_test로 검증해봅시다.","metadata":{}},{"cell_type":"code","source":"history = model.fit(X_train, y_train, batch_size=100, epochs=20, validation_data = (X_test,y_test), verbose=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:20:23.425617Z","iopub.execute_input":"2022-08-14T21:20:23.426278Z","iopub.status.idle":"2022-08-14T21:21:22.810377Z","shell.execute_reply.started":"2022-08-14T21:20:23.426225Z","shell.execute_reply":"2022-08-14T21:21:22.809338Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습 결과**","metadata":{}},{"cell_type":"markdown","source":"학습이 진행되며 변화된 loss와 accuracy 값을 시각화 할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"history_frame = pd.DataFrame(history.history)\nhistory_frame.loc[:, ['loss', 'val_loss']].plot()\nhistory_frame.loc[:, ['accuracy', 'val_accuracy']].plot();","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:22.811907Z","iopub.execute_input":"2022-08-14T21:21:22.812515Z","iopub.status.idle":"2022-08-14T21:21:23.183862Z","shell.execute_reply.started":"2022-08-14T21:21:22.812476Z","shell.execute_reply":"2022-08-14T21:21:23.182739Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"3246개의 파라미터만으로도 약 97.7%의 정확도를 보이는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"max(history.history[\"val_accuracy\"])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.185161Z","iopub.execute_input":"2022-08-14T21:21:23.185569Z","iopub.status.idle":"2022-08-14T21:21:23.192522Z","shell.execute_reply.started":"2022-08-14T21:21:23.185530Z","shell.execute_reply":"2022-08-14T21:21:23.191437Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **lenet-4**","metadata":{}},{"cell_type":"markdown","source":"![image.png](attachment:a4c97a62-44c0-44c3-b4ae-8954ba4aa995.png)","metadata":{},"attachments":{"a4c97a62-44c0-44c3-b4ae-8954ba4aa995.png":{"image/png":"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"}}},{"cell_type":"markdown","source":"lenet-4에서는 padding을 적용합니다.\n\n이미지에 zero padding을 적용한 32*32 이미지를 처리하는 모델을 구현해봅시다.","metadata":{}},{"cell_type":"markdown","source":"## **- sequential api**","metadata":{}},{"cell_type":"markdown","source":"먼저 keras의 sequential을 model로 지정해줍시다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.Sequential()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.194291Z","iopub.execute_input":"2022-08-14T21:21:23.194754Z","iopub.status.idle":"2022-08-14T21:21:23.205173Z","shell.execute_reply.started":"2022-08-14T21:21:23.194709Z","shell.execute_reply":"2022-08-14T21:21:23.204244Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- input layer**","metadata":{}},{"cell_type":"markdown","source":"그리고 입력을 받을 layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Input(shape=[32, 32, 1]))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.206536Z","iopub.execute_input":"2022-08-14T21:21:23.207554Z","iopub.status.idle":"2022-08-14T21:21:23.216825Z","shell.execute_reply.started":"2022-08-14T21:21:23.207489Z","shell.execute_reply":"2022-08-14T21:21:23.215931Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 특징 추출**","metadata":{}},{"cell_type":"markdown","source":"lenet-4의 형태를 참고하여 convolution layer와 pooling layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Conv2D(filters=4, kernel_size=(5, 5), padding='valid', activation='tanh'))\nmodel.add(tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2))\nmodel.add(tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation='tanh'))\nmodel.add(tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.218167Z","iopub.execute_input":"2022-08-14T21:21:23.219131Z","iopub.status.idle":"2022-08-14T21:21:23.253160Z","shell.execute_reply.started":"2022-08-14T21:21:23.219085Z","shell.execute_reply":"2022-08-14T21:21:23.252205Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"특징을 추출하였다면 분류를 위해 flatten layer를 추가해줍니다.\n\nlenet-4은 lenet-1과 다르게 분류과정에 dense layer를 하나 더 추가하여 정확도를 증가시켰습니다.\n\n마지막으로 flatten된 데이터를 softmax를 이용하여 10개의 class로 분류하도록 작성합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Flatten())\nmodel.add(tf.keras.layers.Dense(units=120, activation='tanh'))\nmodel.add(tf.keras.layers.Dense(units=10, activation='softmax'))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.254616Z","iopub.execute_input":"2022-08-14T21:21:23.255090Z","iopub.status.idle":"2022-08-14T21:21:23.286937Z","shell.execute_reply.started":"2022-08-14T21:21:23.255022Z","shell.execute_reply":"2022-08-14T21:21:23.285991Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"그리고 model을 summary해서 살펴보면 51050개의 파라미터로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.288524Z","iopub.execute_input":"2022-08-14T21:21:23.288965Z","iopub.status.idle":"2022-08-14T21:21:23.295751Z","shell.execute_reply.started":"2022-08-14T21:21:23.288923Z","shell.execute_reply":"2022-08-14T21:21:23.294663Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하면 모델의 연결을 시각화하여 볼 수 있습니다.\n\ndense layer가 하나 더 추가된 것을 볼 수 있습니다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.297499Z","iopub.execute_input":"2022-08-14T21:21:23.298363Z","iopub.status.idle":"2022-08-14T21:21:23.426320Z","shell.execute_reply.started":"2022-08-14T21:21:23.298319Z","shell.execute_reply":"2022-08-14T21:21:23.424798Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **- functional api**","metadata":{}},{"cell_type":"markdown","source":"먼저 32*32 이미지를 입력으로 받습니다.","metadata":{}},{"cell_type":"code","source":"x = tf.keras.layers.Input(shape=[32, 32, 1])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.428112Z","iopub.execute_input":"2022-08-14T21:21:23.428516Z","iopub.status.idle":"2022-08-14T21:21:23.436997Z","shell.execute_reply.started":"2022-08-14T21:21:23.428477Z","shell.execute_reply":"2022-08-14T21:21:23.436147Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 특징 추출**","metadata":{}},{"cell_type":"markdown","source":"그리고 5*5 크기를 가진 4개의 filter로 구성된 convolutoin layer를 거칩니다. \n\nactive function은 tanh를 사용합니다.\n\n입력부분을 제외하면 padding은 사용하기 않기 때문에 valid로 설정합니다.\n\n마지막으로 convolution layer를 이전의 input layer인 x와 연결해줍니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Conv2D(filters=4, kernel_size=(5, 5), padding='valid', activation='tanh')(x)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.438181Z","iopub.execute_input":"2022-08-14T21:21:23.438637Z","iopub.status.idle":"2022-08-14T21:21:23.460565Z","shell.execute_reply.started":"2022-08-14T21:21:23.438606Z","shell.execute_reply":"2022-08-14T21:21:23.459496Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이후 2*2 크기의 filter로 averagepooling을 적용합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2)(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.462247Z","iopub.execute_input":"2022-08-14T21:21:23.462739Z","iopub.status.idle":"2022-08-14T21:21:23.470683Z","shell.execute_reply.started":"2022-08-14T21:21:23.462705Z","shell.execute_reply":"2022-08-14T21:21:23.469769Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"다음으로 5*5 크기를 가진 16개의 filter로 구성된 convolutoin layer와 average pooling을 한번 더 거쳐서 Feature Extractor를 완료합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation='tanh')(h)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid',strides=2)(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.471954Z","iopub.execute_input":"2022-08-14T21:21:23.472440Z","iopub.status.idle":"2022-08-14T21:21:23.494766Z","shell.execute_reply.started":"2022-08-14T21:21:23.472407Z","shell.execute_reply":"2022-08-14T21:21:23.493903Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"마지막으로 flatten layer를 이용해 데이터를 1차원으로 만든 다음 dense layer를 이용하여 분류를 진행합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Flatten()(h)\nh = tf.keras.layers.Dense(units=120, activation='tanh')(h)\nY = tf.keras.layers.Dense(units=10, activation='softmax')(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.495942Z","iopub.execute_input":"2022-08-14T21:21:23.496594Z","iopub.status.idle":"2022-08-14T21:21:23.523551Z","shell.execute_reply.started":"2022-08-14T21:21:23.496539Z","shell.execute_reply":"2022-08-14T21:21:23.522425Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 생성**","metadata":{}},{"cell_type":"markdown","source":"functional api를 사용하면 keras의 Model을 이용해 x를 입력으로 받고 Y를 출력해주는 모델을 생성해야 합니다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.models.Model(x, Y,name = 'lenet-4')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.524854Z","iopub.execute_input":"2022-08-14T21:21:23.525234Z","iopub.status.idle":"2022-08-14T21:21:23.537227Z","shell.execute_reply.started":"2022-08-14T21:21:23.525200Z","shell.execute_reply":"2022-08-14T21:21:23.536200Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"summary를 통해 모델을 확인하면 sequential한 모델과 똑같이 51050개의 파라미터로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.538360Z","iopub.execute_input":"2022-08-14T21:21:23.539171Z","iopub.status.idle":"2022-08-14T21:21:23.546564Z","shell.execute_reply.started":"2022-08-14T21:21:23.539136Z","shell.execute_reply":"2022-08-14T21:21:23.545688Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하여 모델의 연결을 시각화해봅시다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.547958Z","iopub.execute_input":"2022-08-14T21:21:23.548520Z","iopub.status.idle":"2022-08-14T21:21:23.680442Z","shell.execute_reply.started":"2022-08-14T21:21:23.548487Z","shell.execute_reply":"2022-08-14T21:21:23.678899Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"32,32 이미지는 기존 28,28 이미지에 padding을 한것과 같습니다.\n\n아래는 28,28 이미지를 받고 처리하는 모델입니다.\n\nsummary를 통해 확인해보면 모델의 파라미터 수가 위의 모델들과 같은 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"x = tf.keras.layers.Input(shape=[28, 28, 1])\n\nh = tf.keras.layers.Conv2D(filters=4, kernel_size=(5, 5), padding='same', activation='tanh')(x)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2)(h)\nh = tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation='tanh')(h)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid',strides=2)(h)\nh = tf.keras.layers.Flatten()(h)\nh = tf.keras.layers.Dense(units=120, activation='tanh')(h)\nY = tf.keras.layers.Dense(units=10, activation='softmax')(h)\n\nmodel = tf.keras.models.Model(x, Y,name = 'lenet-4_zero')\nmodel.compile(loss='categorical_crossentropy', optimizer=tf.keras.optimizers.RMSprop(learning_rate=learning_rate), metrics=['accuracy'])\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.682470Z","iopub.execute_input":"2022-08-14T21:21:23.682879Z","iopub.status.idle":"2022-08-14T21:21:23.759391Z","shell.execute_reply.started":"2022-08-14T21:21:23.682842Z","shell.execute_reply":"2022-08-14T21:21:23.758382Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습**","metadata":{}},{"cell_type":"markdown","source":"fit을 이용하여 X_train과 y_train을 학습시키고 X_test와 y_test로 검증해봅시다.","metadata":{}},{"cell_type":"code","source":"history = model.fit(X_train, y_train, batch_size=100, epochs=20, validation_data = (X_test,y_test), verbose=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:21:23.760984Z","iopub.execute_input":"2022-08-14T21:21:23.761646Z","iopub.status.idle":"2022-08-14T21:22:42.586932Z","shell.execute_reply.started":"2022-08-14T21:21:23.761601Z","shell.execute_reply":"2022-08-14T21:22:42.585359Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습 결과**","metadata":{}},{"cell_type":"markdown","source":"학습이 진행되며 변화된 loss와 accuracy 값을 시각화 할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"history_frame = pd.DataFrame(history.history)\nhistory_frame.loc[:, ['loss', 'val_loss']].plot()\nhistory_frame.loc[:, ['accuracy', 'val_accuracy']].plot();","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:42.588733Z","iopub.execute_input":"2022-08-14T21:22:42.589141Z","iopub.status.idle":"2022-08-14T21:22:43.340606Z","shell.execute_reply.started":"2022-08-14T21:22:42.589101Z","shell.execute_reply":"2022-08-14T21:22:43.339603Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"lenet-1보다 향상된 정확도를 보이는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"max(history.history[\"val_accuracy\"])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.341973Z","iopub.execute_input":"2022-08-14T21:22:43.342537Z","iopub.status.idle":"2022-08-14T21:22:43.349738Z","shell.execute_reply.started":"2022-08-14T21:22:43.342501Z","shell.execute_reply":"2022-08-14T21:22:43.348564Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **lenet-5**","metadata":{}},{"cell_type":"markdown","source":"![image.png](attachment:39ec752e-84f6-428c-8dda-cfbd89ac3119.png)","metadata":{},"attachments":{"39ec752e-84f6-428c-8dda-cfbd89ac3119.png":{"image/png":"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"}}},{"cell_type":"markdown","source":"![image.png](attachment:b2ef50b3-4a25-4686-a4c7-2c6c7fe89556.png)\n\nlenet-5 논문에서는 2번째 convoultion layer에서 연산을 진행할 때 이전 layer에서 넘어온 모든 feature를 고려하지 않고 저자가 선택하여 연결하였습니다.\n\n그러나 본 노트북에서는 이전 layer에서 넘어온 모든 feature를 고려하게하여 작성하였습니다.","metadata":{},"attachments":{"b2ef50b3-4a25-4686-a4c7-2c6c7fe89556.png":{"image/png":"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"}}},{"cell_type":"markdown","source":"## **- sequential api**","metadata":{}},{"cell_type":"markdown","source":"먼저 keras의 sequential을 model로 지정해줍시다.","metadata":{}},{"cell_type":"code","source":"model = tf.keras.Sequential()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.351229Z","iopub.execute_input":"2022-08-14T21:22:43.351702Z","iopub.status.idle":"2022-08-14T21:22:43.363076Z","shell.execute_reply.started":"2022-08-14T21:22:43.351657Z","shell.execute_reply":"2022-08-14T21:22:43.361989Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- input layer**","metadata":{}},{"cell_type":"markdown","source":"그리고 입력을 받을 layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Input(shape=[32, 32, 1]))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.364708Z","iopub.execute_input":"2022-08-14T21:22:43.365728Z","iopub.status.idle":"2022-08-14T21:22:43.376160Z","shell.execute_reply.started":"2022-08-14T21:22:43.365687Z","shell.execute_reply":"2022-08-14T21:22:43.375110Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 특징 추출**","metadata":{}},{"cell_type":"markdown","source":"lenet-5의 형태를 참고하여 convolution layer와 pooling layer를 추가합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Conv2D(filters=6, kernel_size=(5, 5), padding='valid', activation='tanh'))\nmodel.add(tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2))\nmodel.add(tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation='tanh'))\nmodel.add(tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.377788Z","iopub.execute_input":"2022-08-14T21:22:43.378147Z","iopub.status.idle":"2022-08-14T21:22:43.417096Z","shell.execute_reply.started":"2022-08-14T21:22:43.378115Z","shell.execute_reply":"2022-08-14T21:22:43.416247Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"특징을 추출하였다면 분류를 위해 flatten 시켜줍시다.\n\n그리고 lenet-5에서는 lenet-4에서 분류과정에 dense layer를 하나 더 추가하여 정확도를 증가시켰습니다.\n\n마지막으로 flatten된 데이터를 softmax를 이용하여 10개의 class로 분류하도록 작성합니다.","metadata":{}},{"cell_type":"code","source":"model.add(tf.keras.layers.Flatten())\nmodel.add(tf.keras.layers.Dense(units=120, activation='tanh'))\nmodel.add(tf.keras.layers.Dense(units=84, activation='tanh'))\nmodel.add(tf.keras.layers.Dense(units=10, activation='softmax'))","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.419098Z","iopub.execute_input":"2022-08-14T21:22:43.419418Z","iopub.status.idle":"2022-08-14T21:22:43.455991Z","shell.execute_reply.started":"2022-08-14T21:22:43.419388Z","shell.execute_reply":"2022-08-14T21:22:43.455227Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"model을 summary해서 살펴보면 61706개의 파라미터로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.457284Z","iopub.execute_input":"2022-08-14T21:22:43.457993Z","iopub.status.idle":"2022-08-14T21:22:43.463732Z","shell.execute_reply.started":"2022-08-14T21:22:43.457954Z","shell.execute_reply":"2022-08-14T21:22:43.462969Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하면 모델의 연결을 시각화하여 볼 수 있습니다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.465099Z","iopub.execute_input":"2022-08-14T21:22:43.465557Z","iopub.status.idle":"2022-08-14T21:22:43.619245Z","shell.execute_reply.started":"2022-08-14T21:22:43.465525Z","shell.execute_reply":"2022-08-14T21:22:43.617996Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **- functional api**","metadata":{}},{"cell_type":"markdown","source":"lenet-5는 lenet-4와 같이 28 * 28 이미지에 zero padding을 적용한 32 * 32 이미지를 입력으로 받습니다.","metadata":{}},{"cell_type":"code","source":"x = tf.keras.layers.Input(shape=[32, 32, 1])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.621391Z","iopub.execute_input":"2022-08-14T21:22:43.622005Z","iopub.status.idle":"2022-08-14T21:22:43.630366Z","shell.execute_reply.started":"2022-08-14T21:22:43.621965Z","shell.execute_reply":"2022-08-14T21:22:43.629328Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 특징추출**","metadata":{}},{"cell_type":"markdown","source":"그리고 5*5 크기를 가진 6개의 filter로 구성된 convolutoin layer를 거칩니다. \n\nactive function은 tanh를 사용합니다.\n\n입력부분을 제외하면 padding은 사용하기 않기 때문에 valid로 설정합니다.\n\n마지막으로 convolution layer를 input layer인 x와 연결해줍니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Conv2D(filters=6, kernel_size=(5, 5), padding='valid', activation='tanh')(x)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.632431Z","iopub.execute_input":"2022-08-14T21:22:43.632811Z","iopub.status.idle":"2022-08-14T21:22:43.657067Z","shell.execute_reply.started":"2022-08-14T21:22:43.632780Z","shell.execute_reply":"2022-08-14T21:22:43.655956Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이후 2*2 크기의 filter로 averagepooling을 적용합니다.\n","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2)(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.658395Z","iopub.execute_input":"2022-08-14T21:22:43.659164Z","iopub.status.idle":"2022-08-14T21:22:43.668981Z","shell.execute_reply.started":"2022-08-14T21:22:43.659127Z","shell.execute_reply":"2022-08-14T21:22:43.668132Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"다음으로 5*5 크기를 가진 16개의 filter로 구성된 convolutoin layer와 average pooling을 한번 더 거쳐서 Feature Extractor를 완료합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation='tanh')(h)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid',strides=2)(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.671342Z","iopub.execute_input":"2022-08-14T21:22:43.671850Z","iopub.status.idle":"2022-08-14T21:22:43.696751Z","shell.execute_reply.started":"2022-08-14T21:22:43.671800Z","shell.execute_reply":"2022-08-14T21:22:43.695611Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 분류**","metadata":{}},{"cell_type":"markdown","source":"마지막으로 flatten layer를 이용해 데이터를 1차원으로 만든 다음 dense layer를 이용하여 분류를 진행합니다.","metadata":{}},{"cell_type":"code","source":"h = tf.keras.layers.Flatten()(h)\nh = tf.keras.layers.Dense(units=120, activation='tanh')(h)\nh = tf.keras.layers.Dense(units=84, activation='tanh')(h)\nY = tf.keras.layers.Dense(units=10, activation='softmax')(h)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.698322Z","iopub.execute_input":"2022-08-14T21:22:43.698833Z","iopub.status.idle":"2022-08-14T21:22:43.735476Z","shell.execute_reply.started":"2022-08-14T21:22:43.698787Z","shell.execute_reply":"2022-08-14T21:22:43.734730Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 생성**","metadata":{}},{"cell_type":"markdown","source":"keras의 Model을 이용해 x를 입력으로 받고 Y를 출력해주는 모델을 생성합시다.\n","metadata":{}},{"cell_type":"code","source":"model = tf.keras.models.Model(x, Y,name = 'lenet-5')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.737078Z","iopub.execute_input":"2022-08-14T21:22:43.738118Z","iopub.status.idle":"2022-08-14T21:22:43.747125Z","shell.execute_reply.started":"2022-08-14T21:22:43.738069Z","shell.execute_reply":"2022-08-14T21:22:43.746323Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 살펴보기**","metadata":{}},{"cell_type":"markdown","source":"summary를 통해 모델을 확인하면 sequential한 모델과 같은 구조로 이루어져 있는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"model.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.748753Z","iopub.execute_input":"2022-08-14T21:22:43.749505Z","iopub.status.idle":"2022-08-14T21:22:43.759785Z","shell.execute_reply.started":"2022-08-14T21:22:43.749460Z","shell.execute_reply":"2022-08-14T21:22:43.758763Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"plot_model을 이용하여 모델의 연결을 시각화해봅시다.","metadata":{}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.761240Z","iopub.execute_input":"2022-08-14T21:22:43.761781Z","iopub.status.idle":"2022-08-14T21:22:43.911612Z","shell.execute_reply.started":"2022-08-14T21:22:43.761746Z","shell.execute_reply":"2022-08-14T21:22:43.910145Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"32,32 이미지는 기존 28,28 이미지에 padding을 한것과 같습니다.\n\n아래는 28,28 이미지를 받고 처리하는 lenet-5 모델입니다.","metadata":{}},{"cell_type":"code","source":"x = tf.keras.layers.Input(shape=[28, 28, 1])\n\nh = tf.keras.layers.Conv2D(filters=6, kernel_size=(5, 5), padding='same', activation='tanh')(x)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid', strides=2)(h)\nh = tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation='tanh')(h)\nh = tf.keras.layers.AveragePooling2D(pool_size=(2, 2), padding='valid',strides=2)(h)\nh = tf.keras.layers.Flatten()(h)\nh = tf.keras.layers.Dense(units=120, activation='tanh')(h)\nh = tf.keras.layers.Dense(units=84, activation='tanh')(h)\nY = tf.keras.layers.Dense(units=10, activation='softmax')(h)\n\nmodel = tf.keras.models.Model(x, Y,name = 'lenet-5_zero')\nmodel.compile(loss='categorical_crossentropy', optimizer=tf.keras.optimizers.RMSprop(learning_rate=learning_rate), metrics=['accuracy'])\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.914164Z","iopub.execute_input":"2022-08-14T21:22:43.914656Z","iopub.status.idle":"2022-08-14T21:22:43.995292Z","shell.execute_reply.started":"2022-08-14T21:22:43.914606Z","shell.execute_reply":"2022-08-14T21:22:43.994112Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습**","metadata":{}},{"cell_type":"markdown","source":"fit을 이용하여 X_train과 y_train을 학습시키고 X_test와 y_test로 검증해봅시다.","metadata":{}},{"cell_type":"code","source":"history = model.fit(X_train, y_train, batch_size=100, epochs=20, validation_data = (X_test,y_test), verbose=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:22:43.996868Z","iopub.execute_input":"2022-08-14T21:22:43.997227Z","iopub.status.idle":"2022-08-14T21:24:10.763849Z","shell.execute_reply.started":"2022-08-14T21:22:43.997195Z","shell.execute_reply":"2022-08-14T21:24:10.762950Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- 모델 학습 결과**","metadata":{}},{"cell_type":"markdown","source":"학습이 진행되며 변화된 loss와 accuracy 값을 시각화 할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"history_frame = pd.DataFrame(history.history)\nhistory_frame.loc[:, ['loss', 'val_loss']].plot()\nhistory_frame.loc[:, ['accuracy', 'val_accuracy']].plot();","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:10.765177Z","iopub.execute_input":"2022-08-14T21:24:10.766178Z","iopub.status.idle":"2022-08-14T21:24:11.184655Z","shell.execute_reply.started":"2022-08-14T21:24:10.766137Z","shell.execute_reply":"2022-08-14T21:24:11.183681Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"정확도가 향상된 것을 볼 수 있습니다.","metadata":{}},{"cell_type":"code","source":"max(history.history[\"val_accuracy\"])","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:11.185791Z","iopub.execute_input":"2022-08-14T21:24:11.186134Z","iopub.status.idle":"2022-08-14T21:24:11.192414Z","shell.execute_reply.started":"2022-08-14T21:24:11.186102Z","shell.execute_reply":"2022-08-14T21:24:11.191608Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **가중치 시각화**","metadata":{}},{"cell_type":"markdown","source":"데이터를 이용해 학습을 거친 모델은 숫자를 분류할 수 있게됩니다.\n\n대부분의 사람들은 모델이 숫자를 분류하는 과정을 궁금해하지 않지만 이 노트북을 작성하며 Convolutional Neural Network에 대해 공부하다보니\n\n입력값이 1번째 layer의 filter에서 정확히 얼마의 weight을 거쳐서 다음 layer로 전달되며,\n\n3번째 layer에서는 얼마의 weight이 적용되는지 궁금하였습니다.\n\n그러나 가중치를 시각화하거나 가중치가 적용된 값을 시각화하는 사이트만 있을 뿐 코드로 설명된 자료는 찾기 힘들었습니다.\n\n그래서 각 layer 별로 생기는 각 filter별 weight값을 얻어내었습니다.\n\n그러나 숫자로 된 값은 한번에 알아보기 힘들기 때문에 matplotlib를 이용하여 시각화를 진행하였습니다.\n\n결과를 보면 각 filter별로 어떤 부위에 강한 weight값이 적용되었는지를 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"for i in [1,3]:\n    print(\"layer\",i)\n    conv = model.layers[i]\n    fig, axs = plt.subplots(1,conv.weights[1].shape[0],figsize=(15,2))\n    for j in range(conv.weights[1].shape[0]):\n        axs[j].imshow(conv.get_weights()[0][:,:,0,j],cmap='jet',vmin=-0.5,vmax=0.5)\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:11.193956Z","iopub.execute_input":"2022-08-14T21:24:11.194578Z","iopub.status.idle":"2022-08-14T21:24:13.001854Z","shell.execute_reply.started":"2022-08-14T21:24:11.194543Z","shell.execute_reply":"2022-08-14T21:24:13.000675Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **적용 결과 시각화**","metadata":{}},{"cell_type":"markdown","source":"그러나 각 layer의 weight 값만 보고서는 입력값이 어떻게 변하는지 알기 힘들었습니다.\n\n그래서 n번째 layer까지 도달하는 딥러닝 모델을 제작하고 입력값을 입력하였습니다.\n\n마지막으로 모델의 출력값을 시각화하여 입력값이 다음 layer로 전달되는 과정을 확인하였습니다.","metadata":{}},{"cell_type":"markdown","source":"### **-- first layer (input layer)**","metadata":{}},{"cell_type":"markdown","source":"첫번째 layer는 input layer이므로 입력값이 그대로 출력되는 것을 알 수 있습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[0].output)\nfig, axs = plt.subplots(1,10,figsize=(15,8))\nfor i in range(0,100,10):\n    inputs = X_train[i]\n    feature_map = conv_model.predict(inputs)\n    axs[i//10].imshow(feature_map.reshape(28,28),cmap='gray')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:13.003771Z","iopub.execute_input":"2022-08-14T21:24:13.004487Z","iopub.status.idle":"2022-08-14T21:24:14.456643Z","shell.execute_reply.started":"2022-08-14T21:24:13.004443Z","shell.execute_reply":"2022-08-14T21:24:14.455595Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **- 특징 추출**","metadata":{}},{"cell_type":"markdown","source":"### **-- second layer**","metadata":{}},{"cell_type":"markdown","source":"2번째 layer는 첫번째 convolution layer이며 2번째 layer까지의 weight과 bias를 적용한 값을 시각화합니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[1].output)\nfor i in range(0,100,10):\n    inputs = X_train[i].reshape(28,28,1)\n    feature_map = conv_model.predict(inputs)\n\n    fig, axs = plt.subplots(1,6,figsize=(15,8))\n    \n    for j in range(6):\n        axs[j].imshow(feature_map[:,:,0,j],cmap='gray')\n        \n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:14.465460Z","iopub.execute_input":"2022-08-14T21:24:14.466101Z","iopub.status.idle":"2022-08-14T21:24:21.445658Z","shell.execute_reply.started":"2022-08-14T21:24:14.466061Z","shell.execute_reply":"2022-08-14T21:24:21.444243Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- third layer**","metadata":{}},{"cell_type":"markdown","source":"3번째 layer는 첫번째 pooling layer입니다.\n\nconvolution layer를 거친 값을 pooling한 값을 시각화하여 확인할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[2].output)\nfor i in range(0,100,10):\n    inputs = X_train[i].reshape(1,28,28)\n    feature_map = conv_model.predict(inputs)\n\n    fig, axs = plt.subplots(1,6,figsize=(15,8))\n    \n    for j in range(6):\n        axs[j].imshow(feature_map[0,:,:,j],cmap='gray')\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:21.448195Z","iopub.execute_input":"2022-08-14T21:24:21.448583Z","iopub.status.idle":"2022-08-14T21:24:27.907326Z","shell.execute_reply.started":"2022-08-14T21:24:21.448547Z","shell.execute_reply":"2022-08-14T21:24:27.905945Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- fourth layer**","metadata":{}},{"cell_type":"markdown","source":"4번째 layer는 두번째 convolution layer입니다.\n\n4번재 layer까지의 weight과 bias, pooling이 적용된 값을 시각화하여 확인할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[3].output)\nfor i in range(0,100,10):\n    inputs = X_train[i].reshape(1,28,28)\n    feature_map = conv_model.predict(inputs)\n\n    fig, axs = plt.subplots(1,16,figsize=(15,8))\n    for j in range(16):\n        axs[j].imshow(feature_map[0,:,:,j],cmap='gray')\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:27.908802Z","iopub.execute_input":"2022-08-14T21:24:27.909225Z","iopub.status.idle":"2022-08-14T21:24:41.514456Z","shell.execute_reply.started":"2022-08-14T21:24:27.909177Z","shell.execute_reply":"2022-08-14T21:24:41.513305Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- fifth layer**","metadata":{}},{"cell_type":"markdown","source":"5번째 layer는 두번째 pooling layer입니다.\n\n5번재 layer까지의 convolution, pooling layer가 적용된 값을 시각화하여 확인할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[4].output)\nfor i in range(0,100,10):\n    inputs = X_train[i].reshape(1,28,28)\n    feature_map = conv_model.predict(inputs)\n\n    fig, axs = plt.subplots(1,16,figsize=(15,8))\n    for j in range(16):\n        axs[j].imshow(feature_map[0,:,:,j],cmap='gray')\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:41.515712Z","iopub.execute_input":"2022-08-14T21:24:41.516112Z","iopub.status.idle":"2022-08-14T21:24:55.490361Z","shell.execute_reply.started":"2022-08-14T21:24:41.516074Z","shell.execute_reply":"2022-08-14T21:24:55.489133Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"convolution은 딥러닝을 이용하여 Feature Extractor를 진행하므로 네트워크가 깊어지면 사람은 시각화한 값을 해석하기 힘들어집니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 분류**","metadata":{}},{"cell_type":"markdown","source":"이후에는 flatten을 진행하여 1차원으로 만들기 때문에 하나의 값을 기준으로 layer가 적용된 값을 확인하겠습니다.","metadata":{}},{"cell_type":"code","source":"plt.imshow(X_train[0], cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:55.492132Z","iopub.execute_input":"2022-08-14T21:24:55.492745Z","iopub.status.idle":"2022-08-14T21:24:55.687419Z","shell.execute_reply.started":"2022-08-14T21:24:55.492709Z","shell.execute_reply":"2022-08-14T21:24:55.686437Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- sixth layer**","metadata":{}},{"cell_type":"markdown","source":"6번째 layer는 flatten layer입니다.\n\n5번째 layer의 값을 1차원으로 flatten한 값을 시각화하여 확인할 수 있습니다.\n\n784개의 값으로 flatten되기 때문에 100개의 값만 선정하여 시각화하였습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[5].output)\nfor i in range(0,100,10):\n    plt.figure(figsize=(20,5))\n    feature_map = conv_model.predict(X_train[i].reshape(1,28,28))\n    plt.imshow(feature_map[0][150:250].reshape(1,100), cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:55.688748Z","iopub.execute_input":"2022-08-14T21:24:55.689106Z","iopub.status.idle":"2022-08-14T21:24:57.953396Z","shell.execute_reply.started":"2022-08-14T21:24:55.689073Z","shell.execute_reply":"2022-08-14T21:24:57.952129Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- seventh layer**","metadata":{}},{"cell_type":"markdown","source":"7번째 layer는 dense layer입니다.\n\nfeature extraction을 진행한 후 classification을 진행하는 단계입니다.\n\n120개의 unit의 각 weight과 bias가 적용된 값을 시각화 할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[6].output)\nfor i in range(0,100,10):\n    plt.figure(figsize=(20,5))\n    feature_map = conv_model.predict(X_train[i].reshape(1,28,28))\n    plt.imshow(feature_map[0].reshape(1,120), cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:24:57.954865Z","iopub.execute_input":"2022-08-14T21:24:57.955377Z","iopub.status.idle":"2022-08-14T21:25:00.606673Z","shell.execute_reply.started":"2022-08-14T21:24:57.955326Z","shell.execute_reply":"2022-08-14T21:25:00.605561Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- eighth layer**","metadata":{}},{"cell_type":"markdown","source":"8번째 layer는 dense layer입니다.\n\nclassification을 진행하며 softmax를 이용하여 최종 10개의 선택지로 분류하기 전 마지막 layer입니다.\n\n84개의 unit의 각 weight과 bias가 적용된 값을 시각화 할 수 있습니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[7].output)\nfor i in range(0,100,10):\n    plt.figure(figsize=(20,5))\n    feature_map = conv_model.predict(X_train[i].reshape(1,28,28))\n    plt.imshow(feature_map[0].reshape(1,84), cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:00.607941Z","iopub.execute_input":"2022-08-14T21:25:00.608307Z","iopub.status.idle":"2022-08-14T21:25:03.133907Z","shell.execute_reply.started":"2022-08-14T21:25:00.608275Z","shell.execute_reply":"2022-08-14T21:25:03.132871Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **-- ninth layer**","metadata":{}},{"cell_type":"markdown","source":"마지막 9번째 layer는 dense layer입니다.\n\n활성화 함수로 softmax를 사용하여 최종적으로 10개의 선택지 중 하나로 분류합니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[8].output)\nfor i in range(0,100,10):\n    plt.figure(figsize=(20,5))\n    feature_map = conv_model.predict(X_train[i].reshape(1,28,28))\n    plt.imshow(feature_map[0].reshape(1,10), cmap='gray')","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:03.135169Z","iopub.execute_input":"2022-08-14T21:25:03.135511Z","iopub.status.idle":"2022-08-14T21:25:05.513723Z","shell.execute_reply.started":"2022-08-14T21:25:03.135474Z","shell.execute_reply":"2022-08-14T21:25:05.512754Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# **머신러닝 모델을 이용한 분류 실험**","metadata":{}},{"cell_type":"markdown","source":"데이터가 모델을 통과하는 과정을 시각화하니 lenet의 장점은 convolution layer를 통한 특징 추출에 있다고 생각되었습니다.\n\n그래서 분류기로 머신러닝 모델을 이용해보면 어떨까라는 생각을 하게되었습니다.","metadata":{}},{"cell_type":"markdown","source":"## **- 새로운 입력 데이터 생성**\n\nlenet-5의 특징 추출 부분을 제외한 부분을 입력 데이터로 사용하였습니다.","metadata":{}},{"cell_type":"markdown","source":"먼저 10개의 값 중 하나로 분류하기 위해 to_categorical을 적용한 값을 원래대로 돌려놓습니다.","metadata":{}},{"cell_type":"code","source":"y_train = np.argmax(y_train,axis=1)\ny_test = np.argmax(y_test,axis=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:05.514975Z","iopub.execute_input":"2022-08-14T21:25:05.515355Z","iopub.status.idle":"2022-08-14T21:25:05.521697Z","shell.execute_reply.started":"2022-08-14T21:25:05.515321Z","shell.execute_reply":"2022-08-14T21:25:05.520283Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이후 특징 추출이 끝난 6번째 layer와 이후 dense layer의 출력값을 합쳐 input으로 만듭니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[5].output)\noutput_feature1 = conv_model.predict(X_train)\nconv_model = tf.keras.Model(model.input, model.layers[6].output)\noutput_feature2 = conv_model.predict(X_train)\nconv_model = tf.keras.Model(model.input, model.layers[7].output)\noutput_feature3 = conv_model.predict(X_train)\nconv_model = tf.keras.Model(model.input, model.layers[8].output)\noutput_feature4 = conv_model.predict(X_train)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:05.522927Z","iopub.execute_input":"2022-08-14T21:25:05.523363Z","iopub.status.idle":"2022-08-14T21:25:17.682000Z","shell.execute_reply.started":"2022-08-14T21:25:05.523330Z","shell.execute_reply":"2022-08-14T21:25:17.680730Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"new_train1 = pd.DataFrame(output_feature1)\nnew_train2 = pd.DataFrame(output_feature2)\nnew_train3 = pd.DataFrame(output_feature3)\nnew_train4 = pd.DataFrame(output_feature4)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:17.683801Z","iopub.execute_input":"2022-08-14T21:25:17.684314Z","iopub.status.idle":"2022-08-14T21:25:17.691313Z","shell.execute_reply.started":"2022-08-14T21:25:17.684267Z","shell.execute_reply":"2022-08-14T21:25:17.690105Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"new_train = pd.concat([new_train1,new_train2,new_train3,new_train4],axis=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:17.693138Z","iopub.execute_input":"2022-08-14T21:25:17.693934Z","iopub.status.idle":"2022-08-14T21:25:17.826892Z","shell.execute_reply.started":"2022-08-14T21:25:17.693887Z","shell.execute_reply":"2022-08-14T21:25:17.825684Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"new_train.columns = list(range(614))\nnew_train.rename(columns = lambda x:\"column_\"+str(x), inplace = True)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:17.828597Z","iopub.execute_input":"2022-08-14T21:25:17.828964Z","iopub.status.idle":"2022-08-14T21:25:17.835501Z","shell.execute_reply.started":"2022-08-14T21:25:17.828933Z","shell.execute_reply":"2022-08-14T21:25:17.834661Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"검증 데이터에 대해서도 똑같은 작업을 진행합니다.","metadata":{}},{"cell_type":"code","source":"conv_model = tf.keras.Model(model.input, model.layers[5].output)\noutput_feature1 = conv_model.predict(X_test)\nconv_model = tf.keras.Model(model.input, model.layers[6].output)\noutput_feature2 = conv_model.predict(X_test)\nconv_model = tf.keras.Model(model.input, model.layers[7].output)\noutput_feature3 = conv_model.predict(X_test)\nconv_model = tf.keras.Model(model.input, model.layers[8].output)\noutput_feature4 = conv_model.predict(X_test)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:17.836889Z","iopub.execute_input":"2022-08-14T21:25:17.837513Z","iopub.status.idle":"2022-08-14T21:25:21.227399Z","shell.execute_reply.started":"2022-08-14T21:25:17.837476Z","shell.execute_reply":"2022-08-14T21:25:21.226570Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"new_test1 = pd.DataFrame(output_feature1)\nnew_test2 = pd.DataFrame(output_feature2)\nnew_test3 = pd.DataFrame(output_feature3)\nnew_test4 = pd.DataFrame(output_feature4)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:21.228677Z","iopub.execute_input":"2022-08-14T21:25:21.229675Z","iopub.status.idle":"2022-08-14T21:25:21.235757Z","shell.execute_reply.started":"2022-08-14T21:25:21.229634Z","shell.execute_reply":"2022-08-14T21:25:21.234642Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"new_test = pd.concat([new_test1,new_test2,new_test3,new_test4],axis=1)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:21.237234Z","iopub.execute_input":"2022-08-14T21:25:21.237574Z","iopub.status.idle":"2022-08-14T21:25:21.259699Z","shell.execute_reply.started":"2022-08-14T21:25:21.237542Z","shell.execute_reply":"2022-08-14T21:25:21.258614Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"new_test.columns = list(range(614))\nnew_test.rename(columns = lambda x:\"column_\"+str(x), inplace = True)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:21.262058Z","iopub.execute_input":"2022-08-14T21:25:21.262597Z","iopub.status.idle":"2022-08-14T21:25:21.271051Z","shell.execute_reply.started":"2022-08-14T21:25:21.262543Z","shell.execute_reply":"2022-08-14T21:25:21.270151Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"마지막으로 support vector machine을 이용하여 학습시켜보았습니다.","metadata":{}},{"cell_type":"code","source":"final_model = SVC()\nfinal_model.fit(new_train,y_train)\npred = final_model.predict(new_test)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:25:21.275446Z","iopub.execute_input":"2022-08-14T21:25:21.275866Z","iopub.status.idle":"2022-08-14T21:26:02.236237Z","shell.execute_reply.started":"2022-08-14T21:25:21.275829Z","shell.execute_reply":"2022-08-14T21:26:02.235115Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"이 노트북에서 다룬 알고리즘 중 가장 높은 정확도를 보이는 것을 알 수 있습니다.\n\n분류기를 전환하는 것도 실험해볼만 한 것 같습니다.\n\n추후에 실험을 자유롭게 진행할 수 있는 시간, 공간 등의 요소가 갖춰지면 다시 연구해보아야겠습니다. :)","metadata":{}},{"cell_type":"code","source":"accuracy_score(pred,y_test)","metadata":{"execution":{"iopub.status.busy":"2022-08-14T21:26:02.237597Z","iopub.execute_input":"2022-08-14T21:26:02.237963Z","iopub.status.idle":"2022-08-14T21:26:02.248064Z","shell.execute_reply.started":"2022-08-14T21:26:02.237921Z","shell.execute_reply":"2022-08-14T21:26:02.247023Z"},"trusted":true},"execution_count":null,"outputs":[]}]}