{
  "id": 107588,
  "title": "Surprisingly low public score",
  "url": "/competitions/aptos2019-blindness-detection/discussion/107588",
  "author_name": "",
  "post_date": "2019-09-05T08:58:15.274457900Z",
  "votes": null,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Dear Kagglers,\nI've got 98% accuracy with my kernel on the public training dataset, so I'd expect pretty high score after submission, but my score was only 0.011.  I understand that probably score would't be so much high as the metrics is different from 'accuracy', but so low score is really surprising for me.</p>\n\n<p>I'm a novice on Kaggle competitions, so do anybody please give me a clue what was wrong?</p>\n\n<p>Thank you in advance, and best regards,\nBoris </p>",
  "messages": [
    {
      "id": "618507",
      "postDate": "09/05/2019 08:58:15",
      "content": "<p>Dear Kagglers,\nI've got 98% accuracy with my kernel on the public training dataset, so I'd expect pretty high score after submission, but my score was only 0.011.  I understand that probably score would't be so much high as the metrics is different from 'accuracy', but so low score is really surprising for me.</p>\n\n<p>I'm a novice on Kaggle competitions, so do anybody please give me a clue what was wrong?</p>\n\n<p>Thank you in advance, and best regards,\nBoris </p>",
      "rawMarkdown": "Dear Kagglers,\nI've got 98% accuracy with my kernel on the public training dataset, so I'd expect pretty high score after submission, but my score was only 0.011.  I understand that probably score would't be so much high as the metrics is different from 'accuracy', but so low score is really surprising for me.\n\nI'm a novice on Kaggle competitions, so do anybody please give me a clue what was wrong?\n\nThank you in advance, and best regards,\nBoris",
      "votes": null
    },
    {
      "id": "618514",
      "postDate": "09/05/2019 09:00:27",
      "content": "<p>if you share your code someone might be able to help, like this is very very hard...</p>",
      "rawMarkdown": "if you share your code someone might be able to help, like this is very very hard...",
      "votes": null
    },
    {
      "id": "618528",
      "postDate": "09/05/2019 09:12:31",
      "content": "<p>Sorry, I forgot to attach my kernel submission log (see below):</p>\n\n<p>Using TensorFlow backend.\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:516: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:517: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:519: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:520: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\nEpoch 1/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 1.3594 - acc: 0.6663\nEpoch 2/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.7146 - acc: 0.7400\nEpoch 3/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.6569 - acc: 0.7597\nEpoch 4/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.6148 - acc: 0.7744\nEpoch 5/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5784 - acc: 0.7873\nEpoch 6/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5408 - acc: 0.8031\nEpoch 7/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.5016 - acc: 0.8151\nEpoch 8/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4707 - acc: 0.8326\nEpoch 9/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4445 - acc: 0.8419\nEpoch 10/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4111 - acc: 0.8495\nEpoch 11/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3833 - acc: 0.8624\nEpoch 12/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3565 - acc: 0.8798\nEpoch 13/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3385 - acc: 0.8848\nEpoch 14/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3067 - acc: 0.8979\nEpoch 15/100\n3662/3662 [==============================] - 42s 11ms/step - loss: 0.2855 - acc: 0.9058\nEpoch 16/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2673 - acc: 0.9181\nEpoch 17/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2513 - acc: 0.9194\nEpoch 18/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2276 - acc: 0.9285\nEpoch 19/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2148 - acc: 0.9347\nEpoch 20/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1958 - acc: 0.9429\nEpoch 21/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1880 - acc: 0.9459\nEpoch 22/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1682 - acc: 0.9508\nEpoch 23/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1558 - acc: 0.9582\nEpoch 24/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1457 - acc: 0.9604\nEpoch 25/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1393 - acc: 0.9642\nEpoch 26/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1304 - acc: 0.9645\nEpoch 27/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1192 - acc: 0.9727\nEpoch 28/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1161 - acc: 0.9683\nEpoch 29/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1029 - acc: 0.9700\nEpoch 30/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0987 - acc: 0.9749\nEpoch 31/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0974 - acc: 0.9746\nEpoch 32/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0892 - acc: 0.9782\nEpoch 33/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0857 - acc: 0.9801\nEpoch 34/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0801 - acc: 0.9771\nEpoch 35/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0762 - acc: 0.9795\nEpoch 36/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0720 - acc: 0.9817\nEpoch 37/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0680 - acc: 0.9831\nEpoch 38/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0666 - acc: 0.9817\nEpoch 39/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0606 - acc: 0.9842\nEpoch 40/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0676 - acc: 0.9825\nEpoch 41/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0629 - acc: 0.9809\nEpoch 42/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0648 - acc: 0.9820\nEpoch 43/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0646 - acc: 0.9825\nEpoch 44/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0564 - acc: 0.9836\nEpoch 45/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0684 - acc: 0.9795\nEpoch 46/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0650 - acc: 0.9801\nEpoch 47/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0543 - acc: 0.9842\nEpoch 48/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0547 - acc: 0.9853\nEpoch 49/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0550 - acc: 0.9831\nEpoch 50/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0497 - acc: 0.9872\nEpoch 51/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0551 - acc: 0.9842\nEpoch 52/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0528 - acc: 0.9872\nEpoch 53/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0467 - acc: 0.9855\nEpoch 54/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0464 - acc: 0.9869\nEpoch 55/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9850\nEpoch 56/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9855\nEpoch 57/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0440 - acc: 0.9855\nEpoch 58/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9836\nEpoch 59/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0481 - acc: 0.9844\nEpoch 60/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0472 - acc: 0.9842\nEpoch 61/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0474 - acc: 0.9855\nEpoch 62/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0473 - acc: 0.9847\nEpoch 63/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0469 - acc: 0.9850\nEpoch 64/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9842\nEpoch 65/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0445 - acc: 0.9855\nEpoch 66/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0500 - acc: 0.9844\nEpoch 67/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9850\nEpoch 68/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0421 - acc: 0.9861\nEpoch 69/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0439 - acc: 0.9855\nEpoch 70/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0423 - acc: 0.9861\nEpoch 71/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0451 - acc: 0.9828\nEpoch 72/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0479 - acc: 0.9847\nEpoch 73/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0457 - acc: 0.9858\nEpoch 74/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0437 - acc: 0.9850\nEpoch 75/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0449 - acc: 0.9839\nEpoch 76/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0444 - acc: 0.9853\nEpoch 77/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0458 - acc: 0.9844\nEpoch 78/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0433 - acc: 0.9874\nEpoch 79/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0480 - acc: 0.9839\nEpoch 80/100\n3300/3662 [==========================&gt;...] - ETA: 3s - loss: 0.0405 - acc: 0.9842</p>",
      "rawMarkdown": "Sorry, I forgot to attach my kernel submission log (see below):\n\nUsing TensorFlow backend.\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:516: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:517: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:519: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:520: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\nEpoch 1/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 1.3594 - acc: 0.6663\nEpoch 2/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.7146 - acc: 0.7400\nEpoch 3/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.6569 - acc: 0.7597\nEpoch 4/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.6148 - acc: 0.7744\nEpoch 5/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5784 - acc: 0.7873\nEpoch 6/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5408 - acc: 0.8031\nEpoch 7/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.5016 - acc: 0.8151\nEpoch 8/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4707 - acc: 0.8326\nEpoch 9/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4445 - acc: 0.8419\nEpoch 10/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4111 - acc: 0.8495\nEpoch 11/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3833 - acc: 0.8624\nEpoch 12/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3565 - acc: 0.8798\nEpoch 13/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3385 - acc: 0.8848\nEpoch 14/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3067 - acc: 0.8979\nEpoch 15/100\n3662/3662 [==============================] - 42s 11ms/step - loss: 0.2855 - acc: 0.9058\nEpoch 16/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2673 - acc: 0.9181\nEpoch 17/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2513 - acc: 0.9194\nEpoch 18/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2276 - acc: 0.9285\nEpoch 19/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2148 - acc: 0.9347\nEpoch 20/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1958 - acc: 0.9429\nEpoch 21/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1880 - acc: 0.9459\nEpoch 22/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1682 - acc: 0.9508\nEpoch 23/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1558 - acc: 0.9582\nEpoch 24/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1457 - acc: 0.9604\nEpoch 25/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1393 - acc: 0.9642\nEpoch 26/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1304 - acc: 0.9645\nEpoch 27/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1192 - acc: 0.9727\nEpoch 28/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1161 - acc: 0.9683\nEpoch 29/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1029 - acc: 0.9700\nEpoch 30/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0987 - acc: 0.9749\nEpoch 31/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0974 - acc: 0.9746\nEpoch 32/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0892 - acc: 0.9782\nEpoch 33/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0857 - acc: 0.9801\nEpoch 34/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0801 - acc: 0.9771\nEpoch 35/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0762 - acc: 0.9795\nEpoch 36/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0720 - acc: 0.9817\nEpoch 37/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0680 - acc: 0.9831\nEpoch 38/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0666 - acc: 0.9817\nEpoch 39/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0606 - acc: 0.9842\nEpoch 40/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0676 - acc: 0.9825\nEpoch 41/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0629 - acc: 0.9809\nEpoch 42/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0648 - acc: 0.9820\nEpoch 43/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0646 - acc: 0.9825\nEpoch 44/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0564 - acc: 0.9836\nEpoch 45/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0684 - acc: 0.9795\nEpoch 46/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0650 - acc: 0.9801\nEpoch 47/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0543 - acc: 0.9842\nEpoch 48/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0547 - acc: 0.9853\nEpoch 49/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0550 - acc: 0.9831\nEpoch 50/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0497 - acc: 0.9872\nEpoch 51/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0551 - acc: 0.9842\nEpoch 52/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0528 - acc: 0.9872\nEpoch 53/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0467 - acc: 0.9855\nEpoch 54/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0464 - acc: 0.9869\nEpoch 55/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9850\nEpoch 56/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9855\nEpoch 57/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0440 - acc: 0.9855\nEpoch 58/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9836\nEpoch 59/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0481 - acc: 0.9844\nEpoch 60/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0472 - acc: 0.9842\nEpoch 61/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0474 - acc: 0.9855\nEpoch 62/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0473 - acc: 0.9847\nEpoch 63/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0469 - acc: 0.9850\nEpoch 64/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9842\nEpoch 65/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0445 - acc: 0.9855\nEpoch 66/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0500 - acc: 0.9844\nEpoch 67/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9850\nEpoch 68/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0421 - acc: 0.9861\nEpoch 69/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0439 - acc: 0.9855\nEpoch 70/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0423 - acc: 0.9861\nEpoch 71/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0451 - acc: 0.9828\nEpoch 72/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0479 - acc: 0.9847\nEpoch 73/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0457 - acc: 0.9858\nEpoch 74/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0437 - acc: 0.9850\nEpoch 75/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0449 - acc: 0.9839\nEpoch 76/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0444 - acc: 0.9853\nEpoch 77/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0458 - acc: 0.9844\nEpoch 78/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0433 - acc: 0.9874\nEpoch 79/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0480 - acc: 0.9839\nEpoch 80/100\n3300/3662 [==========================&gt;...] - ETA: 3s - loss: 0.0405 - acc: 0.9842",
      "votes": null
    },
    {
      "id": "618741",
      "postDate": "09/05/2019 12:55:16",
      "content": "<p>Looks like you are facing overfit issues. You should split a validation fold to estimate the loss instead of only  train set.</p>",
      "rawMarkdown": "Looks like you are facing overfit issues. You should split a validation fold to estimate the loss instead of only  train set.",
      "votes": null
    },
    {
      "id": "619015",
      "postDate": "09/05/2019 18:39:49",
      "content": "<p>Thanks a lot, InfiniteWing!\nIndeed, I've seen a sign of overfitting when looked at nearly constant validation loss..</p>",
      "rawMarkdown": "Thanks a lot, InfiniteWing!\nIndeed, I've seen a sign of overfitting when looked at nearly constant validation loss..",
      "votes": null
    },
    {
      "id": "619279",
      "postDate": "09/06/2019 03:08:59",
      "content": "<p>Hi <a href=\"/polishch\">@polishch</a>, accuracy is not a good metric to save checkpoint for this competition, you should use Kappa score.</p>",
      "rawMarkdown": "Hi @polishch, accuracy is not a good metric to save checkpoint for this competition, you should use Kappa score.",
      "votes": null
    },
    {
      "id": "619292",
      "postDate": "09/06/2019 03:33:56",
      "content": "<p>Checked the how to submit successfully discussions, the id_code in test.csv is not correlated with sample_submission.csv. I think it's something wrong with your final csv rather than model.</p>",
      "rawMarkdown": "Checked the how to submit successfully discussions, the id_code in test.csv is not correlated with sample_submission.csv. I think it's something wrong with your final csv rather than model.",
      "votes": null
    },
    {
      "id": "619301",
      "postDate": "09/06/2019 03:46:20",
      "content": "<p>do not rely on accuracy score. If you treat the problem as regression, you have to implement MSE loss (loss function) and also monitor the required metrics (quadratic weighted kappa). there are many useful kernels published for the function of loss and metrics. </p>",
      "rawMarkdown": "do not rely on accuracy score. If you treat the problem as regression, you have to implement MSE loss (loss function) and also monitor the required metrics (quadratic weighted kappa). there are many useful kernels published for the function of loss and metrics.",
      "votes": null
    },
    {
      "id": "619606",
      "postDate": "09/06/2019 11:04:03",
      "content": "<p>Thank you jionie!</p>\n\n<p>Now I use sample_submission.csv, it seems my score increased because of that (the model is also a bit different, but I'm in doubt it can explain eight times increase in score..)</p>",
      "rawMarkdown": "Thank you jionie!\n\nNow I use sample_submission.csv, it seems my score increased because of that (the model is also a bit different, but I'm in doubt it can explain eight times increase in score..)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 618514,
      "author_name": "valanm",
      "author_url": "",
      "post_date": "09/05/2019 09:00:27",
      "content": "<p>if you share your code someone might be able to help, like this is very very hard...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 618528,
      "author_name": "polishch",
      "author_url": "",
      "post_date": "09/05/2019 09:12:31",
      "content": "<p>Sorry, I forgot to attach my kernel submission log (see below):</p>\n\n<p>Using TensorFlow backend.\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:516: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:517: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:519: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:520: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\nEpoch 1/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 1.3594 - acc: 0.6663\nEpoch 2/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.7146 - acc: 0.7400\nEpoch 3/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.6569 - acc: 0.7597\nEpoch 4/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.6148 - acc: 0.7744\nEpoch 5/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5784 - acc: 0.7873\nEpoch 6/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5408 - acc: 0.8031\nEpoch 7/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.5016 - acc: 0.8151\nEpoch 8/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4707 - acc: 0.8326\nEpoch 9/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4445 - acc: 0.8419\nEpoch 10/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4111 - acc: 0.8495\nEpoch 11/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3833 - acc: 0.8624\nEpoch 12/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3565 - acc: 0.8798\nEpoch 13/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3385 - acc: 0.8848\nEpoch 14/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3067 - acc: 0.8979\nEpoch 15/100\n3662/3662 [==============================] - 42s 11ms/step - loss: 0.2855 - acc: 0.9058\nEpoch 16/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2673 - acc: 0.9181\nEpoch 17/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2513 - acc: 0.9194\nEpoch 18/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2276 - acc: 0.9285\nEpoch 19/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2148 - acc: 0.9347\nEpoch 20/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1958 - acc: 0.9429\nEpoch 21/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1880 - acc: 0.9459\nEpoch 22/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1682 - acc: 0.9508\nEpoch 23/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1558 - acc: 0.9582\nEpoch 24/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1457 - acc: 0.9604\nEpoch 25/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1393 - acc: 0.9642\nEpoch 26/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1304 - acc: 0.9645\nEpoch 27/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1192 - acc: 0.9727\nEpoch 28/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1161 - acc: 0.9683\nEpoch 29/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1029 - acc: 0.9700\nEpoch 30/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0987 - acc: 0.9749\nEpoch 31/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0974 - acc: 0.9746\nEpoch 32/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0892 - acc: 0.9782\nEpoch 33/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0857 - acc: 0.9801\nEpoch 34/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0801 - acc: 0.9771\nEpoch 35/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0762 - acc: 0.9795\nEpoch 36/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0720 - acc: 0.9817\nEpoch 37/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0680 - acc: 0.9831\nEpoch 38/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0666 - acc: 0.9817\nEpoch 39/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0606 - acc: 0.9842\nEpoch 40/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0676 - acc: 0.9825\nEpoch 41/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0629 - acc: 0.9809\nEpoch 42/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0648 - acc: 0.9820\nEpoch 43/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0646 - acc: 0.9825\nEpoch 44/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0564 - acc: 0.9836\nEpoch 45/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0684 - acc: 0.9795\nEpoch 46/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0650 - acc: 0.9801\nEpoch 47/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0543 - acc: 0.9842\nEpoch 48/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0547 - acc: 0.9853\nEpoch 49/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0550 - acc: 0.9831\nEpoch 50/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0497 - acc: 0.9872\nEpoch 51/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0551 - acc: 0.9842\nEpoch 52/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0528 - acc: 0.9872\nEpoch 53/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0467 - acc: 0.9855\nEpoch 54/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0464 - acc: 0.9869\nEpoch 55/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9850\nEpoch 56/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9855\nEpoch 57/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0440 - acc: 0.9855\nEpoch 58/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9836\nEpoch 59/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0481 - acc: 0.9844\nEpoch 60/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0472 - acc: 0.9842\nEpoch 61/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0474 - acc: 0.9855\nEpoch 62/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0473 - acc: 0.9847\nEpoch 63/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0469 - acc: 0.9850\nEpoch 64/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9842\nEpoch 65/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0445 - acc: 0.9855\nEpoch 66/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0500 - acc: 0.9844\nEpoch 67/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9850\nEpoch 68/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0421 - acc: 0.9861\nEpoch 69/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0439 - acc: 0.9855\nEpoch 70/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0423 - acc: 0.9861\nEpoch 71/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0451 - acc: 0.9828\nEpoch 72/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0479 - acc: 0.9847\nEpoch 73/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0457 - acc: 0.9858\nEpoch 74/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0437 - acc: 0.9850\nEpoch 75/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0449 - acc: 0.9839\nEpoch 76/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0444 - acc: 0.9853\nEpoch 77/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0458 - acc: 0.9844\nEpoch 78/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0433 - acc: 0.9874\nEpoch 79/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0480 - acc: 0.9839\nEpoch 80/100\n3300/3662 [==========================&gt;...] - ETA: 3s - loss: 0.0405 - acc: 0.9842</p>",
      "votes": null,
      "replies": [
        {
          "id": 618741,
          "author_name": "infinitewing",
          "author_url": "",
          "post_date": "09/05/2019 12:55:16",
          "content": "<p>Looks like you are facing overfit issues. You should split a validation fold to estimate the loss instead of only  train set.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 619015,
          "author_name": "polishch",
          "author_url": "",
          "post_date": "09/05/2019 18:39:49",
          "content": "<p>Thanks a lot, InfiniteWing!\nIndeed, I've seen a sign of overfitting when looked at nearly constant validation loss..</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 619279,
      "author_name": "dattran2346",
      "author_url": "",
      "post_date": "09/06/2019 03:08:59",
      "content": "<p>Hi <a href=\"/polishch\">@polishch</a>, accuracy is not a good metric to save checkpoint for this competition, you should use Kappa score.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 619292,
      "author_name": "jionie",
      "author_url": "",
      "post_date": "09/06/2019 03:33:56",
      "content": "<p>Checked the how to submit successfully discussions, the id_code in test.csv is not correlated with sample_submission.csv. I think it's something wrong with your final csv rather than model.</p>",
      "votes": null,
      "replies": [
        {
          "id": 619606,
          "author_name": "polishch",
          "author_url": "",
          "post_date": "09/06/2019 11:04:03",
          "content": "<p>Thank you jionie!</p>\n\n<p>Now I use sample_submission.csv, it seems my score increased because of that (the model is also a bit different, but I'm in doubt it can explain eight times increase in score..)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 619301,
      "author_name": "projdev",
      "author_url": "",
      "post_date": "09/06/2019 03:46:20",
      "content": "<p>do not rely on accuracy score. If you treat the problem as regression, you have to implement MSE loss (loss function) and also monitor the required metrics (quadratic weighted kappa). there are many useful kernels published for the function of loss and metrics. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "618507": "Dear Kagglers,\nI've got 98% accuracy with my kernel on the public training dataset, so I'd expect pretty high score after submission, but my score was only 0.011.  I understand that probably score would't be so much high as the metrics is different from 'accuracy', but so low score is really surprising for me.\n\nI'm a novice on Kaggle competitions, so do anybody please give me a clue what was wrong?\n\nThank you in advance, and best regards,\nBoris",
    "618514": "if you share your code someone might be able to help, like this is very very hard...",
    "618528": "Sorry, I forgot to attach my kernel submission log (see below):\n\nUsing TensorFlow backend.\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:516: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:517: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:519: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:520: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\nEpoch 1/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 1.3594 - acc: 0.6663\nEpoch 2/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.7146 - acc: 0.7400\nEpoch 3/100\n3662/3662 [==============================] - 39s 11ms/step - loss: 0.6569 - acc: 0.7597\nEpoch 4/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.6148 - acc: 0.7744\nEpoch 5/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5784 - acc: 0.7873\nEpoch 6/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.5408 - acc: 0.8031\nEpoch 7/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.5016 - acc: 0.8151\nEpoch 8/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4707 - acc: 0.8326\nEpoch 9/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4445 - acc: 0.8419\nEpoch 10/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.4111 - acc: 0.8495\nEpoch 11/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3833 - acc: 0.8624\nEpoch 12/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3565 - acc: 0.8798\nEpoch 13/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3385 - acc: 0.8848\nEpoch 14/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.3067 - acc: 0.8979\nEpoch 15/100\n3662/3662 [==============================] - 42s 11ms/step - loss: 0.2855 - acc: 0.9058\nEpoch 16/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2673 - acc: 0.9181\nEpoch 17/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2513 - acc: 0.9194\nEpoch 18/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2276 - acc: 0.9285\nEpoch 19/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.2148 - acc: 0.9347\nEpoch 20/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1958 - acc: 0.9429\nEpoch 21/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1880 - acc: 0.9459\nEpoch 22/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1682 - acc: 0.9508\nEpoch 23/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.1558 - acc: 0.9582\nEpoch 24/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1457 - acc: 0.9604\nEpoch 25/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1393 - acc: 0.9642\nEpoch 26/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1304 - acc: 0.9645\nEpoch 27/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1192 - acc: 0.9727\nEpoch 28/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1161 - acc: 0.9683\nEpoch 29/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.1029 - acc: 0.9700\nEpoch 30/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0987 - acc: 0.9749\nEpoch 31/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0974 - acc: 0.9746\nEpoch 32/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0892 - acc: 0.9782\nEpoch 33/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0857 - acc: 0.9801\nEpoch 34/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0801 - acc: 0.9771\nEpoch 35/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0762 - acc: 0.9795\nEpoch 36/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0720 - acc: 0.9817\nEpoch 37/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0680 - acc: 0.9831\nEpoch 38/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0666 - acc: 0.9817\nEpoch 39/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0606 - acc: 0.9842\nEpoch 40/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0676 - acc: 0.9825\nEpoch 41/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0629 - acc: 0.9809\nEpoch 42/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0648 - acc: 0.9820\nEpoch 43/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0646 - acc: 0.9825\nEpoch 44/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0564 - acc: 0.9836\nEpoch 45/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0684 - acc: 0.9795\nEpoch 46/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0650 - acc: 0.9801\nEpoch 47/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0543 - acc: 0.9842\nEpoch 48/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0547 - acc: 0.9853\nEpoch 49/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0550 - acc: 0.9831\nEpoch 50/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0497 - acc: 0.9872\nEpoch 51/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0551 - acc: 0.9842\nEpoch 52/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0528 - acc: 0.9872\nEpoch 53/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0467 - acc: 0.9855\nEpoch 54/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0464 - acc: 0.9869\nEpoch 55/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9850\nEpoch 56/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9855\nEpoch 57/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0440 - acc: 0.9855\nEpoch 58/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9836\nEpoch 59/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0481 - acc: 0.9844\nEpoch 60/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0472 - acc: 0.9842\nEpoch 61/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0474 - acc: 0.9855\nEpoch 62/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0473 - acc: 0.9847\nEpoch 63/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0469 - acc: 0.9850\nEpoch 64/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0483 - acc: 0.9842\nEpoch 65/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0445 - acc: 0.9855\nEpoch 66/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0500 - acc: 0.9844\nEpoch 67/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0488 - acc: 0.9850\nEpoch 68/100\n3662/3662 [==============================] - 41s 11ms/step - loss: 0.0421 - acc: 0.9861\nEpoch 69/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0439 - acc: 0.9855\nEpoch 70/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0423 - acc: 0.9861\nEpoch 71/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0451 - acc: 0.9828\nEpoch 72/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0479 - acc: 0.9847\nEpoch 73/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0457 - acc: 0.9858\nEpoch 74/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0437 - acc: 0.9850\nEpoch 75/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0449 - acc: 0.9839\nEpoch 76/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0444 - acc: 0.9853\nEpoch 77/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0458 - acc: 0.9844\nEpoch 78/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0433 - acc: 0.9874\nEpoch 79/100\n3662/3662 [==============================] - 40s 11ms/step - loss: 0.0480 - acc: 0.9839\nEpoch 80/100\n3300/3662 [==========================&gt;...] - ETA: 3s - loss: 0.0405 - acc: 0.9842",
    "618741": "Looks like you are facing overfit issues. You should split a validation fold to estimate the loss instead of only  train set.",
    "619015": "Thanks a lot, InfiniteWing!\nIndeed, I've seen a sign of overfitting when looked at nearly constant validation loss..",
    "619279": "Hi @polishch, accuracy is not a good metric to save checkpoint for this competition, you should use Kappa score.",
    "619292": "Checked the how to submit successfully discussions, the id_code in test.csv is not correlated with sample_submission.csv. I think it's something wrong with your final csv rather than model.",
    "619301": "do not rely on accuracy score. If you treat the problem as regression, you have to implement MSE loss (loss function) and also monitor the required metrics (quadratic weighted kappa). there are many useful kernels published for the function of loss and metrics.",
    "619606": "Thank you jionie!\n\nNow I use sample_submission.csv, it seems my score increased because of that (the model is also a bit different, but I'm in doubt it can explain eight times increase in score..)"
  },
  "source": "meta"
}