{"cells":[{"metadata":{"trusted":true},"cell_type":"code","source":"import copy\nimport gc\nimport glob\nimport os\nimport time\n\nimport cv2\nimport IPython\nimport IPython.display\nimport joblib\nimport librosa\nimport librosa.display\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nimport scipy\nfrom joblib import Parallel, delayed\nfrom sklearn.model_selection import KFold, StratifiedKFold, train_test_split\nfrom tqdm import tqdm\n\n%matplotlib inline","execution_count":1,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"pd.options.display.max_columns = 128\npd.options.display.max_rows = 128\nplt.rcParams['figure.figsize'] = (15, 8)","execution_count":2,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"class EasyDict(dict):\n def __init__(self, d=None, **kwargs):\n if d is None:\n d = {}\n if kwargs:\n d.update(**kwargs)\n for k, v in d.items():\n setattr(self, k, v)\n # Class attributes\n for k in self.__class__.__dict__.keys():\n if not (k.startswith('__') and k.endswith('__')) and not k in ('update', 'pop'):\n setattr(self, k, getattr(self, k))\n\n def __setattr__(self, name, value):\n if isinstance(value, (list, tuple)):\n value = [self.__class__(x)\n if isinstance(x, dict) else x for x in value]\n elif isinstance(value, dict) and not isinstance(value, self.__class__):\n value = self.__class__(value)\n super(EasyDict, self).__setattr__(name, value)\n super(EasyDict, self).__setitem__(name, value)\n\n __setitem__ = __setattr__\n\n def update(self, e=None, **f):\n d = e or dict()\n d.update(f)\n for k in d:\n setattr(self, k, d[k])\n\n def pop(self, k, d=None):\n delattr(self, k)\n return super(EasyDict, self).pop(k, d)","execution_count":3,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Configuration and global parameters:"},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df = pd.read_csv('../input/train_curated.csv')\nsample_submission = pd.read_csv('../input/sample_submission.csv')\nprint('train: {}'.format(train_df.shape))\nprint('test: {}'.format(sample_submission.shape))\n\nROOT = '../input/'\ntest_root = os.path.join(ROOT, 'test/')\ntrain_root = os.path.join(ROOT, 'train_curated/')\n\n\nCONFIG = EasyDict()\nCONFIG.hop_length = 347 # to make time steps 128\nCONFIG.fmin = 20\nCONFIG.fmax = 44100 / 2\nCONFIG.n_fft = 480\n\nN_SAMPLES = 48\nSAMPLE_DIM = 256\n\nTRAINING_CONFIG = {\n 'sample_dim': (N_SAMPLES, SAMPLE_DIM),\n 'padding_mode': cv2.BORDER_REFLECT,\n}\n\nprint(CONFIG)\nprint(TRAINING_CONFIG)\n\ntrain_df.head()","execution_count":4,"outputs":[{"output_type":"stream","text":"train: (4970, 2)\ntest: (1120, 81)\n{'hop_length': 347, 'fmin': 20, 'fmax': 22050.0, 'n_fft': 480}\n{'sample_dim': (48, 256), 'padding_mode': 2}\n","name":"stdout"},{"output_type":"execute_result","execution_count":4,"data":{"text/plain":" fname labels\n0 0006ae4e.wav Bark\n1 0019ef41.wav Raindrop\n2 001ec0ad.wav Finger_snapping\n3 0026c7cb.wav Run\n4 0026f116.wav Finger_snapping","text/html":"
\n\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n
fnamelabels
00006ae4e.wavBark
10019ef41.wavRaindrop
2001ec0ad.wavFinger_snapping
30026c7cb.wavRun
40026f116.wavFinger_snapping
\n
"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## Data Processing class:"},{"metadata":{"trusted":true},"cell_type":"code","source":"# Preprocessing functions inspired by:\n# https://github.com/xiaozhouwang/tensorflow_speech_recognition_solution/blob/master/data.py\nclass DataProcessor(object):\n \n def __init__(self, debug=False):\n self.debug = debug\n \n # Placeholders for global statistics\n self.mel_mean = None\n self.mel_std = None\n self.mel_max = None\n self.mfcc_max = None\n \n def createMel(self, filename, params, normalize=False):\n \"\"\"\n Create Mel Spectrogram sample out of raw wavfile\n \"\"\"\n y, sr = librosa.load(filename, sr=None)\n mel = librosa.feature.melspectrogram(y, sr, n_mels=N_SAMPLES, **params)\n mel = librosa.power_to_db(mel)\n if normalize:\n if self.mel_mean is not None and self.mel_std is not None:\n mel = (mel - self.mel_mean) / self.mel_std\n else:\n sample_mean = np.mean(mel)\n sample_std = np.std(mel)\n mel = (mel - sample_mean) / sample_std\n if self.mel_max is not None:\n mel = mel / self.mel_max\n else:\n mel = mel / np.max(np.abs(mel))\n return mel\n \n def createMfcc(self, filename, params, normalize=False):\n \"\"\"\n Create MFCC sample out of raw wavfile\n \"\"\"\n y, sr = librosa.load(filename, sr=None)\n nonzero_idx = [y > 0]\n y[nonzero_idx] = np.log(y[nonzero_idx])\n mfcc = librosa.feature.mfcc(y, sr, n_mfcc=N_SAMPLES, **params)\n if normalize:\n if self.mfcc_max is not None:\n mfcc = mfcc / self.mfcc_max\n else:\n mfcc = mfcc / np.max(np.abs(mfcc))\n return mfcc\n \n def createLogspec(self, filename, params,\n normalize=False,\n window_size=20,\n step_size=10, eps=1e-10):\n \"\"\"\n Create log spectrogram,\n based on \n https://www.kaggle.com/voglinio/keras-2d-model-5-fold-log-specgram-curated-only\n \"\"\"\n \n y, sr = librosa.load(filename, sr=None)\n nperseg = int(round(window_size * sr / 1e3))\n noverlap = int(round(step_size * sr / 1e3))\n freqs, times, spec = scipy.signal.spectrogram(\n y,\n fs=sr,\n window='hann',\n nperseg=nperseg,\n noverlap=noverlap,\n detrend=False)\n spec = np.log(spec.astype(np.float32) + eps)\n return spec\n \n \n def prepareSample(self, root, row, \n preprocFunc, \n preprocParams, trainingParams, \n test_mode=False, normalize=False, \n proc_mode='split'):\n \"\"\"\n Prepare sample for model training.\n Function takes row of DataFrame, extracts filename and labels and processes them.\n \n If proc_mode is 'split':\n Outputs sets of arrays of constant shape padded to TRAINING_CONFIG shape\n with selected padding mode, also specified in TRAINING_CONFIG.\n This approach prevents loss of information caused by trimming the audio sample,\n instead it splits it into equally-sized parts and pads them.\n To account for creation of multiple samples, number of labels are multiplied to a number\n equal to number of created samples.\n \n If proc_mode is 'resize':\n Resizes the original processed sample to (SAMPLE_DIM, N_SAMPLES) shape.\n \"\"\"\n \n assert proc_mode in ['split', 'resize'], 'proc_must be one of split or resize'\n \n filename = os.path.join(root, row['fname'])\n if not test_mode:\n labels = row['labels']\n \n sample = preprocFunc(filename, preprocParams, normalize=normalize)\n # print(sample.min(), sample.max())\n \n if proc_mode == 'split':\n sample_split = np.array_split(\n sample, np.ceil(sample.shape[1] / SAMPLE_DIM), axis=1)\n samples_pad = []\n for i in sample_split:\n padding_dim = SAMPLE_DIM - i.shape[1]\n sample_pad = cv2.copyMakeBorder(i, 0, 0, 0, padding_dim, trainingParams['padding_mode'])\n samples_pad.append(sample_pad)\n samples_pad = np.asarray(samples_pad)\n if not test_mode:\n labels = [labels] * len(samples_pad)\n labels = np.asarray(labels)\n return samples_pad, labels\n return samples_pad\n elif proc_mode == 'resize':\n sample_pad = cv2.resize(sample, (SAMPLE_DIM, N_SAMPLES), interpolation=cv2.INTER_NEAREST)\n sample_pad = np.expand_dims(sample_pad, axis=0)\n if not test_mode:\n labels = np.asarray(labels)\n return sample_pad, labels\n return sample_pad\n elif proc_mode == 'raw':\n if not test_model:\n return sample, labels\n return sample\n\n\nprocessor = DataProcessor()","execution_count":5,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"DATA_PREFIX = 'proc'\ntrain_filename = 'train_curated_{}.joblib'.format(DATA_PREFIX)\ntest_filename = 'test_{}.joblib'.format(DATA_PREFIX)\n\n\n# Train processing/loading:\nif os.path.isfile(train_filename):\n print('load processed train:')\n train_dict = joblib.load(train_filename)\n X_train = train_dict['X']\n y_train = train_dict['y']\n print(X_train.shape, y_train.shape)\nelse:\n print('process train...')\n output = Parallel(n_jobs=-3, verbose=1)(\n delayed(processor.prepareSample)(\n train_root, \n train_df.iloc[f, :],\n processor.createLogspec,\n CONFIG,\n TRAINING_CONFIG,\n test_mode=False,\n proc_mode='resize',\n ) for f in range(train_df.shape[0])) # change to number of sample in train data for full processing\n X_train = np.array([x[0] for x in output])\n y_train = np.array([x[1] for x in output])\n y_train = pd.Series(y_train).str.get_dummies(sep=',')\n print(X_train.shape, y_train.shape)\n # Save output for quicker experiments\n train_dict = {\n 'X': X_train,\n 'y': y_train,\n }\n joblib.dump(train_dict, train_filename)\n \n\n# Test processing/loading:\nif os.path.isfile(test_filename):\n print('load processed test:')\n test_dict = joblib.load(test_filename)\n X_test = test_dict['X']\n print(X_test.shape)\nelse:\n print('process test...')\n X_test = Parallel(n_jobs=-3, verbose=1)(\n delayed(processor.prepareSample)(\n test_root, \n sample_submission.iloc[f, :],\n processor.createLogspec,\n CONFIG,\n TRAINING_CONFIG,\n test_mode=True,\n proc_mode='resize',\n ) for f in range(sample_submission.shape[0])) # change to number of sample in test data for full processing\n X_test = np.array(X_test)\n print(X_test.shape)\n test_dict = {\n 'X': X_test,\n }\n joblib.dump(test_dict, test_filename)\n \n \n# Switch channel axis from 2nd to last\nX_train = np.moveaxis(X_train, 1, -1)\nX_test = np.moveaxis(X_test, 1, -1)\nprint(X_train.shape, y_train.shape)","execution_count":6,"outputs":[{"output_type":"stream","text":"process train...\n","name":"stdout"},{"output_type":"stream","text":"[Parallel(n_jobs=-3)]: Using backend SequentialBackend with 1 concurrent workers.\n[Parallel(n_jobs=-3)]: Done 4970 out of 4970 | elapsed: 3.0min finished\n","name":"stderr"},{"output_type":"stream","text":"(4970, 1, 48, 256) (4970, 80)\nprocess test...\n","name":"stdout"},{"output_type":"stream","text":"[Parallel(n_jobs=-3)]: Using backend SequentialBackend with 1 concurrent workers.\n","name":"stderr"},{"output_type":"stream","text":"(1120, 1, 48, 256)\n(4970, 48, 256, 1) (4970, 80)\n","name":"stdout"},{"output_type":"stream","text":"[Parallel(n_jobs=-3)]: Done 1120 out of 1120 | elapsed: 53.3s finished\n","name":"stderr"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Distribution of multilabel labels\nprint('Multilabel class distribution:')\nprint(y_train.sum(axis=1).value_counts())\n\n# Most of the samples belong to only one class.\n# There are some (around 15%), which belong to two classes.\n# Occurrence of samples belonging to more than two classes at once\n# is quite rare, around 1.5%.","execution_count":7,"outputs":[{"output_type":"stream","text":"Multilabel class distribution:\n1 4269\n2 627\n3 69\n4 4\n6 1\ndtype: int64\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"from collections import Counter\nfrom pprint import pprint\n\ndf_lab = (y_train.loc[y_train.sum(axis=1) > 1] > 0)\nmultilabel_combs = []\nfor i in range(df_lab.shape[0]):\n row_label = df_lab.iloc[i, :][df_lab.iloc[i, :] > 0].index.tolist()\n multilabel_combs.append(row_label)\n\nmultilabel_comb_counter = Counter(list(map(lambda x: ' + '.join(x), multilabel_combs)))\npprint(multilabel_comb_counter.most_common(20))\n# 20 most common combinations of labels","execution_count":8,"outputs":[{"output_type":"stream","text":"[('Acoustic_guitar + Strum', 69),\n ('Drip + Raindrop', 52),\n ('Sink_(filling_or_washing) + Water_tap_and_faucet', 46),\n ('Applause + Cheering + Crowd', 36),\n ('Cutlery_and_silverware + Dishes_and_pots_and_pans', 29),\n ('Chink_and_clink + Dishes_and_pots_and_pans', 27),\n ('Accelerating_and_revving_and_vroom + Race_car_and_auto_racing', 26),\n ('Female_speech_and_woman_speaking + Yell', 22),\n ('Slam + Squeak', 21),\n ('Screaming + Yell', 20),\n ('Applause + Cheering', 15),\n ('Meow + Purr', 13),\n ('Accelerating_and_revving_and_vroom + Motorcycle', 12),\n ('Squeak + Walk_and_footsteps', 12),\n ('Car_passing_by + Traffic_noise_and_roadway_noise', 10),\n ('Applause + Crowd', 10),\n ('Bass_guitar + Electric_guitar', 10),\n ('Gurgling + Toilet_flush', 10),\n ('Female_singing + Male_singing', 9),\n ('Cheering + Crowd', 8)]\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"import numpy as np\nimport sklearn.metrics\n\n\n# Based on https://www.kaggle.com/voglinio/keras-2d-model-5-fold-log-specgram-curated-only\n# Core calculation of label precisions for one test sample.\ndef _one_sample_positive_class_precisions(scores, truth):\n \"\"\"Calculate precisions for each true class for a single sample.\n\n Args:\n scores: np.array of (num_classes,) giving the individual classifier scores.\n truth: np.array of (num_classes,) bools indicating which classes are true.\n\n Returns:\n pos_class_indices: np.array of indices of the true classes for this sample.\n pos_class_precisions: np.array of precisions corresponding to each of those\n classes.\n \"\"\"\n num_classes = scores.shape[0]\n pos_class_indices = np.flatnonzero(truth > 0)\n # Only calculate precisions if there are some true classes.\n if not len(pos_class_indices):\n return pos_class_indices, np.zeros(0)\n # Retrieval list of classes for this sample.\n retrieved_classes = np.argsort(scores)[::-1]\n # class_rankings[top_scoring_class_index] == 0 etc.\n class_rankings = np.zeros(num_classes, dtype=np.int)\n class_rankings[retrieved_classes] = range(num_classes)\n # Which of these is a true label?\n retrieved_class_true = np.zeros(num_classes, dtype=np.bool)\n retrieved_class_true[class_rankings[pos_class_indices]] = True\n # Num hits for every truncated retrieval list.\n retrieved_cumulative_hits = np.cumsum(retrieved_class_true)\n # Precision of retrieval list truncated at each hit, in order of pos_labels.\n precision_at_hits = (\n retrieved_cumulative_hits[class_rankings[pos_class_indices]] /\n (1 + class_rankings[pos_class_indices].astype(np.float)))\n return pos_class_indices, precision_at_hits\n\n# All-in-one calculation of per-class lwlrap.\n\n\ndef calculate_per_class_lwlrap(truth, scores):\n \"\"\"Calculate label-weighted label-ranking average precision.\n\n Arguments:\n truth: np.array of (num_samples, num_classes) giving boolean ground-truth\n of presence of that class in that sample.\n scores: np.array of (num_samples, num_classes) giving the classifier-under-\n test's real-valued score for each class for each sample.\n\n Returns:\n per_class_lwlrap: np.array of (num_classes,) giving the lwlrap for each\n class.\n weight_per_class: np.array of (num_classes,) giving the prior of each\n class within the truth labels. Then the overall unbalanced lwlrap is\n simply np.sum(per_class_lwlrap * weight_per_class)\n \"\"\"\n assert truth.shape == scores.shape\n num_samples, num_classes = scores.shape\n # Space to store a distinct precision value for each class on each sample.\n # Only the classes that are true for each sample will be filled in.\n precisions_for_samples_by_classes = np.zeros((num_samples, num_classes))\n for sample_num in range(num_samples):\n pos_class_indices, precision_at_hits = (\n _one_sample_positive_class_precisions(scores[sample_num, :],\n truth[sample_num, :]))\n precisions_for_samples_by_classes[sample_num, pos_class_indices] = (\n precision_at_hits)\n labels_per_class = np.sum(truth > 0, axis=0)\n weight_per_class = labels_per_class / float(np.sum(labels_per_class))\n # Form average of each column, i.e. all the precisions assigned to labels in\n # a particular class.\n per_class_lwlrap = (np.sum(precisions_for_samples_by_classes, axis=0) /\n np.maximum(1, labels_per_class))\n # overall_lwlrap = simple average of all the actual per-class, per-sample precisions\n # = np.sum(precisions_for_samples_by_classes) / np.sum(precisions_for_samples_by_classes > 0)\n # also = weighted mean of per-class lwlraps, weighted by class label prior across samples\n # = np.sum(per_class_lwlrap * weight_per_class)\n return per_class_lwlrap, weight_per_class\n\n# Calculate the overall lwlrap using sklearn.metrics function.\n\n\ndef calculate_overall_lwlrap_sklearn(truth, scores):\n \"\"\"Calculate the overall lwlrap using sklearn.metrics.lrap.\"\"\"\n # sklearn doesn't correctly apply weighting to samples with no labels, so just skip them.\n sample_weight = np.sum(truth > 0, axis=1)\n nonzero_weight_sample_indices = np.flatnonzero(sample_weight > 0)\n overall_lwlrap = sklearn.metrics.label_ranking_average_precision_score(\n truth[nonzero_weight_sample_indices, :] > 0,\n scores[nonzero_weight_sample_indices, :],\n sample_weight=sample_weight[nonzero_weight_sample_indices])\n return overall_lwlrap\n\n\n# Accumulator object version.\n\nclass lwlrap_accumulator(object):\n \"\"\"Accumulate batches of test samples into per-class and overall lwlrap.\"\"\"\n\n def __init__(self):\n self.num_classes = 0\n self.total_num_samples = 0\n\n def accumulate_samples(self, batch_truth, batch_scores):\n \"\"\"Cumulate a new batch of samples into the metric.\n\n Args:\n truth: np.array of (num_samples, num_classes) giving boolean\n ground-truth of presence of that class in that sample for this batch.\n scores: np.array of (num_samples, num_classes) giving the\n classifier-under-test's real-valued score for each class for each\n sample.\n \"\"\"\n assert batch_scores.shape == batch_truth.shape\n num_samples, num_classes = batch_truth.shape\n if not self.num_classes:\n self.num_classes = num_classes\n self._per_class_cumulative_precision = np.zeros(self.num_classes)\n self._per_class_cumulative_count = np.zeros(self.num_classes,\n dtype=np.int)\n assert num_classes == self.num_classes\n for truth, scores in zip(batch_truth, batch_scores):\n pos_class_indices, precision_at_hits = (\n _one_sample_positive_class_precisions(scores, truth))\n self._per_class_cumulative_precision[pos_class_indices] += (\n precision_at_hits)\n self._per_class_cumulative_count[pos_class_indices] += 1\n self.total_num_samples += num_samples\n\n def per_class_lwlrap(self):\n \"\"\"Return a vector of the per-class lwlraps for the accumulated samples.\"\"\"\n return (self._per_class_cumulative_precision /\n np.maximum(1, self._per_class_cumulative_count))\n\n def per_class_weight(self):\n \"\"\"Return a normalized weight vector for the contributions of each class.\"\"\"\n return (self._per_class_cumulative_count /\n float(np.sum(self._per_class_cumulative_count)))\n\n def overall_lwlrap(self):\n \"\"\"Return the scalar overall lwlrap for cumulated samples.\"\"\"\n return np.sum(self.per_class_lwlrap() * self.per_class_weight())","execution_count":9,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# import tensorflow.keras as keras\n# from tensorflow.keras import layers\n# from tensorflow.keras.callbacks import *\n# from tensorflow.keras.layers import *\n# from tensorflow.keras.optimizers import *\n# from tensorflow.keras.losses import *\n# from tensorflow.keras.models import Model\n\nfrom keras import layers\nfrom keras.callbacks import *\nfrom keras.layers import *\nfrom keras.optimizers import *\nfrom keras.losses import *\nfrom keras.models import Model\n\n\n# Based on https://github.com/keras-team/keras-applications/blob/master/keras_applications/resnet_common.py\ndef block1(x, filters, kernel_size=3, stride=1,\n conv_shortcut=True, name=None):\n \"\"\"A residual block.\n # Arguments\n x: input tensor.\n filters: integer, filters of the bottleneck layer.\n kernel_size: default 3, kernel size of the bottleneck layer.\n stride: default 1, stride of the first layer.\n conv_shortcut: default True, use convolution shortcut if True,\n otherwise identity shortcut.\n name: string, block label.\n # Returns\n Output tensor for the residual block.\n \"\"\"\n bn_axis = 3\n\n if conv_shortcut is True:\n shortcut = layers.Conv2D(4 * filters, 1, strides=stride,\n name=name + '_0_conv')(x)\n shortcut = layers.BatchNormalization(axis=bn_axis, epsilon=1.001e-5,\n name=name + '_0_bn')(shortcut)\n else:\n shortcut = x\n\n x = layers.Conv2D(filters, 1, strides=stride, name=name + '_1_conv')(x)\n x = layers.BatchNormalization(axis=bn_axis, epsilon=1.001e-5,\n name=name + '_1_bn')(x)\n x = layers.PReLU(name=name + '_1_prelu')(x)\n\n x = layers.Conv2D(filters, kernel_size, padding='SAME',\n name=name + '_2_conv')(x)\n x = layers.BatchNormalization(axis=bn_axis, epsilon=1.001e-5,\n name=name + '_2_bn')(x)\n x = layers.PReLU(name=name + '_2_prelu')(x)\n\n x = layers.Conv2D(4 * filters, 1, name=name + '_3_conv')(x)\n x = layers.BatchNormalization(axis=bn_axis, epsilon=1.001e-5,\n name=name + '_3_bn')(x)\n\n x = layers.Add(name=name + '_add')([shortcut, x])\n x = layers.PReLU(name=name + '_out')(x)\n \n return x\n\n\ndef stack1(x, filters, blocks, stride1=2, name=None):\n \"\"\"A set of stacked residual blocks.\n # Arguments\n x: input tensor.\n filters: integer, filters of the bottleneck layer in a block.\n blocks: integer, blocks in the stacked blocks.\n stride1: default 2, stride of the first layer in the first block.\n name: string, stack label.\n # Returns\n Output tensor for the stacked blocks.\n \"\"\"\n x = block1(x, filters, stride=stride1, name=name + '_block1')\n for i in range(2, blocks + 1):\n x = block1(x, filters, conv_shortcut=False, name=name + '_block' + str(i))\n return x\n\n\ndef ResNetlike(input_shape, num_classes):\n \n use_bias = False\n num_blocks = 1\n strides = (2, 4)\n \n input_layer = layers.Input(input_shape)\n # x = layers.ZeroPadding2D(padding=((3, 3), (3, 3)), name='conv1_pad')(input_layer)\n x = layers.Conv2D(64, (3, 7), strides=2, use_bias=use_bias, name='conv1_conv')(input_layer)\n x = layers.BatchNormalization(axis=bn_axis, epsilon=1.001e-5,\n name='conv1_bn')(x)\n x = layers.PReLU(name='conv1_prelu')(x)\n \n x = stack1(x, 64, num_blocks, stride1=1, name='conv2')\n x = stack1(x, 64, num_blocks, stride1=strides, name='conv3')\n x = stack1(x, 64, num_blocks, stride1=strides, name='conv4')\n \n x = layers.GlobalAveragePooling2D(name='avg_pool')(x)\n x = layers.Dense(num_classes, activation='softmax', name='probs')(x)\n \n model = Model(input_layer, x, name='resnetlike')\n model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n \n return model","execution_count":10,"outputs":[{"output_type":"stream","text":"Using TensorFlow backend.\n","name":"stderr"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"from sklearn.preprocessing import MultiLabelBinarizer, LabelEncoder\n\n\n# Inspired by https://www.kaggle.com/yekenot/pooled-gru-fasttext\nclass LwlRapEvaluation(Callback):\n def __init__(self, validation_data=(), interval=1):\n super(Callback, self).__init__()\n self.interval = interval\n self.X_val, self.y_val = validation_data\n\n def on_epoch_end(self, epoch, logs={}):\n if epoch % self.interval == 0:\n y_pred = self.model.predict(self.X_val, verbose=0)\n score = calculate_overall_lwlrap_sklearn(self.y_val, y_pred)\n print(\"\\n LWLRAP - epoch: {} - score: {:.4f} \\n\".format(epoch +1, score))\n return\n\n\n# Based on https://www.kaggle.com/voglinio/keras-2d-model-5-fold-log-specgram-curated-only\ndef create_unique_labels(all_labels):\n label_dict = {}\n all_labels_set = []\n first_labels_set = []\n for labs in all_labels:\n lab = labs.split(',')\n for l in lab:\n if l in label_dict:\n label_dict[l] = label_dict[l] + 1\n else:\n label_dict[l]= 0\n\n all_labels_set.append(set(lab))\n first_labels_set.append(lab[0])\n classes = list(label_dict.keys())\n \n return label_dict, classes, all_labels_set, first_labels_set\n\n\nlabel_dict, classes, all_labels_set, first_labels_set = create_unique_labels(train_df.labels)\nbinarize = MultiLabelBinarizer(classes=classes)\ny_cat = binarize.fit_transform(all_labels_set)","execution_count":11,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"RUN_NAME = 'ResNetLike_mel'\nBATCH_SIZE = 32\nNUM_EPOCHS = 35\nNFOLDS = 5\nSEED = 1337\nLOAD_MODEL = False\nTRAIN_MODEL = True\nDEBUG = False\nSAVE_SUBMISSION = True\n\n\n# y_cat for categorical_crossentropy\n# y_train.values for binary_crossentropy\nlabels_set = y_cat \nkfold = KFold(n_splits=NFOLDS, random_state=SEED)\nnum_classes = 80\nbn_axis = 3\n\noof_train = np.zeros((X_train.shape[0], num_classes))\noof_test = np.zeros((X_test.shape[0], num_classes, NFOLDS))\nvalid_scores = []\n\n\n# KFold training\nfor fold_idx, (train_idx, valid_idx) in enumerate(kfold.split(X_train)):\n X_tr, y_tr = X_train[train_idx], labels_set[train_idx]\n X_val, y_val = X_train[valid_idx], labels_set[valid_idx]\n print(X_tr.shape, y_tr.shape)\n print(X_val.shape, y_val.shape)\n \n checkpoint_name = '{}_{}.h5'.format(RUN_NAME, fold_idx)\n lwl_callback = LwlRapEvaluation(validation_data=(X_val, y_val))\n \n callbacks_list = [\n ModelCheckpoint(checkpoint_name, save_best_only=True, save_weights_only=True),\n ReduceLROnPlateau(patience=5, factor=0.2),\n EarlyStopping(patience=8, monitor='val_loss', restore_best_weights=True),\n lwl_callback,\n ]\n \n model = ResNetlike(X_train.shape[1:], y_train.shape[-1])\n # model = BaselineModel(X_train.shape[1:])\n # model.summary()\n if LOAD_MODEL:\n print('loading: {}'.format(checkpoint_name))\n model.load_weights(checkpoint_name)\n if TRAIN_MODEL:\n print('training model...')\n model.fit(\n X_tr, y_tr, \n validation_data=(X_val, y_val),\n batch_size=BATCH_SIZE,\n epochs=NUM_EPOCHS,\n verbose=1,\n callbacks=callbacks_list)\n print('loading best weights from current fold')\n model.load_weights(checkpoint_name)\n \n val_pred = model.predict(X_val, batch_size=BATCH_SIZE)\n oof_train[valid_idx, :] = val_pred\n oof_test[:, :, fold_idx] = model.predict(X_test, batch_size=BATCH_SIZE)\n \n val_lwlrap = calculate_overall_lwlrap_sklearn(y_val, val_pred)\n valid_scores.append(val_lwlrap)\n print(\"lwlrap fold: {:.4f}\".format(val_lwlrap))\n # break\n \n \noof_lwl = calculate_overall_lwlrap_sklearn(labels_set, oof_train)\nprint('OOF LWLRAP: {:.4f}'.format(oof_lwl))","execution_count":13,"outputs":[{"output_type":"stream","text":"(3976, 48, 256, 1) (3976, 80)\n(994, 48, 256, 1) (994, 80)\ntraining model...\nTrain on 3976 samples, validate on 994 samples\nEpoch 1/35\n3976/3976 [==============================] - 11s 3ms/step - loss: 4.5081 - acc: 0.0978 - val_loss: 18.0130 - val_acc: 0.0221\n\n LWLRAP - epoch: 1 - score: 0.0782 \n\nEpoch 2/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.6574 - acc: 0.2165 - val_loss: 13.2957 - val_acc: 0.0191\n\n LWLRAP - epoch: 2 - score: 0.0975 \n\nEpoch 3/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.2246 - acc: 0.2918 - val_loss: 16.9711 - val_acc: 0.0191\n\n LWLRAP - epoch: 3 - score: 0.0830 \n\nEpoch 4/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.9238 - acc: 0.3549 - val_loss: 15.0006 - val_acc: 0.0161\n\n LWLRAP - epoch: 4 - score: 0.0762 \n\nEpoch 5/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.6551 - acc: 0.4007 - val_loss: 14.5814 - val_acc: 0.0201\n\n LWLRAP - epoch: 5 - score: 0.0893 \n\nEpoch 6/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.3557 - acc: 0.4605 - val_loss: 12.7641 - val_acc: 0.0201\n\n LWLRAP - epoch: 6 - score: 0.0921 \n\nEpoch 7/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.1144 - acc: 0.5053 - val_loss: 12.1922 - val_acc: 0.0302\n\n LWLRAP - epoch: 7 - score: 0.1022 \n\nEpoch 8/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.8421 - acc: 0.5757 - val_loss: 14.1850 - val_acc: 0.0332\n\n LWLRAP - epoch: 8 - score: 0.0910 \n\nEpoch 9/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.5682 - acc: 0.6466 - val_loss: 14.2001 - val_acc: 0.0211\n\n LWLRAP - epoch: 9 - score: 0.0865 \n\nEpoch 10/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.3243 - acc: 0.7100 - val_loss: 9.8503 - val_acc: 0.0724\n\n LWLRAP - epoch: 10 - score: 0.1629 \n\nEpoch 11/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.0845 - acc: 0.7651 - val_loss: 14.5469 - val_acc: 0.0292\n\n LWLRAP - epoch: 11 - score: 0.1000 \n\nEpoch 12/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.9425 - acc: 0.8036 - val_loss: 10.9716 - val_acc: 0.0322\n\n LWLRAP - epoch: 12 - score: 0.1097 \n\nEpoch 13/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.8246 - acc: 0.8418 - val_loss: 15.0683 - val_acc: 0.0231\n\n LWLRAP - epoch: 13 - score: 0.1027 \n\nEpoch 14/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6624 - acc: 0.8785 - val_loss: 16.1527 - val_acc: 0.0252\n\n LWLRAP - epoch: 14 - score: 0.1055 \n\nEpoch 15/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6359 - acc: 0.8836 - val_loss: 9.9333 - val_acc: 0.0644\n\n LWLRAP - epoch: 15 - score: 0.1513 \n\nEpoch 16/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4843 - acc: 0.9173 - val_loss: 9.1433 - val_acc: 0.1288\n\n LWLRAP - epoch: 16 - score: 0.2522 \n\nEpoch 17/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3849 - acc: 0.9248 - val_loss: 6.3827 - val_acc: 0.1942\n\n LWLRAP - epoch: 17 - score: 0.3422 \n\nEpoch 18/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3595 - acc: 0.9278 - val_loss: 8.7835 - val_acc: 0.1579\n\n LWLRAP - epoch: 18 - score: 0.2742 \n\nEpoch 19/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3464 - acc: 0.9291 - val_loss: 11.0564 - val_acc: 0.1137\n\n LWLRAP - epoch: 19 - score: 0.2394 \n\nEpoch 20/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3380 - acc: 0.9258 - val_loss: 5.5691 - val_acc: 0.2143\n\n LWLRAP - epoch: 20 - score: 0.3642 \n\nEpoch 21/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3341 - acc: 0.9243 - val_loss: 12.9984 - val_acc: 0.0755\n\n LWLRAP - epoch: 21 - score: 0.1697 \n\nEpoch 22/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3308 - acc: 0.9276 - val_loss: 9.8545 - val_acc: 0.1127\n\n LWLRAP - epoch: 22 - score: 0.2188 \n\nEpoch 23/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3268 - acc: 0.9281 - val_loss: 9.2912 - val_acc: 0.1338\n\n LWLRAP - epoch: 23 - score: 0.2553 \n\nEpoch 24/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3219 - acc: 0.9261 - val_loss: 10.8660 - val_acc: 0.1107\n\n LWLRAP - epoch: 24 - score: 0.2226 \n\nEpoch 25/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3213 - acc: 0.9283 - val_loss: 6.1751 - val_acc: 0.2133\n\n LWLRAP - epoch: 25 - score: 0.3431 \n\nEpoch 26/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3234 - acc: 0.9218 - val_loss: 8.5180 - val_acc: 0.1680\n\n LWLRAP - epoch: 26 - score: 0.2867 \n\nEpoch 27/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2966 - acc: 0.9374 - val_loss: 4.9604 - val_acc: 0.2716\n\n LWLRAP - epoch: 27 - score: 0.4300 \n\nEpoch 28/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2950 - acc: 0.9266 - val_loss: 6.4939 - val_acc: 0.2254\n\n LWLRAP - epoch: 28 - score: 0.3811 \n\nEpoch 29/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2907 - acc: 0.9263 - val_loss: 9.9135 - val_acc: 0.1811\n\n LWLRAP - epoch: 29 - score: 0.3072 \n\nEpoch 30/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2917 - acc: 0.9306 - val_loss: 8.6049 - val_acc: 0.1821\n\n LWLRAP - epoch: 30 - score: 0.3000 \n\nEpoch 31/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2881 - acc: 0.9258 - val_loss: 6.2872 - val_acc: 0.2384\n\n LWLRAP - epoch: 31 - score: 0.3815 \n\nEpoch 32/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2859 - acc: 0.9298 - val_loss: 8.0057 - val_acc: 0.1942\n\n LWLRAP - epoch: 32 - score: 0.3313 \n\nEpoch 33/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2847 - acc: 0.9281 - val_loss: 3.0566 - val_acc: 0.4135\n\n LWLRAP - epoch: 33 - score: 0.5672 \n\nEpoch 34/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2858 - acc: 0.9288 - val_loss: 3.0095 - val_acc: 0.4205\n\n LWLRAP - epoch: 34 - score: 0.5723 \n\nEpoch 35/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2837 - acc: 0.9341 - val_loss: 3.0776 - val_acc: 0.4095\n\n LWLRAP - epoch: 35 - score: 0.5637 \n\nloading best weights from current fold\nlwlrap fold: 0.5723\n(3976, 48, 256, 1) (3976, 80)\n(994, 48, 256, 1) (994, 80)\ntraining model...\nTrain on 3976 samples, validate on 994 samples\nEpoch 1/35\n3976/3976 [==============================] - 12s 3ms/step - loss: 4.4912 - acc: 0.1099 - val_loss: 17.7345 - val_acc: 0.0131\n\n LWLRAP - epoch: 1 - score: 0.0655 \n\nEpoch 2/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.6820 - acc: 0.2145 - val_loss: 12.9391 - val_acc: 0.0161\n\n LWLRAP - epoch: 2 - score: 0.0869 \n\nEpoch 3/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.2310 - acc: 0.2938 - val_loss: 17.5431 - val_acc: 0.0131\n\n LWLRAP - epoch: 3 - score: 0.0675 \n\nEpoch 4/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.9186 - acc: 0.3461 - val_loss: 14.1680 - val_acc: 0.0171\n\n LWLRAP - epoch: 4 - score: 0.0905 \n\nEpoch 5/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.6381 - acc: 0.4115 - val_loss: 17.2281 - val_acc: 0.0111\n\n LWLRAP - epoch: 5 - score: 0.0647 \n\nEpoch 6/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.3844 - acc: 0.4590 - val_loss: 13.7789 - val_acc: 0.0161\n\n LWLRAP - epoch: 6 - score: 0.0764 \n\nEpoch 7/35\n3976/3976 [==============================] - 8s 2ms/step - loss: 2.1333 - acc: 0.5108 - val_loss: 11.3177 - val_acc: 0.0131\n\n LWLRAP - epoch: 7 - score: 0.0813 \n\nEpoch 8/35\n3976/3976 [==============================] - 8s 2ms/step - loss: 1.8822 - acc: 0.5589 - val_loss: 12.6311 - val_acc: 0.0191\n\n LWLRAP - epoch: 8 - score: 0.0889 \n\nEpoch 9/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.6305 - acc: 0.6222 - val_loss: 10.7489 - val_acc: 0.0292\n\n LWLRAP - epoch: 9 - score: 0.1000 \n\nEpoch 10/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.3915 - acc: 0.6854 - val_loss: 10.0830 - val_acc: 0.0362\n\n LWLRAP - epoch: 10 - score: 0.1095 \n\nEpoch 11/35\n","name":"stdout"},{"output_type":"stream","text":"3976/3976 [==============================] - 7s 2ms/step - loss: 1.1634 - acc: 0.7399 - val_loss: 14.8250 - val_acc: 0.0302\n\n LWLRAP - epoch: 11 - score: 0.0901 \n\nEpoch 12/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.9420 - acc: 0.8068 - val_loss: 14.0706 - val_acc: 0.0191\n\n LWLRAP - epoch: 12 - score: 0.0959 \n\nEpoch 13/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7937 - acc: 0.8519 - val_loss: 10.9775 - val_acc: 0.0352\n\n LWLRAP - epoch: 13 - score: 0.1158 \n\nEpoch 14/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7512 - acc: 0.8546 - val_loss: 14.6807 - val_acc: 0.0221\n\n LWLRAP - epoch: 14 - score: 0.0974 \n\nEpoch 15/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6811 - acc: 0.8732 - val_loss: 11.3777 - val_acc: 0.0473\n\n LWLRAP - epoch: 15 - score: 0.1346 \n\nEpoch 16/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4891 - acc: 0.9135 - val_loss: 8.4840 - val_acc: 0.1388\n\n LWLRAP - epoch: 16 - score: 0.2621 \n\nEpoch 17/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4109 - acc: 0.9258 - val_loss: 8.4312 - val_acc: 0.1479\n\n LWLRAP - epoch: 17 - score: 0.2658 \n\nEpoch 18/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3851 - acc: 0.9245 - val_loss: 11.5984 - val_acc: 0.0734\n\n LWLRAP - epoch: 18 - score: 0.1959 \n\nEpoch 19/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3705 - acc: 0.9147 - val_loss: 8.0491 - val_acc: 0.1388\n\n LWLRAP - epoch: 19 - score: 0.2678 \n\nEpoch 20/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3624 - acc: 0.9253 - val_loss: 13.0432 - val_acc: 0.0493\n\n LWLRAP - epoch: 20 - score: 0.1563 \n\nEpoch 21/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3562 - acc: 0.9245 - val_loss: 3.9541 - val_acc: 0.3028\n\n LWLRAP - epoch: 21 - score: 0.4595 \n\nEpoch 22/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3511 - acc: 0.9258 - val_loss: 9.5933 - val_acc: 0.1408\n\n LWLRAP - epoch: 22 - score: 0.2611 \n\nEpoch 23/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3423 - acc: 0.9208 - val_loss: 8.3903 - val_acc: 0.1378\n\n LWLRAP - epoch: 23 - score: 0.2446 \n\nEpoch 24/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3449 - acc: 0.9223 - val_loss: 9.3713 - val_acc: 0.1006\n\n LWLRAP - epoch: 24 - score: 0.2228 \n\nEpoch 25/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3446 - acc: 0.9271 - val_loss: 8.1476 - val_acc: 0.1479\n\n LWLRAP - epoch: 25 - score: 0.2670 \n\nEpoch 26/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3361 - acc: 0.9230 - val_loss: 11.6523 - val_acc: 0.0775\n\n LWLRAP - epoch: 26 - score: 0.1991 \n\nEpoch 27/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3281 - acc: 0.9245 - val_loss: 8.7643 - val_acc: 0.1600\n\n LWLRAP - epoch: 27 - score: 0.2784 \n\nEpoch 28/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3142 - acc: 0.9258 - val_loss: 4.1688 - val_acc: 0.3169\n\n LWLRAP - epoch: 28 - score: 0.4690 \n\nEpoch 29/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3089 - acc: 0.9271 - val_loss: 4.0470 - val_acc: 0.3290\n\n LWLRAP - epoch: 29 - score: 0.4595 \n\nloading best weights from current fold\nlwlrap fold: 0.4595\n(3976, 48, 256, 1) (3976, 80)\n(994, 48, 256, 1) (994, 80)\ntraining model...\nTrain on 3976 samples, validate on 994 samples\nEpoch 1/35\n3976/3976 [==============================] - 13s 3ms/step - loss: 4.4886 - acc: 0.0971 - val_loss: 18.4201 - val_acc: 0.0121\n\n LWLRAP - epoch: 1 - score: 0.0695 \n\nEpoch 2/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.6733 - acc: 0.2118 - val_loss: 8.7769 - val_acc: 0.0352\n\n LWLRAP - epoch: 2 - score: 0.1048 \n\nEpoch 3/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.2693 - acc: 0.2892 - val_loss: 13.1905 - val_acc: 0.0201\n\n LWLRAP - epoch: 3 - score: 0.0970 \n\nEpoch 4/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.9345 - acc: 0.3375 - val_loss: 17.9374 - val_acc: 0.0151\n\n LWLRAP - epoch: 4 - score: 0.0714 \n\nEpoch 5/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.6794 - acc: 0.3871 - val_loss: 15.5190 - val_acc: 0.0191\n\n LWLRAP - epoch: 5 - score: 0.0845 \n\nEpoch 6/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.4061 - acc: 0.4369 - val_loss: 11.5776 - val_acc: 0.0272\n\n LWLRAP - epoch: 6 - score: 0.0989 \n\nEpoch 7/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.1636 - acc: 0.4902 - val_loss: 18.2851 - val_acc: 0.0131\n\n LWLRAP - epoch: 7 - score: 0.0735 \n\nEpoch 8/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.6872 - acc: 0.6313 - val_loss: 10.0217 - val_acc: 0.0986\n\n LWLRAP - epoch: 8 - score: 0.2101 \n\nEpoch 9/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.5060 - acc: 0.6735 - val_loss: 7.7934 - val_acc: 0.1268\n\n LWLRAP - epoch: 9 - score: 0.2317 \n\nEpoch 10/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.3822 - acc: 0.7150 - val_loss: 5.0467 - val_acc: 0.1579\n\n LWLRAP - epoch: 10 - score: 0.3131 \n\nEpoch 11/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.2767 - acc: 0.7389 - val_loss: 5.1305 - val_acc: 0.1670\n\n LWLRAP - epoch: 11 - score: 0.3049 \n\nEpoch 12/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.1759 - acc: 0.7696 - val_loss: 13.2783 - val_acc: 0.0543\n\n LWLRAP - epoch: 12 - score: 0.1475 \n\nEpoch 13/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.0980 - acc: 0.7958 - val_loss: 7.8284 - val_acc: 0.1258\n\n LWLRAP - epoch: 13 - score: 0.2392 \n\nEpoch 14/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.9923 - acc: 0.8247 - val_loss: 10.1160 - val_acc: 0.1036\n\n LWLRAP - epoch: 14 - score: 0.2279 \n\nEpoch 15/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.9130 - acc: 0.8290 - val_loss: 9.3173 - val_acc: 0.1036\n\n LWLRAP - epoch: 15 - score: 0.1981 \n\nEpoch 16/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.8078 - acc: 0.8717 - val_loss: 4.4585 - val_acc: 0.2626\n\n LWLRAP - epoch: 16 - score: 0.4197 \n\nEpoch 17/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7785 - acc: 0.8788 - val_loss: 6.3783 - val_acc: 0.1761\n\n LWLRAP - epoch: 17 - score: 0.3283 \n\nEpoch 18/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7615 - acc: 0.8765 - val_loss: 3.0787 - val_acc: 0.3773\n\n LWLRAP - epoch: 18 - score: 0.5394 \n\nEpoch 19/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7414 - acc: 0.8836 - val_loss: 2.9717 - val_acc: 0.3984\n\n LWLRAP - epoch: 19 - score: 0.5555 \n\nEpoch 20/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7273 - acc: 0.8876 - val_loss: 4.6422 - val_acc: 0.2495\n\n LWLRAP - epoch: 20 - score: 0.3996 \n\nEpoch 21/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7014 - acc: 0.8898 - val_loss: 4.6625 - val_acc: 0.2565\n\n LWLRAP - epoch: 21 - score: 0.4133 \n\nEpoch 22/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7035 - acc: 0.8959 - val_loss: 2.9910 - val_acc: 0.3833\n\n LWLRAP - epoch: 22 - score: 0.5499 \n\nEpoch 23/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6939 - acc: 0.8911 - val_loss: 12.1379 - val_acc: 0.1006\n\n LWLRAP - epoch: 23 - score: 0.1898 \n\nEpoch 24/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6746 - acc: 0.8944 - val_loss: 3.0281 - val_acc: 0.3773\n\n LWLRAP - epoch: 24 - score: 0.5434 \n\nEpoch 25/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6545 - acc: 0.9062 - val_loss: 2.8843 - val_acc: 0.4205\n\n LWLRAP - epoch: 25 - score: 0.5763 \n\nEpoch 26/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6537 - acc: 0.9007 - val_loss: 2.9557 - val_acc: 0.4004\n\n LWLRAP - epoch: 26 - score: 0.5674 \n\nEpoch 27/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6536 - acc: 0.9024 - val_loss: 2.8553 - val_acc: 0.4175\n\n LWLRAP - epoch: 27 - score: 0.5791 \n\nEpoch 28/35\n","name":"stdout"},{"output_type":"stream","text":"3976/3976 [==============================] - 7s 2ms/step - loss: 0.6497 - acc: 0.9017 - val_loss: 2.9386 - val_acc: 0.4074\n\n LWLRAP - epoch: 28 - score: 0.5672 \n\nEpoch 29/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6381 - acc: 0.8999 - val_loss: 2.9227 - val_acc: 0.4145\n\n LWLRAP - epoch: 29 - score: 0.5707 \n\nEpoch 30/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6345 - acc: 0.9014 - val_loss: 2.9752 - val_acc: 0.3954\n\n LWLRAP - epoch: 30 - score: 0.5547 \n\nEpoch 31/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6271 - acc: 0.9067 - val_loss: 3.2608 - val_acc: 0.3622\n\n LWLRAP - epoch: 31 - score: 0.5286 \n\nEpoch 32/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6273 - acc: 0.9082 - val_loss: 2.9809 - val_acc: 0.3954\n\n LWLRAP - epoch: 32 - score: 0.5627 \n\nEpoch 33/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6196 - acc: 0.9074 - val_loss: 2.8463 - val_acc: 0.4235\n\n LWLRAP - epoch: 33 - score: 0.5816 \n\nEpoch 34/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6333 - acc: 0.9024 - val_loss: 2.8327 - val_acc: 0.4165\n\n LWLRAP - epoch: 34 - score: 0.5810 \n\nEpoch 35/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6203 - acc: 0.9074 - val_loss: 2.8381 - val_acc: 0.4135\n\n LWLRAP - epoch: 35 - score: 0.5786 \n\nloading best weights from current fold\nlwlrap fold: 0.5810\n(3976, 48, 256, 1) (3976, 80)\n(994, 48, 256, 1) (994, 80)\ntraining model...\nTrain on 3976 samples, validate on 994 samples\nEpoch 1/35\n3976/3976 [==============================] - 13s 3ms/step - loss: 4.5014 - acc: 0.1004 - val_loss: 17.7738 - val_acc: 0.0201\n\n LWLRAP - epoch: 1 - score: 0.0710 \n\nEpoch 2/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.6592 - acc: 0.2213 - val_loss: 12.3252 - val_acc: 0.0080\n\n LWLRAP - epoch: 2 - score: 0.0812 \n\nEpoch 3/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.2380 - acc: 0.2935 - val_loss: 16.8519 - val_acc: 0.0181\n\n LWLRAP - epoch: 3 - score: 0.0799 \n\nEpoch 4/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.9057 - acc: 0.3488 - val_loss: 11.1960 - val_acc: 0.0241\n\n LWLRAP - epoch: 4 - score: 0.1056 \n\nEpoch 5/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.6619 - acc: 0.3999 - val_loss: 18.2489 - val_acc: 0.0161\n\n LWLRAP - epoch: 5 - score: 0.0677 \n\nEpoch 6/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.3880 - acc: 0.4633 - val_loss: 16.0797 - val_acc: 0.0262\n\n LWLRAP - epoch: 6 - score: 0.0895 \n\nEpoch 7/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.1311 - acc: 0.5010 - val_loss: 14.8452 - val_acc: 0.0161\n\n LWLRAP - epoch: 7 - score: 0.0934 \n\nEpoch 8/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.8815 - acc: 0.5739 - val_loss: 18.2731 - val_acc: 0.0070\n\n LWLRAP - epoch: 8 - score: 0.0742 \n\nEpoch 9/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.6605 - acc: 0.6119 - val_loss: 13.5265 - val_acc: 0.0322\n\n LWLRAP - epoch: 9 - score: 0.1089 \n\nEpoch 10/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.1695 - acc: 0.7641 - val_loss: 10.5381 - val_acc: 0.0563\n\n LWLRAP - epoch: 10 - score: 0.1637 \n\nEpoch 11/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.0128 - acc: 0.8038 - val_loss: 10.4230 - val_acc: 0.0845\n\n LWLRAP - epoch: 11 - score: 0.2033 \n\nEpoch 12/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.9150 - acc: 0.8360 - val_loss: 7.1067 - val_acc: 0.1449\n\n LWLRAP - epoch: 12 - score: 0.2796 \n\nEpoch 13/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.8391 - acc: 0.8566 - val_loss: 9.5867 - val_acc: 0.0815\n\n LWLRAP - epoch: 13 - score: 0.2170 \n\nEpoch 14/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7831 - acc: 0.8619 - val_loss: 7.9145 - val_acc: 0.1388\n\n LWLRAP - epoch: 14 - score: 0.2717 \n\nEpoch 15/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7224 - acc: 0.8773 - val_loss: 5.5979 - val_acc: 0.1801\n\n LWLRAP - epoch: 15 - score: 0.3464 \n\nEpoch 16/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6655 - acc: 0.8913 - val_loss: 6.1202 - val_acc: 0.1600\n\n LWLRAP - epoch: 16 - score: 0.3113 \n\nEpoch 17/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6270 - acc: 0.8913 - val_loss: 8.3577 - val_acc: 0.1036\n\n LWLRAP - epoch: 17 - score: 0.2305 \n\nEpoch 18/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.5771 - acc: 0.9049 - val_loss: 6.9501 - val_acc: 0.1288\n\n LWLRAP - epoch: 18 - score: 0.2821 \n\nEpoch 19/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.5414 - acc: 0.9100 - val_loss: 7.8700 - val_acc: 0.1328\n\n LWLRAP - epoch: 19 - score: 0.2768 \n\nEpoch 20/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.5019 - acc: 0.9100 - val_loss: 5.8246 - val_acc: 0.1640\n\n LWLRAP - epoch: 20 - score: 0.2997 \n\nEpoch 21/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4651 - acc: 0.9190 - val_loss: 5.6788 - val_acc: 0.1942\n\n LWLRAP - epoch: 21 - score: 0.3388 \n\nEpoch 22/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4507 - acc: 0.9266 - val_loss: 4.3285 - val_acc: 0.2495\n\n LWLRAP - epoch: 22 - score: 0.4350 \n\nEpoch 23/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4411 - acc: 0.9220 - val_loss: 3.7830 - val_acc: 0.3008\n\n LWLRAP - epoch: 23 - score: 0.4836 \n\nEpoch 24/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4336 - acc: 0.9173 - val_loss: 3.6060 - val_acc: 0.3008\n\n LWLRAP - epoch: 24 - score: 0.4874 \n\nEpoch 25/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4267 - acc: 0.9200 - val_loss: 2.9122 - val_acc: 0.3773\n\n LWLRAP - epoch: 25 - score: 0.5599 \n\nEpoch 26/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4178 - acc: 0.9251 - val_loss: 6.7097 - val_acc: 0.2133\n\n LWLRAP - epoch: 26 - score: 0.3531 \n\nEpoch 27/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4114 - acc: 0.9243 - val_loss: 2.9512 - val_acc: 0.3602\n\n LWLRAP - epoch: 27 - score: 0.5395 \n\nEpoch 28/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4127 - acc: 0.9318 - val_loss: 3.7340 - val_acc: 0.3300\n\n LWLRAP - epoch: 28 - score: 0.5023 \n\nEpoch 29/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3990 - acc: 0.9213 - val_loss: 6.5019 - val_acc: 0.1982\n\n LWLRAP - epoch: 29 - score: 0.3555 \n\nEpoch 30/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4036 - acc: 0.9253 - val_loss: 3.7654 - val_acc: 0.2988\n\n LWLRAP - epoch: 30 - score: 0.4634 \n\nEpoch 31/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3858 - acc: 0.9238 - val_loss: 2.8586 - val_acc: 0.3934\n\n LWLRAP - epoch: 31 - score: 0.5699 \n\nEpoch 32/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3860 - acc: 0.9258 - val_loss: 2.7038 - val_acc: 0.4125\n\n LWLRAP - epoch: 32 - score: 0.5843 \n\nEpoch 33/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3909 - acc: 0.9208 - val_loss: 2.6885 - val_acc: 0.4225\n\n LWLRAP - epoch: 33 - score: 0.5933 \n\nEpoch 34/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3885 - acc: 0.9251 - val_loss: 2.8296 - val_acc: 0.3934\n\n LWLRAP - epoch: 34 - score: 0.5737 \n\nEpoch 35/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3831 - acc: 0.9303 - val_loss: 2.7887 - val_acc: 0.3984\n\n LWLRAP - epoch: 35 - score: 0.5719 \n\nloading best weights from current fold\nlwlrap fold: 0.5933\n(3976, 48, 256, 1) (3976, 80)\n(994, 48, 256, 1) (994, 80)\ntraining model...\nTrain on 3976 samples, validate on 994 samples\nEpoch 1/35\n3976/3976 [==============================] - 14s 4ms/step - loss: 4.4866 - acc: 0.0973 - val_loss: 18.4801 - val_acc: 0.0111\n\n LWLRAP - epoch: 1 - score: 0.0685 \n\nEpoch 2/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 3.6365 - acc: 0.2130 - val_loss: 17.1820 - val_acc: 0.0151\n\n LWLRAP - epoch: 2 - score: 0.0778 \n\nEpoch 3/35\n","name":"stdout"},{"output_type":"stream","text":"3976/3976 [==============================] - 7s 2ms/step - loss: 3.2071 - acc: 0.2995 - val_loss: 17.6085 - val_acc: 0.0201\n\n LWLRAP - epoch: 3 - score: 0.0829 \n\nEpoch 4/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.8813 - acc: 0.3481 - val_loss: 15.1711 - val_acc: 0.0402\n\n LWLRAP - epoch: 4 - score: 0.0959 \n\nEpoch 5/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.6332 - acc: 0.4054 - val_loss: 13.9732 - val_acc: 0.0392\n\n LWLRAP - epoch: 5 - score: 0.1048 \n\nEpoch 6/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.3725 - acc: 0.4532 - val_loss: 11.9056 - val_acc: 0.0252\n\n LWLRAP - epoch: 6 - score: 0.0970 \n\nEpoch 7/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 2.1659 - acc: 0.4927 - val_loss: 16.2132 - val_acc: 0.0201\n\n LWLRAP - epoch: 7 - score: 0.0913 \n\nEpoch 8/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.9057 - acc: 0.5606 - val_loss: 14.1725 - val_acc: 0.0221\n\n LWLRAP - epoch: 8 - score: 0.1060 \n\nEpoch 9/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.6500 - acc: 0.6124 - val_loss: 13.9326 - val_acc: 0.0362\n\n LWLRAP - epoch: 9 - score: 0.1010 \n\nEpoch 10/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.4104 - acc: 0.6733 - val_loss: 11.2530 - val_acc: 0.0322\n\n LWLRAP - epoch: 10 - score: 0.1160 \n\nEpoch 11/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 1.1804 - acc: 0.7437 - val_loss: 14.4463 - val_acc: 0.0453\n\n LWLRAP - epoch: 11 - score: 0.1092 \n\nEpoch 12/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.9808 - acc: 0.7885 - val_loss: 12.5131 - val_acc: 0.0312\n\n LWLRAP - epoch: 12 - score: 0.1297 \n\nEpoch 13/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.8137 - acc: 0.8325 - val_loss: 14.1633 - val_acc: 0.0433\n\n LWLRAP - epoch: 13 - score: 0.1182 \n\nEpoch 14/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.7384 - acc: 0.8554 - val_loss: 15.5256 - val_acc: 0.0292\n\n LWLRAP - epoch: 14 - score: 0.1044 \n\nEpoch 15/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.6307 - acc: 0.8815 - val_loss: 14.2363 - val_acc: 0.0272\n\n LWLRAP - epoch: 15 - score: 0.1078 \n\nEpoch 16/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4836 - acc: 0.9127 - val_loss: 12.8985 - val_acc: 0.0915\n\n LWLRAP - epoch: 16 - score: 0.1795 \n\nEpoch 17/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.4027 - acc: 0.9173 - val_loss: 8.7482 - val_acc: 0.1408\n\n LWLRAP - epoch: 17 - score: 0.2729 \n\nEpoch 18/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3750 - acc: 0.9178 - val_loss: 9.4290 - val_acc: 0.1429\n\n LWLRAP - epoch: 18 - score: 0.2547 \n\nEpoch 19/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3590 - acc: 0.9268 - val_loss: 11.5904 - val_acc: 0.1177\n\n LWLRAP - epoch: 19 - score: 0.2270 \n\nEpoch 20/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3386 - acc: 0.9271 - val_loss: 4.8226 - val_acc: 0.2596\n\n LWLRAP - epoch: 20 - score: 0.4321 \n\nEpoch 21/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3398 - acc: 0.9225 - val_loss: 8.6329 - val_acc: 0.1650\n\n LWLRAP - epoch: 21 - score: 0.3088 \n\nEpoch 22/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3415 - acc: 0.9258 - val_loss: 10.0746 - val_acc: 0.0966\n\n LWLRAP - epoch: 22 - score: 0.2494 \n\nEpoch 23/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3345 - acc: 0.9245 - val_loss: 10.9956 - val_acc: 0.0946\n\n LWLRAP - epoch: 23 - score: 0.2333 \n\nEpoch 24/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3339 - acc: 0.9240 - val_loss: 14.1146 - val_acc: 0.0573\n\n LWLRAP - epoch: 24 - score: 0.1499 \n\nEpoch 25/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3272 - acc: 0.9251 - val_loss: 12.7259 - val_acc: 0.0895\n\n LWLRAP - epoch: 25 - score: 0.1853 \n\nEpoch 26/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3128 - acc: 0.9238 - val_loss: 3.5480 - val_acc: 0.3592\n\n LWLRAP - epoch: 26 - score: 0.5307 \n\nEpoch 27/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.3033 - acc: 0.9238 - val_loss: 4.8667 - val_acc: 0.2817\n\n LWLRAP - epoch: 27 - score: 0.4501 \n\nEpoch 28/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2977 - acc: 0.9291 - val_loss: 4.4277 - val_acc: 0.2757\n\n LWLRAP - epoch: 28 - score: 0.4388 \n\nEpoch 29/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2968 - acc: 0.9225 - val_loss: 3.3773 - val_acc: 0.3853\n\n LWLRAP - epoch: 29 - score: 0.5406 \n\nEpoch 30/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2933 - acc: 0.9200 - val_loss: 6.3341 - val_acc: 0.1932\n\n LWLRAP - epoch: 30 - score: 0.3505 \n\nEpoch 31/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2938 - acc: 0.9266 - val_loss: 3.7254 - val_acc: 0.3471\n\n LWLRAP - epoch: 31 - score: 0.5058 \n\nEpoch 32/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2931 - acc: 0.9278 - val_loss: 7.0667 - val_acc: 0.1932\n\n LWLRAP - epoch: 32 - score: 0.3592 \n\nEpoch 33/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2924 - acc: 0.9193 - val_loss: 7.9965 - val_acc: 0.1861\n\n LWLRAP - epoch: 33 - score: 0.3344 \n\nEpoch 34/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2902 - acc: 0.9291 - val_loss: 4.4428 - val_acc: 0.2897\n\n LWLRAP - epoch: 34 - score: 0.4543 \n\nEpoch 35/35\n3976/3976 [==============================] - 7s 2ms/step - loss: 0.2862 - acc: 0.9276 - val_loss: 2.9665 - val_acc: 0.4215\n\n LWLRAP - epoch: 35 - score: 0.5824 \n\nloading best weights from current fold\nlwlrap fold: 0.5824\nOOF LWLRAP: 0.5583\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"oof_test_mean = oof_test.mean(axis=-1)\nprint(oof_test_mean.shape)\n\nsort_idx = np.argsort(classes).astype(int)\noof_test_mean_sorted = oof_test_mean[:, sort_idx]\nsample_submission.iloc[:, 1:] = oof_test_mean_sorted\nif SAVE_SUBMISSION:\n # sample_submission.to_csv('{}_lwl_{:.4f}.csv'.format(RUN_NAME, oof_lwl, index=False))\n sample_submission.to_csv('submission.csv', index=False)\n print(sample_submission.shape)\n \nsample_submission.head()","execution_count":14,"outputs":[{"output_type":"stream","text":"(1120, 80)\n(1120, 81)\n","name":"stdout"},{"output_type":"execute_result","execution_count":14,"data":{"text/plain":" fname Accelerating_and_revving_and_vroom Accordion \\\n0 000ccb97.wav 7.630934e-06 1.126024e-06 \n1 0012633b.wav 7.945251e-03 6.793572e-07 \n2 001ed5f1.wav 4.914838e-04 5.959930e-07 \n3 00294be0.wav 3.017973e-05 4.028038e-07 \n4 003fde7a.wav 4.506943e-11 8.921707e-08 \n\n Acoustic_guitar Applause Bark Bass_drum Bass_guitar \\\n0 9.557936e-06 7.191704e-06 5.745985e-06 2.253374e-03 6.174082e-06 \n1 2.767486e-07 4.266541e-05 3.476241e-04 5.217420e-06 1.051208e-06 \n2 4.253546e-06 2.102950e-05 1.089857e-04 5.557509e-02 3.716875e-05 \n3 6.008819e-06 1.118851e-05 2.601729e-03 3.143482e-06 7.871578e-06 \n4 1.036718e-06 7.856484e-08 1.572457e-07 1.505096e-08 9.948780e-09 \n\n Bathtub_(filling_or_washing) Bicycle_bell Burping_and_eructation \\\n0 2.355557e-02 3.019212e-03 0.001116 \n1 1.277254e-02 2.783864e-07 0.000156 \n2 2.423663e-04 1.361906e-05 0.000168 \n3 2.782175e-02 2.041693e-06 0.000217 \n4 2.414148e-09 8.056401e-01 0.000003 \n\n Bus Buzz Car_passing_by Cheering \\\n0 1.127185e-04 4.156595e-04 1.125294e-07 6.142814e-06 \n1 9.149866e-05 8.942229e-03 1.185120e-04 4.317892e-05 \n2 1.586857e-03 4.781486e-06 2.390796e-05 2.770816e-05 \n3 2.748477e-04 3.740818e-03 1.197445e-06 6.105997e-06 \n4 3.039046e-07 7.447269e-10 1.652458e-12 5.385623e-08 \n\n Chewing_and_mastication Child_speech_and_kid_speaking Chink_and_clink \\\n0 7.523261e-03 7.807558e-06 2.217286e-03 \n1 2.516754e-02 1.578456e-04 3.880692e-07 \n2 6.731230e-03 1.194379e-05 1.097298e-04 \n3 2.260438e-02 3.871577e-02 2.511732e-05 \n4 1.137780e-07 1.309705e-09 1.604554e-01 \n\n Chirp_and_tweet Church_bell Clapping Computer_keyboard Crackle \\\n0 0.019559 0.000002 0.000231 0.000821 2.206258e-03 \n1 0.000085 0.000016 0.000015 0.000534 6.352894e-02 \n2 0.000047 0.000064 0.001740 0.007760 1.945635e-04 \n3 0.000671 0.000168 0.000002 0.002720 8.103669e-05 \n4 0.000052 0.000009 0.000016 0.000004 5.334810e-10 \n\n Cricket Crowd Cupboard_open_or_close Cutlery_and_silverware \\\n0 0.116611 7.487527e-06 5.106073e-05 0.054577 \n1 0.000111 2.674633e-04 3.477763e-04 0.000014 \n2 0.000021 8.320537e-05 7.620898e-02 0.001282 \n3 0.001542 2.616761e-05 3.992716e-05 0.000067 \n4 0.000006 4.348518e-08 2.881525e-08 0.002008 \n\n Dishes_and_pots_and_pans Drawer_open_or_close Drip Electric_guitar \\\n0 0.003496 1.380159e-04 0.007563 8.464166e-06 \n1 0.000004 8.569744e-02 0.000048 5.711990e-07 \n2 0.000364 3.849829e-02 0.000050 4.099506e-04 \n3 0.000026 4.051256e-03 0.000066 2.201624e-06 \n4 0.027753 8.389362e-11 0.000029 6.607043e-07 \n\n Fart Female_singing Female_speech_and_woman_speaking \\\n0 3.790464e-04 1.796575e-05 7.161554e-04 \n1 6.306472e-04 1.985967e-05 7.372833e-04 \n2 1.384116e-02 1.452739e-07 5.066368e-04 \n3 3.012429e-04 5.623194e-04 1.727634e-02 \n4 5.820608e-09 4.084433e-08 4.364073e-08 \n\n Fill_(with_liquid) Finger_snapping Frying_(food) Gasp \\\n0 1.204408e-03 2.411856e-02 0.010561 1.141314e-03 \n1 1.520966e-03 3.733742e-08 0.000255 8.873779e-05 \n2 3.991843e-05 1.983964e-04 0.000032 8.880346e-06 \n3 5.731898e-04 1.193709e-07 0.000065 2.994187e-05 \n4 2.580490e-07 2.657285e-06 0.000001 4.677611e-08 \n\n Glockenspiel Gong Gurgling Harmonica Hi-hat \\\n0 3.995070e-06 6.569998e-04 8.718882e-04 8.443955e-05 8.851263e-03 \n1 1.981254e-12 1.304141e-07 3.926818e-03 2.210444e-06 1.873633e-06 \n2 1.217020e-06 5.853775e-06 7.599831e-05 8.467414e-07 1.173972e-03 \n3 3.593231e-09 2.430043e-06 2.005642e-02 2.965128e-04 3.607208e-06 \n4 8.467334e-04 1.364170e-06 7.464735e-08 5.795021e-04 8.360082e-07 \n\n Hiss Keys_jangling Knock Male_singing \\\n0 3.842605e-02 0.292134 1.648564e-05 3.101176e-06 \n1 3.355232e-03 0.000029 5.773754e-05 2.419432e-04 \n2 4.086373e-04 0.000173 9.300863e-02 2.091942e-06 \n3 3.375200e-04 0.000116 4.762887e-04 5.201666e-05 \n4 4.462873e-07 0.000136 1.325708e-08 1.742664e-11 \n\n Male_speech_and_man_speaking Marimba_and_xylophone Mechanical_fan \\\n0 7.125412e-04 5.951346e-06 3.640461e-05 \n1 8.899620e-04 2.745302e-09 2.020297e-04 \n2 2.537540e-03 1.043665e-03 8.105561e-05 \n3 5.360056e-04 2.270470e-06 3.666047e-06 \n4 7.846682e-09 1.296280e-03 5.261956e-07 \n\n Meow Microwave_oven Motorcycle Printer Purr \\\n0 4.123669e-04 0.000210 9.366042e-07 0.000145 1.422188e-04 \n1 1.672870e-03 0.000725 2.838649e-02 0.002354 1.803961e-01 \n2 4.742294e-05 0.026263 1.132772e-03 0.006985 5.558643e-05 \n3 5.781172e-03 0.000033 2.963027e-04 0.002549 4.489451e-01 \n4 7.912064e-09 0.000184 2.701101e-11 0.000014 1.005599e-11 \n\n Race_car_and_auto_racing Raindrop Run Scissors Screaming \\\n0 5.197266e-08 0.011268 9.356048e-06 0.128158 1.711372e-04 \n1 1.049741e-04 0.000003 3.934470e-02 0.001237 3.008439e-05 \n2 9.856125e-05 0.000033 1.259541e-02 0.004615 5.979117e-05 \n3 3.897196e-06 0.000048 6.293690e-03 0.000108 1.262180e-05 \n4 1.347549e-11 0.000034 2.243251e-11 0.000003 1.544679e-07 \n\n Shatter Sigh Sink_(filling_or_washing) Skateboard \\\n0 0.140231 3.475408e-04 0.007002 1.862208e-06 \n1 0.000019 2.220316e-04 0.000832 1.536778e-03 \n2 0.000834 3.193068e-06 0.000080 7.580077e-04 \n3 0.000009 2.911150e-05 0.001492 2.008946e-04 \n4 0.000191 3.297388e-09 0.000004 3.136905e-10 \n\n Slam Sneeze Squeak Stream Strum Tap \\\n0 4.675094e-04 1.831639e-03 0.002119 3.051140e-04 3.147200e-06 0.000089 \n1 1.044683e-03 1.713921e-04 0.005715 1.108031e-04 9.399754e-08 0.000007 \n2 3.379928e-01 2.925850e-03 0.133867 3.261873e-06 4.635730e-06 0.014612 \n3 7.094769e-05 9.866642e-05 0.000492 3.537733e-05 3.252812e-06 0.000263 \n4 2.702228e-07 3.415866e-07 0.000007 2.012157e-10 3.186759e-07 0.000035 \n\n Tick-tock Toilet_flush Traffic_noise_and_roadway_noise \\\n0 0.001445 2.000678e-04 1.137943e-06 \n1 0.000114 1.511131e-02 1.009377e-04 \n2 0.049321 3.780541e-05 6.191605e-05 \n3 0.002591 7.771356e-04 3.500645e-06 \n4 0.000681 4.628647e-10 1.206940e-08 \n\n Trickle_and_dribble Walk_and_footsteps Water_tap_and_faucet \\\n0 1.875239e-03 3.074506e-05 4.461727e-03 \n1 1.844445e-05 1.155871e-01 1.066032e-03 \n2 2.500981e-06 9.177884e-02 6.771300e-06 \n3 5.882365e-04 1.856796e-02 8.137919e-04 \n4 7.399385e-07 1.406988e-10 4.495905e-07 \n\n Waves_and_surf Whispering Writing Yell Zipper_(clothing) \n0 1.493053e-05 3.263234e-03 6.970076e-02 1.609087e-04 4.525526e-04 \n1 2.837723e-02 1.594179e-04 9.546124e-03 2.999285e-04 3.472851e-01 \n2 2.638801e-04 4.096762e-04 9.777345e-04 3.149970e-04 8.839027e-03 \n3 3.540992e-05 9.361802e-04 3.208230e-01 2.883691e-03 3.899243e-02 \n4 8.433536e-14 8.252657e-08 1.030768e-09 8.219400e-09 6.913077e-12 ","text/html":"
\n\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n
fnameAccelerating_and_revving_and_vroomAccordionAcoustic_guitarApplauseBarkBass_drumBass_guitarBathtub_(filling_or_washing)Bicycle_bellBurping_and_eructationBusBuzzCar_passing_byCheeringChewing_and_masticationChild_speech_and_kid_speakingChink_and_clinkChirp_and_tweetChurch_bellClappingComputer_keyboardCrackleCricketCrowdCupboard_open_or_closeCutlery_and_silverwareDishes_and_pots_and_pansDrawer_open_or_closeDripElectric_guitarFartFemale_singingFemale_speech_and_woman_speakingFill_(with_liquid)Finger_snappingFrying_(food)GaspGlockenspielGongGurglingHarmonicaHi-hatHissKeys_janglingKnockMale_singingMale_speech_and_man_speakingMarimba_and_xylophoneMechanical_fanMeowMicrowave_ovenMotorcyclePrinterPurrRace_car_and_auto_racingRaindropRunScissorsScreamingShatterSighSink_(filling_or_washing)SkateboardSlamSneezeSqueakStreamStrumTapTick-tockToilet_flushTraffic_noise_and_roadway_noiseTrickle_and_dribbleWalk_and_footstepsWater_tap_and_faucetWaves_and_surfWhisperingWritingYellZipper_(clothing)
0000ccb97.wav7.630934e-061.126024e-069.557936e-067.191704e-065.745985e-062.253374e-036.174082e-062.355557e-023.019212e-030.0011161.127185e-044.156595e-041.125294e-076.142814e-067.523261e-037.807558e-062.217286e-030.0195590.0000020.0002310.0008212.206258e-030.1166117.487527e-065.106073e-050.0545770.0034961.380159e-040.0075638.464166e-063.790464e-041.796575e-057.161554e-041.204408e-032.411856e-020.0105611.141314e-033.995070e-066.569998e-048.718882e-048.443955e-058.851263e-033.842605e-020.2921341.648564e-053.101176e-067.125412e-045.951346e-063.640461e-054.123669e-040.0002109.366042e-070.0001451.422188e-045.197266e-080.0112689.356048e-060.1281581.711372e-040.1402313.475408e-040.0070021.862208e-064.675094e-041.831639e-030.0021193.051140e-043.147200e-060.0000890.0014452.000678e-041.137943e-061.875239e-033.074506e-054.461727e-031.493053e-053.263234e-036.970076e-021.609087e-044.525526e-04
10012633b.wav7.945251e-036.793572e-072.767486e-074.266541e-053.476241e-045.217420e-061.051208e-061.277254e-022.783864e-070.0001569.149866e-058.942229e-031.185120e-044.317892e-052.516754e-021.578456e-043.880692e-070.0000850.0000160.0000150.0005346.352894e-020.0001112.674633e-043.477763e-040.0000140.0000048.569744e-020.0000485.711990e-076.306472e-041.985967e-057.372833e-041.520966e-033.733742e-080.0002558.873779e-051.981254e-121.304141e-073.926818e-032.210444e-061.873633e-063.355232e-030.0000295.773754e-052.419432e-048.899620e-042.745302e-092.020297e-041.672870e-030.0007252.838649e-020.0023541.803961e-011.049741e-040.0000033.934470e-020.0012373.008439e-050.0000192.220316e-040.0008321.536778e-031.044683e-031.713921e-040.0057151.108031e-049.399754e-080.0000070.0001141.511131e-021.009377e-041.844445e-051.155871e-011.066032e-032.837723e-021.594179e-049.546124e-032.999285e-043.472851e-01
2001ed5f1.wav4.914838e-045.959930e-074.253546e-062.102950e-051.089857e-045.557509e-023.716875e-052.423663e-041.361906e-050.0001681.586857e-034.781486e-062.390796e-052.770816e-056.731230e-031.194379e-051.097298e-040.0000470.0000640.0017400.0077601.945635e-040.0000218.320537e-057.620898e-020.0012820.0003643.849829e-020.0000504.099506e-041.384116e-021.452739e-075.066368e-043.991843e-051.983964e-040.0000328.880346e-061.217020e-065.853775e-067.599831e-058.467414e-071.173972e-034.086373e-040.0001739.300863e-022.091942e-062.537540e-031.043665e-038.105561e-054.742294e-050.0262631.132772e-030.0069855.558643e-059.856125e-050.0000331.259541e-020.0046155.979117e-050.0008343.193068e-060.0000807.580077e-043.379928e-012.925850e-030.1338673.261873e-064.635730e-060.0146120.0493213.780541e-056.191605e-052.500981e-069.177884e-026.771300e-062.638801e-044.096762e-049.777345e-043.149970e-048.839027e-03
300294be0.wav3.017973e-054.028038e-076.008819e-061.118851e-052.601729e-033.143482e-067.871578e-062.782175e-022.041693e-060.0002172.748477e-043.740818e-031.197445e-066.105997e-062.260438e-023.871577e-022.511732e-050.0006710.0001680.0000020.0027208.103669e-050.0015422.616761e-053.992716e-050.0000670.0000264.051256e-030.0000662.201624e-063.012429e-045.623194e-041.727634e-025.731898e-041.193709e-070.0000652.994187e-053.593231e-092.430043e-062.005642e-022.965128e-043.607208e-063.375200e-040.0001164.762887e-045.201666e-055.360056e-042.270470e-063.666047e-065.781172e-030.0000332.963027e-040.0025494.489451e-013.897196e-060.0000486.293690e-030.0001081.262180e-050.0000092.911150e-050.0014922.008946e-047.094769e-059.866642e-050.0004923.537733e-053.252812e-060.0002630.0025917.771356e-043.500645e-065.882365e-041.856796e-028.137919e-043.540992e-059.361802e-043.208230e-012.883691e-033.899243e-02
4003fde7a.wav4.506943e-118.921707e-081.036718e-067.856484e-081.572457e-071.505096e-089.948780e-092.414148e-098.056401e-010.0000033.039046e-077.447269e-101.652458e-125.385623e-081.137780e-071.309705e-091.604554e-010.0000520.0000090.0000160.0000045.334810e-100.0000064.348518e-082.881525e-080.0020080.0277538.389362e-110.0000296.607043e-075.820608e-094.084433e-084.364073e-082.580490e-072.657285e-060.0000014.677611e-088.467334e-041.364170e-067.464735e-085.795021e-048.360082e-074.462873e-070.0001361.325708e-081.742664e-117.846682e-091.296280e-035.261956e-077.912064e-090.0001842.701101e-110.0000141.005599e-111.347549e-110.0000342.243251e-110.0000031.544679e-070.0001913.297388e-090.0000043.136905e-102.702228e-073.415866e-070.0000072.012157e-103.186759e-070.0000350.0006814.628647e-101.206940e-087.399385e-071.406988e-104.495905e-078.433536e-148.252657e-081.030768e-098.219400e-096.913077e-12
\n
"},"metadata":{}}]},{"metadata":{"trusted":false},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":false},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.6.7"}},"nbformat":4,"nbformat_minor":1}