Class3__MNIST NN Tensorflow 2.0.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Deep learning : Neural network with many layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.1.0'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
      "11493376/11490434 [==============================] - 28s 2us/step\n"
     ]
    }
   ],
   "source": [
    "mnist = tf.keras.datasets.mnist\n",
    "\n",
    "(x_train_full, y_train_full), (x_test, y_test) = mnist.load_data()\n",
    "x_train_full, x_test = x_train_full / 255.0, x_test / 255.0\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train_full.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000,)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_full.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_full[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# one-hot encoding\n",
    "y_train_full_onehot = np.zeros((y_train_full.size, y_train_full.max()+1))\n",
    "y_train_full_onehot[np.arange(y_train_full.size),y_train_full] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_full[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 0., 0., 0., 1., 0., 0., 0., 0.])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# after one-hot encoding\n",
    "y_train_full_onehot[0,]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# among train data -> train & validate\n",
    "x_train, x_valid  = x_train_full[0:50000, :, :], x_train_full[50000:, :, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train, y_valid  = y_train_full_onehot[0:50000, :], y_train_full_onehot[50000:, :]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### train(50000), valid(10000), test(10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 28, 28)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 10)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(28, 28)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train[0,:,:].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1f86525f8c8>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.imshow(x_train[0,:,:]) # 0th picture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1f87fd5ff88>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[1,:,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1f87fdd6148>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[2,:,:]) # 2nd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 0., 0., 1., 0., 0., 0., 0., 0.])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train[2,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)), # 0th layer\n",
    "  tf.keras.layers.Dense(128, activation='relu'), # 1st layer : 784*128+128\n",
    "  tf.keras.layers.Dense(128, activation='relu'),\n",
    "  tf.keras.layers.Dense(10, activation=\"softmax\")\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 128)               100480    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 118,282\n",
      "Trainable params: 118,282\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten_4 (Flatten)          (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 128)               100480    \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 118,282\n",
      "Trainable params: 118,282\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define L\n",
    "model.compile(loss=\"categorical_crossentropy\", optimizer=\"sgd\", metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# stochastic gradient descent algorithm \n",
    "# gradient descent처럼 모든 데이터를 다 미분하는 게 아니고 \n",
    "# 한 스텝 옮길 때마다 sigma 중에 몇 개만 뽑아서 여러번 (batch) -> 매 iteration마다 randomly chose\n",
    "# 128개의 batch (for each move, parameter is moving), 500번 정도 해야 모든 데이터를 다 사용\n",
    "# epoch 20 바퀴 돌아라 -> epoch을 올릴 수록 accuracy 증가\n",
    "# validation : train이 잘 되는지 정보를 주기 위함 (val_loss, val_accuracy)\n",
    "# accuray : 정확도 - train accuracy : mostly better and better\n",
    "# 따라서 val_accuracy가 계속 증가하는지를 봐야 함"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "50000/50000 [==============================] - 1s 14us/sample - loss: 0.1794 - accuracy: 0.9492 - val_loss: 0.1760 - val_accuracy: 0.9513\n",
      "Epoch 2/20\n",
      "50000/50000 [==============================] - 1s 23us/sample - loss: 0.1747 - accuracy: 0.9498 - val_loss: 0.1729 - val_accuracy: 0.9519\n",
      "Epoch 3/20\n",
      "50000/50000 [==============================] - 3s 60us/sample - loss: 0.1704 - accuracy: 0.9517 - val_loss: 0.1665 - val_accuracy: 0.9537\n",
      "Epoch 4/20\n",
      "50000/50000 [==============================] - 2s 44us/sample - loss: 0.1662 - accuracy: 0.9521 - val_loss: 0.1631 - val_accuracy: 0.9554\n",
      "Epoch 5/20\n",
      "50000/50000 [==============================] - 2s 36us/sample - loss: 0.1621 - accuracy: 0.9538 - val_loss: 0.1613 - val_accuracy: 0.9553\n",
      "Epoch 6/20\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.1583 - accuracy: 0.9550 - val_loss: 0.1571 - val_accuracy: 0.9565\n",
      "Epoch 7/20\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.1545 - accuracy: 0.9563 - val_loss: 0.1541 - val_accuracy: 0.9578\n",
      "Epoch 8/20\n",
      "50000/50000 [==============================] - 2s 31us/sample - loss: 0.1511 - accuracy: 0.9566 - val_loss: 0.1518 - val_accuracy: 0.9584\n",
      "Epoch 9/20\n",
      "50000/50000 [==============================] - 2s 32us/sample - loss: 0.1477 - accuracy: 0.9578 - val_loss: 0.1490 - val_accuracy: 0.9586\n",
      "Epoch 10/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.1444 - accuracy: 0.9589 - val_loss: 0.1468 - val_accuracy: 0.9594\n",
      "Epoch 11/20\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.1412 - accuracy: 0.9598 - val_loss: 0.1450 - val_accuracy: 0.9609\n",
      "Epoch 12/20\n",
      "50000/50000 [==============================] - 1s 30us/sample - loss: 0.1382 - accuracy: 0.9609 - val_loss: 0.1434 - val_accuracy: 0.9606\n",
      "Epoch 13/20\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.1354 - accuracy: 0.9617 - val_loss: 0.1407 - val_accuracy: 0.9609\n",
      "Epoch 14/20\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.1327 - accuracy: 0.9623 - val_loss: 0.1380 - val_accuracy: 0.9618\n",
      "Epoch 15/20\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.1299 - accuracy: 0.9631 - val_loss: 0.1365 - val_accuracy: 0.9631\n",
      "Epoch 16/20\n",
      "50000/50000 [==============================] - 1s 24us/sample - loss: 0.1274 - accuracy: 0.9638 - val_loss: 0.1346 - val_accuracy: 0.9634\n",
      "Epoch 17/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.1249 - accuracy: 0.9652 - val_loss: 0.1335 - val_accuracy: 0.9634\n",
      "Epoch 18/20\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.1224 - accuracy: 0.9652 - val_loss: 0.1316 - val_accuracy: 0.9642\n",
      "Epoch 19/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.1202 - accuracy: 0.9661 - val_loss: 0.1299 - val_accuracy: 0.9637\n",
      "Epoch 20/20\n",
      "50000/50000 [==============================] - 1s 23us/sample - loss: 0.1179 - accuracy: 0.9668 - val_loss: 0.1293 - val_accuracy: 0.9637\n"
     ]
    }
   ],
   "source": [
    "history =  model.fit(x_train, y_train, epochs=20, validation_data=(x_valid, y_valid), batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/30\n",
      "50000/50000 [==============================] - 2s 31us/sample - loss: 0.1157 - accuracy: 0.9676 - val_loss: 0.1294 - val_accuracy: 0.9635\n",
      "Epoch 2/30\n",
      "50000/50000 [==============================] - 2s 49us/sample - loss: 0.1136 - accuracy: 0.9679 - val_loss: 0.1250 - val_accuracy: 0.9646\n",
      "Epoch 3/30\n",
      "50000/50000 [==============================] - 2s 39us/sample - loss: 0.1117 - accuracy: 0.9681 - val_loss: 0.1243 - val_accuracy: 0.9649\n",
      "Epoch 4/30\n",
      "50000/50000 [==============================] - 2s 41us/sample - loss: 0.1095 - accuracy: 0.9694 - val_loss: 0.1229 - val_accuracy: 0.9663\n",
      "Epoch 5/30\n",
      "50000/50000 [==============================] - 2s 31us/sample - loss: 0.1077 - accuracy: 0.9693 - val_loss: 0.1214 - val_accuracy: 0.9665\n",
      "Epoch 6/30\n",
      "50000/50000 [==============================] - 1s 30us/sample - loss: 0.1057 - accuracy: 0.9704 - val_loss: 0.1208 - val_accuracy: 0.9659\n",
      "Epoch 7/30\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.1038 - accuracy: 0.9706 - val_loss: 0.1198 - val_accuracy: 0.9664\n",
      "Epoch 8/30\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.1022 - accuracy: 0.9709 - val_loss: 0.1175 - val_accuracy: 0.9667\n",
      "Epoch 9/30\n",
      "50000/50000 [==============================] - 2s 33us/sample - loss: 0.1004 - accuracy: 0.9718 - val_loss: 0.1164 - val_accuracy: 0.9667\n",
      "Epoch 10/30\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.0989 - accuracy: 0.9721 - val_loss: 0.1151 - val_accuracy: 0.9679\n",
      "Epoch 11/30\n",
      "50000/50000 [==============================] - 1s 27us/sample - loss: 0.0972 - accuracy: 0.9724 - val_loss: 0.1135 - val_accuracy: 0.9689\n",
      "Epoch 12/30\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.0953 - accuracy: 0.9730 - val_loss: 0.1138 - val_accuracy: 0.9673\n",
      "Epoch 13/30\n",
      "50000/50000 [==============================] - 2s 39us/sample - loss: 0.0939 - accuracy: 0.9740 - val_loss: 0.1125 - val_accuracy: 0.9684\n",
      "Epoch 14/30\n",
      "50000/50000 [==============================] - 2s 33us/sample - loss: 0.0925 - accuracy: 0.9743 - val_loss: 0.1116 - val_accuracy: 0.9689\n",
      "Epoch 15/30\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.0909 - accuracy: 0.9747 - val_loss: 0.1107 - val_accuracy: 0.9686\n",
      "Epoch 16/30\n",
      "50000/50000 [==============================] - 2s 30us/sample - loss: 0.0894 - accuracy: 0.9752 - val_loss: 0.1097 - val_accuracy: 0.9686\n",
      "Epoch 17/30\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.0881 - accuracy: 0.9752 - val_loss: 0.1080 - val_accuracy: 0.9691\n",
      "Epoch 18/30\n",
      "50000/50000 [==============================] - 1s 30us/sample - loss: 0.0865 - accuracy: 0.9759 - val_loss: 0.1082 - val_accuracy: 0.9690\n",
      "Epoch 19/30\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.0855 - accuracy: 0.9767 - val_loss: 0.1060 - val_accuracy: 0.9698\n",
      "Epoch 20/30\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.0842 - accuracy: 0.9769 - val_loss: 0.1066 - val_accuracy: 0.9706\n",
      "Epoch 21/30\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.0828 - accuracy: 0.9771 - val_loss: 0.1064 - val_accuracy: 0.9700\n",
      "Epoch 22/30\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.0816 - accuracy: 0.9775 - val_loss: 0.1055 - val_accuracy: 0.9703\n",
      "Epoch 23/30\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.0803 - accuracy: 0.9783 - val_loss: 0.1040 - val_accuracy: 0.9707\n",
      "Epoch 24/30\n",
      "50000/50000 [==============================] - 1s 30us/sample - loss: 0.0791 - accuracy: 0.9783 - val_loss: 0.1038 - val_accuracy: 0.9706\n",
      "Epoch 25/30\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.0780 - accuracy: 0.9787 - val_loss: 0.1032 - val_accuracy: 0.9710\n",
      "Epoch 26/30\n",
      "50000/50000 [==============================] - 2s 35us/sample - loss: 0.0768 - accuracy: 0.9791 - val_loss: 0.1023 - val_accuracy: 0.9709\n",
      "Epoch 27/30\n",
      "50000/50000 [==============================] - 2s 31us/sample - loss: 0.0757 - accuracy: 0.9796 - val_loss: 0.1012 - val_accuracy: 0.9710\n",
      "Epoch 28/30\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.0745 - accuracy: 0.9799 - val_loss: 0.1009 - val_accuracy: 0.9708\n",
      "Epoch 29/30\n",
      "50000/50000 [==============================] - 2s 31us/sample - loss: 0.0736 - accuracy: 0.9801 - val_loss: 0.1004 - val_accuracy: 0.9712\n",
      "Epoch 30/30\n",
      "50000/50000 [==============================] - 2s 36us/sample - loss: 0.0726 - accuracy: 0.9804 - val_loss: 0.0998 - val_accuracy: 0.9712\n"
     ]
    }
   ],
   "source": [
    "history =  model.fit(x_train, y_train, epochs=30, validation_data=(x_valid, y_valid), batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# google colabotory - cloud"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Let's draw learning curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "a = history.history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a.keys() # dictionary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(a[\"accuracy\"], label=\"train\")\n",
    "plt.plot(a[\"val_accuracy\"], label=\"valid\")\n",
    "plt.legend(fontsize='x-large')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "a = history.history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAD4CAYAAAD2FnFTAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/d3fzzAAAACXBIWXMAAAsTAAALEwEAmpwYAAA6HUlEQVR4nO3deXhU1fnA8e+bDRKWhCUESICwhk1EiCCuKIJgrahVBERQQaBVrD+7qGittbai3aR1QVRULIuAC7ijFgVcgLBGApGwJ4QQ1gAhy0ze3x93wCEGMglJJpO8n+fJk5l7z9z7HobMO/ecc88RVcUYY4wpLsjfARhjjKmeLEEYY4wpkSUIY4wxJbIEYYwxpkSWIIwxxpQoxN8BVISmTZtqfHy8v8MwxpiAsnr16v2qGn2m/TUiQcTHx5OUlOTvMIwxJqCIyM6z7bcmJmOMMSWyBGGMMaZEliCMMcaUyBKEMcaYElmCMMYYUyJLEMYYY0pUI4a5lqawsJD09HTy8vL8HUqNEBwcTFRUFE2bNiUoyL5jGFNT1YoEkZ6eToMGDYiPj0dE/B1OQFNVCgsLycrKIj09ndatW/s7JGNqpcwjJ/gsJYum9etw7XktKuUctSJB5OXlWXKoICJCWFgYsbGxpKam+jscY2oNVWXLvmMs3riXxSlZbEg/AsD157e0BHGuLDlULGtaMqbyuYuUtbsOsTgli8Ub97LjQC4AF7SO4sHBnRnYNYYOzepX2vl9ShAiMhiYCgQDr6jqlGL7GwEzgPZAHnCXqn7v2fd/wDhAgWTgTlXNE5HGwFtAPLADGKaqhzyveRgYC7iB+1T103OrpjHGBIbcAhffbj3AZylZfL4pi/3HCggNFi5u35S7L2/H1V1iiGlYt0piKTVBiEgw8DwwEEgHVonIIlVN8So2GVinqjeKSGdP+QEiEgvcB3RV1RMiMg8YDrwOPAR8oapTROQhz/MHRaSrp0w3oCXwuYh0UlV3BdXZGGOqDXeRkpxxhOVbslmetp81Ow9T4C6ifp0QruzcjEFdY+ifEE2DuqFVHpsvVxB9gDRV3QYgInOBoYB3gugKPAWgqptFJF5EYrzOES4ihUAEsMezfSjQ3/P4DeBL4EHP9rmqmg9sF5E0TwzflqeC5kevv/4648aNw+Vy+TsUY2otVWXngVyWp+1n+Zb9fLN1Pzl5zt9k1xYNueOSeC7t0JS+7RpTJyTYr7H6kiBigd1ez9OBvsXKrAduApaLSB+gDRCnqqtF5O/ALuAEsFhVF3teE6OqmQCqmikizbzO912x88UWD0pExgPjgRo9kubqq68mLi6O119//ZyPdeuttzJkyJBzD8oYUyY5eYUs37KfpT84Vwnph04A0DKyLoO7N+fSjtFc3L4JTevX8XOkp/MlQZTUu6vFnk8BporIOpx+hrWAy9M3MRRoCxwG5ovIKFX97zmeD1WdDkwHSExM/Mn+2qSgoICwsLBSy4WHhxMeHl4FERlTu6kqP2QdY0nqPpZs3kfSzkO4i5QGdULo174JEy5vxyUdmtK2ab1qPYDGl6Eo6UArr+dx/NhMBICq5qjqnaraExgNRAPbgauB7aqaraqFwDvAxZ6XZYlICwDP732+nq+2uOOOO/jiiy944403EBFEhNdffx0RYdasWVx77bXUq1ePyZMno6rcfffdtG/fnvDwcNq1a8fkyZPJz88/dbzXX3+dkJCQnzz/+uuv6dWrFxEREVx44YWsXr3aH9U1JqDlFrj4PCWLR95N5tKnl3DNs0uZ8vFmcvJcTLi8HQsm9mPtYwOZPjqR2/vF0y66frVODuDbFcQqoKOItAUycDqQR3oXEJEoIFdVC3BGLC1V1RwR2QVcJCIROE1MA4CTK/ssAsbgXH2MARZ6bZ8tIv/E6aTuCKwsdw1L8Kf3N5KyJ6ciD+mTri0b8sefd/O5/NSpU9m2bRstWrRg6tSpAOTkOHE/+OCDTJkyheeeew4RQVWJiYlh9uzZxMTEsGHDBiZMmEBoaCh/+tOfzniOoqIiHn74YaZOnUp0dDT33Xcfw4YNIzU19bRkYoz5qWP5LhYk7eZ/qdl8t+0ABa4iIsKCubRDUyZd1YH+Cc1oHlk1I44qQ6mfAKrqEpF7gU9xhrnOUNWNIjLRs38a0AWYKSJunM7rsZ59K0RkAbAGcOE0PU33HHoKME9ExuL0Udziec1Gz2inFM9r7qmtI5giIyMJCwsjPDyc5s2bA5yaLmTChAmMGjXqtPJPPvnkqcfx8fFs3bqVF1544awJQlV59tln6dWrFwBPPPEE/fr1Y+vWrSQkJFR0lYypEVSV9zdk8pcPU8jKyadddD1uv6gNV3VuRmJ8I793LlcUn74iqupHwEfFtk3zevwtzjf9kl77R+CPJWw/gHNFUdJr/gL8xZfYyqMs3+Krqz59+vxk28svv8wrr7zCjh07OH78OC6Xi6KiorMeR0Q4//zzTz2PjXXGA2RlZVmCMKYEW7KO8tjCjXy77QDnxUby4qje9GrdyN9hVQprQwhQ9erVO+35/Pnzueeee5gyZQpXXHEFDRs2ZP78+TzyyCNnPU5QUBDBwT9+2znZJlpaYjGmtjme7+Lf/9vCq8u2U69OCE/e0J0RfVoTHFS9+xHOhSWIai4sLAy3u/QWtqVLl3LBBRfwwAMPnNq2Y8eOSozMmNpBVfkoeS9//iCFvTl5DEuM48HBnWlSzYakVgZLENVc27ZtWbJkCVu3biUyMpLCwsISyyUkJPDqq6+ycOFCunfvzgcffMA777xTxdEaU7NszT7GHxduZHnafrq2aMjzt/Wid5ua2ZxUEksQ1dxvfvMbkpOTOf/88zl+/DivvfZaieUmTJhAcnIyd955Jy6Xi+uuu47HH3+cSZMmVXHExgS+3AIXz/0vjZeXbaNuaDBPDO3GbX3b1OjmpJKIauDfY5aYmKhJSUln3L9p0ya6dOlShRHVDvbvamqSoiJl9a5DvLc2gw+TMzmcW8gvesXx0JDORDeomc1JIrJaVRPPtN+uIIwxtdqWrKO8ty6Dhev2kH7oBHVDgxjUtTljLm5D7zaN/R2eX1mCMMbUOlk5eSxat4f31mWwcU8OQQKXdozmgYGdGNStOfXr2EcjWIIwxtQSeYVu3l/vJIVvth5AFc6Pi+Sx67py3fktaNYgcO94riyWIIwxNdqRE4X897udzFi+nQPHC2jdOIJJV3VkaM+WtI+uvNXYagJLEMaYGikrJ48Zy7cza8UujuW76J8QzYTL23NRu8bVfpK86sIShDGmRtm+/zjTl27l7dUZuIqKuK5HSyZe0Z6uLRv6O7SAYwnCGFMjJKcfYdpXW/no+0xCg4MYdmEc4y9rT+smEf4OLWBZgjDGVGvuIiW3wEVugdvz4/U438XRfBfvr9/Dsi37aVAnhF9e0Z47L2lbY+9dqEqWIIwx1c67a9N5+uNUDuUWkO8qfeLI6AZ1eGhIZ0b2bU3DuqFVEGHtYAmihnv99dcZN24cLpezKPqXX37JlVdeye7du4mLizvj60SEN9988ydrThhTmYqKlH99/gP/+V8aF7SOYmjPloSHBVMvLMT5XSeY8NAQ6tUJJiLMeRwRFkzLqHDCQnxZINOUhSWIWubiiy8mMzOTZs2a+TsUY06TV+jmN/PX8+GGTG7pHcdfbjzPPvT9zBJELRMWFnZqdTpjqovso/ncPTOJdbsP8+Dgzky8op0NRa0GLD1XYy+//DKRkZGcOHHitO1PP/00sbGxuN1u7r77btq3b094eDjt2rVj8uTJ5Ofnn/GYX375JSJCenr6qW1LliyhR48e1K1blx49erBkyZJKq5MxxaXuPcoNz3/N5r05TBvVi1/2b2/JoZqonVcQHz8Ee5Or/rzNz4MhU3wuPmzYMO677z7ee+89RowYcWr7yb4BESEmJobZs2cTExPDhg0bmDBhAqGhoWddh9rbnj17uO666xg2bBhz584lIyODX//612WumjHl8WXqPu6dvZaIsGDmTehHj7gof4dkvNgVRDUWGRnJ0KFDmTlz5qlta9asYePGjYwePZqgoCCefPJJ+vbtS3x8PNdffz2///3vmT17ts/neOGFF2jatCkvv/wyXbt2ZeDAgfz1r3+tjOoYc5o3vtnBXa+vonXjCBbee4klh2rIpysIERkMTAWCgVdUdUqx/Y2AGUB7IA+4S1W/F5EE4C2vou2Ax1T1WRF5C0jwbI8CDqtqTxGJBzYBqZ5936nqxPJU7ozK8C3e30aPHs3111/P3r17ad68OW+++Sa9e/emW7dugNMM9corr7Bjxw6OHz+Oy+Uq03rSKSkp9OnTh5CQH/8rXHrppRVeD2NOcrmL+PMHKbzx7U6u7tKMqcMvoJ7NnlotlXoFISLBwPPAEKArMEJEuhYrNhlYp6o9gNE4yQRVTVXVnqraE+gN5ALvevbd6rXvbcB7fcytJ/dVeHIIMNdccw3R0dHMmjULl8vFnDlzGD16NADz58/nnnvu4dZbb+Wjjz5i7dq1PPbYY2dclrQkqvqT9l5r/zWV5WheIeNmJvHGtzsZd2lbXro90ZJDNebLO9MHSFPVbQAiMhcYCqR4lekKPAWgqptFJF5EYlQ1y6vMAJwP/p3eBxfn02gYcFX5q1FzBQcHM3LkSGbOnEmXLl04ePDgqf6IpUuXcsEFF/DAAw+cKr9jx44yHb9bt268+eabuN1ugoODAVi+fHmFxW8MwMHjBcxdtYs3v93JvqP5/OXG7tzWt42/wzKl8KUPIhbY7fU83bPN23rgJgAR6QO0AYrfhTUcmFPC8S8DslR1i9e2tiKyVkS+EpHLSgpKRMaLSJKIJGVnZ/tQjcA1ZswYNmzYwCOPPMKQIUOIjo4GICEhgeTkZBYuXMjWrVuZOnUq77zzTilHO90vf/lLsrOzGT9+PJs2beKLL77gkUceqYxqmFooOf0Iv52/noue+oJnPkmlbdN6zB7X15JDgPDlCqKk9obiC1lPAaaKyDogGVgLuE4dQCQMuB54uIRjjeD0xJEJtFbVAyLSG3hPRLqpas5pAahOB6aDsya1D/UIWD169KBnz56sW7eORx999NT2CRMmkJyczJ133onL5eK6667j8ccfZ9KkST4fOzY2lvfff5/777+fnj170rFjR/79738zYMCAyqiKqQUKXEV8/H0mb3yzgzW7DhMRFsywxDjG9IunY0wDf4dnykBUz/7ZKiL9gMdV9RrP84cBVPWpM5QXYDvQ4+SHuogMBe5R1UHFyoYAGUBvVU3/ycGcMl8Cv1XVpDPFmJiYqElJZ9zNpk2b6NKlyxn3m/Kxf1fjLSsnj1krdjF7xS72H8unbdN63H5RG25OjLP5kaopEVmtqoln2u/LFcQqoKOItMX5MB8OjCx2kiggV1ULgHHA0mLf+ItfJZx0NbDZOzmISDRwUFXdItIO6Ahs8yFOY0wVO5pXyPIt+/kwOZNPvt+LW5UrE5oxul8bLu8YTVCQDXgIZKUmCFV1ici9wKc4w1xnqOpGEZno2T8N6ALMFBE3Tuf12JOvF5EIYCAwoYTDl9QvcTnwhIi4ADcwUVUPlrlmxpgKp6qk7TvGktR9LNmczaodB3EVKZHhoYy5OJ7bL2pDfNN6/g7TVBCfxpep6kfAR8W2TfN6/C3ON/2SXpsLNDnDvjtK2PY2zrBXY0w1cKLAzbfb9rNkczZLUveRfsiZ+qVz8waMu6wdV3VuRq/WUYQE2323NY0NQDbGnOJyF7HrYC5bs4+Ttu8Y3207wLfbDlDgKiIiLJhLOjTlV/070D8hmpZR4f4O11SyWpMgSrohzJRfaYMbTPWWk1fItuzjbN13jK3ZJ3+Os/PAcQrdP763bZvWY1TfNlzZOZo+bRtTJyTYj1GbqlYrEkRwcDCFhYWEhYX5O5Qa48SJE4SG2siUQPPFpiweW7iRjMM/zhAcEiS0aRJB++j6DOwaQ/vo+rSPrke76PpEhtt7XJvVigQRFRVFVlYWsbGxBAVZO+m5UFVOnDhBRkYGMTEx/g7HlMGclbt45N1kEpo3ZNRFbWgfXY/2zerTunEEodZ/YEpQKxJE06ZNSU9PJzU1tfTCplShoaHExMTQsGFDf4difKCqPPv5FqZ+sYUrOkXzwm29bP4j45Na8b8kKCiI1q1b+zsMY6qcy13EI+9+z1tJu7m5dxxP3XSeXS0Yn9WKBGFMbZRb4OLe2Wv53+Z93HtlB34zqJMN1DBlYgnCmBrowLF8xr6RxPr0w/z5hu7cfpFNjmfKzhKEMTXMrgO5jHltJXsOn+DF23ozuHtzf4dkApQlCGNqkO8zjnDHa6sodBcxa1xfEuMb+zskE8AsQRhTQyz9IZtf/nc1URFhzB3flw7NbGptc24sQRhTAyxYnc5Db2+gQ7P6vHFXH2Ia1vV3SKYGsARhTABzuYt46uPNvLp8Oxe3b8K023vb2gumwliCMCZAHTxewL2z1/DN1gOM6deGR6/ravc4mAplCcKYALRxzxHGz1xN9tF8nrm5B8MSW/k7JFMDWYIwJsAsWr+H3y9YT1R4GPMm9qNnqyh/h2RqKEsQxgQId5HyzKebeemrbSS2acQLo3rRrIF1RpvKYwnCmABwOLeASXPWsmzLfm7r25o//rwbYSHW32AqlyUIY6q5zXtzGD9zNZlHTvDUTecxoo9NPGmqhiUIY6oRVcVdpBS6lQJ3EUt/yObBtzdQv04Ic8f3o3ebRv4O0dQiPiUIERkMTAWCgVdUdUqx/Y2AGUB7IA+4S1W/F5EE4C2vou2Ax1T1WRF5HLgbyPbsm6yqH3mO9zAwFnAD96nqp+WsnzHVyua9OTz72RZSMnNwuYsocCuF7iJc7qJTSaG4Xq2jeHFUb7v5zVS5UhOEiAQDzwMDgXRglYgsUtUUr2KTgXWqeqOIdPaUH6CqqUBPr+NkAO96ve5fqvr3YufrCgwHugEtgc9FpJOqustZR2P8Lv1QLv/87AfeXZtB/TohXJnQjDohQYSGBBEaJIQGex4He557HkeGh/Lz81vYWtDGL3y5gugDpKnqNgARmQsMBbwTRFfgKQBV3Swi8SISo6pZXmUGAFtVdWcp5xsKzFXVfGC7iKR5YvjWpxoZU40cPF7Ac/9L47/f7QSBuy9rx6/6tycqwtZHN9WfLwkiFtjt9Twd6FuszHrgJmC5iPQB2gBxgHeCGA7MKfa6e0VkNJAE/EZVD3nO912x88UWD0pExgPjAVstzlQ7uQUuXl22nZeWbiO3wMXNveO4/+pOtIwK93doxvjMlwRR0hJUWuz5FGCqiKwDkoG1gOvUAUTCgOuBh71e8yLwZ8+x/gz8A7jLx/OhqtOB6QCJiYk/2W+MPxS6i5i7chdTv0hj/7F8BnWN4XfXJNAxxmZWNYHHlwSRDnjfxx8H7PEuoKo5wJ0A4qxpuN3zc9IQYI13k5P3YxF5GfjA1/MZU924i5QPkzP5x+JUdh7IpU98Y166vbeNOjIBzZcEsQroKCJtcTqZhwMjvQuISBSQq6oFwDhgqSdpnDSCYs1LItJCVTM9T28Evvc8XgTMFpF/4nRSdwRWlqVSxlSVw7kFzEvazZvf7WT3wRN0bt6AGXckcmVCM1v/2QS8UhOEqrpE5F7gU5xhrjNUdaOITPTsnwZ0AWaKiBun83rsydeLSATOCKgJxQ79jIj0xGk+2nFyv+fY8zzHcQH32AgmU92k7Mlh5rc7eG9dBnmFRfRp25iHh3Thmm7NCQ6yxGBqBlEN/Ob7xMRETUpK8ncYpoYrdBfx6ca9zPxmJyt3HKRuaBA3XhDL6H7xdGnR0N/hGVNmIrJaVRPPtN/upDamFNlH85mzchezVuwkKyefVo3DeeTaLgxLbEVkhC3OY2ouSxDGlMBdpHyzdT/zk9L5+PtMCt3KZR2b8tcbz6N/QjNrRjK1giUIY7zsPHCcBavTeXt1OnuO5NGwbgi39W3D7f3a0D66vr/DM6ZKWYIwtd7xfBcfJWcyf3U6K7cfRAQu6xjNw9d2YWDXGOqG2jQXpnayBGFqJVUlaech5q3azUfJmRwvcBPfJILfXZPATb1iaRFpdzwbYwnC1Dqb9+bwq1lr2JZ9nHphwfysRwtuSWxFYptGdu+CMV4sQZhaJePwCcbMcO67/NvNPbj2vBbUq2N/BsaUxP4yTK1xOLeAMTNWklvgZsHEi0lobvMjGXM2tqitqRXyCt2MfSOJXQdyeXl0oiUHY3xgVxCmxnO5i5g0Zy1rdh3i+ZG9uKhdE3+HZExAsCsIU6OpKn9YuJHPUrJ4/OfduPa8Fv4OyZiAYQnC1Gj//iKNOSt38av+7Rlzcby/wzEmoFiCMDXW3JW7+NfnP/CLXnH87poEf4djTMCxBGFqpM9Tspj8bjJXdIpmyi/Os/sbjCkHSxCmxlm98xD3zllD99hIXritF6HB9t/cmPKwvxxTo6TtO8bYN1bRvGFdZtxxod0EZ8w5sARhaoysnDzGzFhJSJDwxl19aFq/jr9DMiagWYIwNcKSzfu46YVvOJxbwGt39KFNk3r+DsmYgGfX3yagZeXk8cT7KXyYnEmHZvX5z8i+nBcX6e+wjKkRLEGYgOQuUmav2Mkzn6SS7y7iNwM7MeGK9oSF2EWxMRXFp78mERksIqkikiYiD5Wwv5GIvCsiG0RkpYh092xPEJF1Xj85InK/Z9/fRGSz5zXvikiUZ3u8iJzwes20iquuqQlS9uTwixe/4Q8LN9KjVSSf3n85kwZ0tORgTAUr9QpCRIKB54GBQDqwSkQWqWqKV7HJwDpVvVFEOnvKD1DVVKCn13EygHc9r/kMeFhVXSLyNPAw8KBn31ZV7XmulTM1S26Bi6mfb+GV5duJCg/l2Vt7MrRnS7vHwZhK4ksTUx8gTVW3AYjIXGAo4J0gugJPAajqZs9VQIyqZnmVGYDzwb/TU26x177vgJvLXw1T0y3ZvI9H3/uejMMnGH5hKx4a0pmoiDB/h2VMjeZLgogFdns9Twf6FiuzHrgJWC4ifYA2QBzgnSCGA3POcI67gLe8nrcVkbVADvCoqi7zIU5Tw6gqyRlHmPbVVj5K3kuHZvWZN6Effdo29ndoxtQKviSIkq7ftdjzKcBUEVkHJANrAdepA4iEAdfjNCOdfnCRRzxlZ3k2ZQKtVfWAiPQG3hORbqqaU+x144HxAK1bt/ahGiZQ7D+Wz3trM5iflE5q1lHqhgbx20GdGH+5dUIbU5V8SRDpQCuv53HAHu8Cng/vOwHEaRDe7vk5aQiwpliTEyIyBrgOp79CPcfKB/I9j1eLyFagE5BU7JzTgekAiYmJxROWCTCF7iKWbN7H/NXpLNm8D1eR0rNVFH+5sTvX9WhJZHiov0M0ptbxJUGsAjqKSFucTubhwEjvAp4RSLmqWgCMA5YW+8Y/gmLNSyIyGKdT+gpVzfXaHg0cVFW3iLQDOgLbyloxExhS9x5lftJu3luXwf5jBTStX4exl7bl5t5xdIyxVd+M8adSE4RnlNG9wKdAMDBDVTeKyETP/mlAF2CmiLhxOq/Hnny9iETgjICaUOzQzwF1gM88o1C+U9WJwOXAEyLiAtzARFU9eG7VNNXJkdxCFm3Yw4Kk3axPP0JIkDCgSzNu6d2KKxKibXI9Y6oJ8bTsBLTExERNSkoqvaDxG3eR8nXafuavTufTjXspcBXRuXkDbklsxQ09W9LE5k0ypsqJyGpVTTzTfruT2lSqnQeOs2B1Om+vTmfPkTwiw0MZcWErbklsRbeWDe0eBmOqMUsQpsIdz3fxUXIm81ens3L7QYIELusYzeSfdeHqLjHUDQ32d4jGGB9YgjAV5nBuAVM+3sz76/dwvMBN26b1+N01CfyiVxzNI+v6OzxjTBlZgjAV4nBuAaNeXcEPe49xwwUtuSWxFYltGlkTkjEBzBKEOWdHcgtPJYeXRvfmyoRm/g7JGFMBbDyhOSenJYfbLTkYU5NYgjDlduREIbfPWEHq3qNMu70XV3a25GBMTWIJwpTLkROFjH51BZsyc3hxVC+u6hzj75CMMRXM+iBMmeXkFTJ6xkpSMnOYNqo3A7pYcjCmQpX1BuZKGgxiCcKUSU5eIbe/upKUPUd48TZLDsZUqN0r4dvnYNMHoG7fXtPtJrjltUoJxxKE8VlOXiGjPcnh+ZG9uLqrJQdjzpnbBZvfh2+fh/RVUDcKLhwHET6uexLdudJCswRhfHI0r5AxM1byfcYRXritF4O6Nfd3SMb4rsgNmz+E1I+dD96GsdCwJUTGOY/rN4OgUu7wdxfC0UzI2QNH0p3fORmQdwRiukOrvtCiB4T4OK9YXg6sfRO+mwZHdkGjtnDt36HnSAird+51rgCWIEypjnr6HJLTj/C8JQcTSPKPwtpZ8N0LcHgnhDeCwjxwnTi9XFAINGjhJI2TyaPIDTmeRHAkA45l8ZO10sIaOB/m6z2rGQSHQYue0KoPxF3o/G7Y8vTXHN4NK6bBmpmQnwOtL4YhU6DT4NKTVBWzBGF+oqhI2bz3KCu2H2DFtoOs2H6Ao3kunhvZi2ssOZhAcCQDVr4ESa9D/hHn2/2gP0Pn60CC4MQh59v/kQznd07Gj1cGmesh9SMnaZxMFh27QMM4z1VHrGd7LNRt6Jzv6F6n/yB9pfN75ctOXwJAZCsnWcQlQnoSpCx0tne7Efr9CmJ7++WfyBc23bfB5S4iJTPnVDJYuf0gOXnOirFxjcLp27YJN/eOo1/7Jn6O1NQaqs6HdnhjCIvw/XV71jpt+RvfBS2CrkPhonug1YVlPz+Uf3SQqwD2bvBKGqucq5E6kdB7DPSd4DRv+ZlN921+wl2kbNxzhK/TDvDdtgOs3nmIY/lOQmjbtB7XnteCPm0b07ddE2Kjwv0cralVXAXOh/u3zzkfsOA0C538xn7qG3zcj30IDZrDtq+cxLBzudPs02eC8yHcqE354jjXYaMhYc4VQ1wi8Ctn29G9UKdBtelf8IUliFpi14FclqVl83Xafr7ZeoDDuYUAdGhWn6E9W9K3XRP6tm1MTEObddUAx7Jh3minTfznU6FJ+8o934lDkPQarJzudAQ3TYBBT4K74Mc+gJwMyEiC3AMlH6NhnPOaXqOhbmTlxlseDQKvedYSRA11OLeAb7YeYNmW/Xydtp9dB51lv5s3rMvVXWK4rGNTLm7flOgGtpKbKebgNnjzJucbb3AYvHgJXP049BkPQRU8+cLBbfDdi7D2v1CYC+36w/X/gfYDznyuwhM/jiA6+btRW+hyPQTbR1pFsn/NGmbhugxeXb6d5IwjqEL9OiFc1K4JYy9tyyUdmtI+up5NwW3OLGMNzLrFab8fs8hpwll0H3zyIGxaBEOfh8Ztz+0cqrDrO6cZafOHTmfwebdAv3ugeffSXx8a7lzRVPZVjbFO6ppk2ZZsxsxYSaeYBgzp3oJLOzbh/LgoQoJtyi3jgy2fwbwxUK8JjHoHmnZ0tqs63/A/nQxFLhj4BCSOLfvVRF6OkxBWToc9a5y+hcSx0OfugGx+qQmsk7qW2H0wl0lz1tKxWQPe/uXF1Ktjb60pg7WzYNEkiOkKty04/QNbBHrdDu2vdK4mPvqtM1Rz6HPQKP7sx83LgR8+cTqe074Adz40bg8/+wecP7JsI5RMlfPpK4CIDBaRVBFJE5GHStjfSETeFZENIrJSRLp7tieIyDqvnxwRud+zr7GIfCYiWzy/G3kd72HPuVJF5JoKqmuNlVvg4u6ZSRQVKdNH97bkYHynCkv/Bgt/BW0vgzs+OvO3+cg4GPU2/PzfsGcdvHAxrHoFiopOL5d/FDbMhzkj4W8d4J27nXsLLhwLYz+De5OcqSQsOVR7pTYxiUgw8AMwEEgHVgEjVDXFq8zfgGOq+icR6Qw8r6oDSjhOBtBXVXeKyDPAQVWd4kk6jVT1QRHpCswB+gAtgc+BTqpnnrmqNjcxqSqT5qzlw+RMXrvjQvrbgj3GV0Vu+Pj3zof8ecOc/oWQMN9ee3i3c8WxbQm0vRwGPw37UpwrhS2fOVcKDVo69yF0u9G5UayiO7jNOauIJqY+QJqqbvMccC4wFEjxKtMVeApAVTeLSLyIxKhqlleZAcBWVd3peT4U6O95/AbwJfCgZ/tcVc0HtotImieGb32ItdaZvnQbH2zI5HfXJFhyML4rPAFvj4PNH8Alv4YBj5ftAzyqFdz+Lqx+HRY/Ci/2c7Y3aAGJd3qSQh9LCgHOlwQRC+z2ep4O9C1WZj1wE7BcRPoAbYA4wDtBDMe5MjgpRlUzAVQ1U0ROfrrFAt8VO19s8aBEZDwwHqB169Y+VKPmWbYlm6c/2cy15zXnV/1tRIfxUe5BmDMCdq9wvvlfNLF8xxFxkkGHAfD9O850Fq36WlKoQXx5J0saE1m8XWoK0EhE1gGTgLWA69QBRMKA64H5FXQ+VHW6qiaqamJ0dLQPh61Zdh3I5d7ZTqf0324+34auGt8c3g0zBjujiG55rfzJwVtUa7j0fmjTz5JDDePLFUQ60MrreRywx7uAquYAdwKI80m13fNz0hBgTbEmpywRaeG5emgB7PP1fLVdboGL8W8moWqd0qYM9n4Ps26GglyneSj+Un9HZKo5X9L9KqCjiLT1XAkMBxZ5FxCRKM8+gHHAUk/SOGkEpzcv4TnGGM/jMcBCr+3DRaSOiLQFOgIrfa1QTaeq/H7BBlKzjvLvERfQpkngzOti/GjbV/DaEGcm07s+seRgfFLqV09VdYnIvcCnQDAwQ1U3ishEz/5pQBdgpoi4cTqvx558vYhE4IyAmlDs0FOAeSIyFtgF3OI53kYRmec5jgu452wjmGqbk53Svx9sndLGR8kL4N2J0KQDjFpQLWYRNYHB7qQOICfvlB7SvQXPjbzA+h1M6b593rkDus0lMHyWc/eyMR52J3U1tnBdBv/bvI/mkXVp0bAuLaLCaRkZTvPIujSpF0ZQ0I8JwLtT+pmbe1hyMGdXVASf/cGZ76jrULhxOoTaTL2mbCxB+MmqHQf5v7fWERkeyvF8NwXu0+9GDQsOIiayDi0iw2kRWdcz+Z51ShsfuPLhvV/C92876yIMfqraLWVpAoN90vjBkdxC7p+7jlaNI/jwvsuICA3mwPEC9h7JY8+RE6d+Zx7OY++RPFbvPES+q4j/jOxlndLm7PKOwNzbYMcyuPpPzk1wdrVpyskSRBVTVSa/m0xWTh5v//Ji6nuuBqIb1CG6QR3Oi6uGC50Y/zhxGFI/dha/ObkOckSTM3/g52Q6w1izN8ONL8H5w6s0XFPzWIKoYvOSdvNhciYPDu7M+a2i/B2OqY7chZA0A76cAicOnr4vuM6PS216L8EZ3gg++6OzMttt86H9Vf6J3dQoliCqUNq+Yzy+KIVLOjRhwuXt/B1O4Chy+68NXdX5qYo7hFWd9RI+ewwObnUmwbvyUWeVtFPLbqb/+HjnN3B0j7NGA0C9ZnDHh9CyZ+XHamoFSxBVJN/l5r45a6kbGsQ/h/U8bYSSOYsN8+DD38KI2VV7c1dRESTPh//9GQqOQ5frnAno4i+vnGUtM9bA4j/AzuXOeswj50HHQT82J8X2PkOcbjie/eOymxGNKz42U2tZgqgiz3ySSkpmDq+MTiSmoQ039EnaF85onCIXfPIQjP+qaq4kdnwNix+BPWuhxfnQ+iJnMro1M50+gC4/h643QPxl554sjqTDF0/AhrcgoqmzkE6vO3w/blCws36DrchmKoEliCrwZeo+Xl2+nTH92nB11xh/hxMY9qyFeaMhujMk3gUfPgDr58IFt1XeOfenwed/dKbAbhjrdPSeN8xpXio84SSsje86i+Gsft2TLK53rizaXFK2ZJF/FJb/y7mRTRUufQAu/T+o27DSqmdMWdmd1JVs39E8rp26jKb16/DePZdQN9TGo5fq4DZ4dRCEhMPYxc6341eudr5t37cGwip4qG/uQfjqaWfhnJC6zsykF91z5hXPCk9A2udOskj9BAqPO9/+Ow12RhyVpsgFG99xmobOGwYDHnPWVzCmitmd1H5UVKT8Zt56jua5mHP3RZYcfHEsG/77C+dDdNTb0LCFs/2av8KMQfDNf6D/T1a9LR9XPqx4CZb+HQqOQq/R0H8yNCjlKi803Glm6vJzZ2bUtM9g43uQ+pEzAskXLXvCyLfO3LdgTDVgCaISvbp8O8u27OfJG7rTMaaBv8Op/vKPwexhznj+MYsgutOP+1r3daaM+Hoq9BrzY+IoD1VIec8ZFnp4J3QYCIP+DM26lP1YYRFOXF2Hlj8eY6opW92jkiSnH+GZTzczqGsMt/WtnSvelYm7EOaPgcx1zkI2rfr8tMzVjztXFv978tzO9c1/YP4dEFbfWRdh1ILyJQdjajhLEJXgeL6L++aupUm9Ojz9C5tYr1SqsOg+p13/umchYUjJ5Rq3gz7jYd0syNxQvnNt+dzpiO56A0xcZjeUGXMWliAqmKry+KKN7DhwnH/d2pNG9cJKf1Ft98UTsH429H8Yeo85e9nLf+fcNbz4ESexlMX+NFhwFzTrBje8YBPYGVMK64OoAIXuIlZuP8jijXtZnJJF5pE87r2yA/3aN/F3aNXfiumw/J/Q+w644sHSy4dHOZ3UH/8efvgUEgb7dp68IzBnuDMUdcTsih8JZUwNZAminI7nu1j6QzaLU7L4YlMWOXku6oQEcXmnaH47KIEbLoj1d4jV38b3nA/6hGvh2n/4Puto4l2wcrqz3kGHARAcevbyRW54exwc2g6jF0KU9QkZ4wtLEGWw/1g+X2zKYvHGLJal7afAVURURCgDuzZnULcYLuvYlIgw+yf1yY6v4Z3xTmf0L14t201mwaEw8M8wd4Rzw1qfu89e/n9/hi2LnbuUbS1mY3xmn2ZnoaqkZOawZPM+lqRms2bXIVQhNiqc2/q2ZlDX5lwY34iQYOvKKZOcPTDvdmjUBkbMPfMNaWeTMMSZ6uLLp6DHsDPfoJa8wLljufedcOG4c4vbmFrGEkQxR/MK+TptP0s2Z7MkdR/7juYD0CMukvuu6sigbjF0bdHQRiaVl9sFC8ZCYR7cOqv8k8uJwKAnYXp/WPYPGPjET8vsWQsL74HWF8OQZ84pbGNqI58ShIgMBqYCwcArqjql2P5GwAygPZAH3KWq33v2RQGvAN0B9ez7VkTeAhI8h4gCDqtqTxGJBzYBqZ5936nqxPJWsDSqytbsY6cSwqodByl0Kw3qhHB5p2j6J0RzRUI0zRrYBHsV4sunYNc3zjxH3jfClUfLnnD+CPjuRadfolH8j/uO7XNWVqsXDcNmQoiNJjOmrEpNECISDDwPDATSgVUiskhVU7yKTQbWqeqNItLZU36AZ99U4BNVvVlEwoAIAFW91esc/wCOeB1vq6r2LH+1fLN+92Humb2G9EMnAEiIacBdl7blyoRm9G7TiFBrOqpYaV843/YvGFVxq50N+IMzJ9Lnf3JusANnCo23bnfmWBr7KdSPrphzGVPL+HIF0QdIU9VtACIyFxgKeCeIrsBTAKq6WUTiRSQGOAFcDtzh2VcAFHgfXJy2mmFAld+x1KpxBJ2bN2DiFe25snMzYqPCqzqE2iMn0+mUju4MQ/5Wccdt2BIungRLn4GLfgVxifDRb2H3d3DzDGe6bmNMufjyFTkW2O31PN2zzdt64CYAEekDtAHigHZANvCaiKwVkVdEpPgA9MuALFXd4rWtraf8VyJyWUlBich4EUkSkaTs7GwfqvFTjeuF8cqYCxl1URtLDpWpyA3v3A2FuXDL6+XrlD6bS34N9WPg08mw8mVn3YbLfgPdf1Gx5zGmlvElQZTUG1v8FtYpQCMRWQdMAtYCLpwrlF7Ai6p6AXAcKD4V5whgjtfzTKC1p/wDwGwR+ckk+ao6XVUTVTUxOtqaEKq1r56GHcvg2r9Ds84Vf/w69eGqRyF9JXz8O2fa7SsfrfjzGFPL+NLElA54T1YfB+zxLqCqOcCdcKrJaLvnJwJIV9UVnqIL8EoQIhKCc+XR2+tY+UC+5/FqEdkKdAKq54IP5uy2fQlfPQPnj6zcxX563gZJM5zRUTe9XDVrSBtTw/mSIFYBHUWkLZABDAdGehfwjFTK9fQxjAOWepJGjojsFpEEVU3F6bj27ru4Gtisqulex4oGDqqqW0TaAR2BbeWuofGfo1nw9t3QtBP87O+Ve66gYLjzY5BgG7FkTAUpNUGoqktE7gU+xRnmOkNVN4rIRM/+aUAXYKaIuHESwFivQ0wCZnlGMG3Dc6XhMZzTm5fA6dR+QkRcgBuYqKoHy1U74z9FbnhnnLO05uiFVTP3Uaj1IxlTkWzJUVM5vnwavvwrXP8fZ6U2Y0y1U9qSo9ZQayre9mXw1RTocStccLu/ozHGlJMlCFOxjmU7M6c2bg8/+6fvM7QaY6odm4vJnJ0q7EuBQzt8K7/iJcg7DKPedoafGmMCliUI81Mnk8LGd52fA2lle/3Pp0Lz7pUTmzGmyliCMA5V2LfJSQgp78H+H0CCnCm1+90DLXv51lxUp4GzdrQxJuBZgqiuju2D3Sudu4N3r3JWQ+s7AfpNKtviOqXZt8lZ2W3ju7A/1UkKbS6BvhOhy/U20Z0xtZgliOrA7YKs770Swko4vNPZFxTqTDjXtBN8/jhs+gBuePHcpspWhY3vwNK/O01JiLPSWt/xnqTQrCJqZYwJcJYg/GnT+/DdNNizxpnIDqB+c2h1obOMZlwfJzmE1v3xQ/3D38K0S+GqR6Dfvc4dxGWxa4UzqV1GEjTr5syP1OV6aBBT8fUzxgQ0SxD+cmCrMxy0YaxzI1nchc76zJGtSm7rF3FmJ42/DD74P/jsMc/VxAvQtGPp5zu4zbkCSVnoJKGhzzuL7ZQ1wRhjag1LEP6gCu//GoLD4I4PoWEL319bvxnc+l9nreWPf+e5mnjUWQuhpA/7E4ecpqQVL0FwKPR/2Fk/oSqmvjDGBDRLEP6w9r/O9NfXPVu25HCSCPS4Bdpe7lxNLH7Uaa4a+gI07eCUcRVA0qvOVNsnDjszqV75aPnOZ4yplSxBVLWjWbD4EWekUK8x53asBjEwfBZsmAcf/x6mXQIDHnOaqT7/o9Os1O5KGPSk3ZdgjCkzSxBV7ePfO2sW/HxqxaxZIALn3wrtroD373c6oMFZ2vO2BdDhapvuwhhTLpYgqtLmD52b0K76g28dy2XRoDmMmOOMdCo8AT2GV+z9EsaYWsc+QapKXo4zRLVZN2cN5cpwcqSTMcZUAEsQVeWLP8Gxvc4IpOBQf0djjDGlsum+q8Ku72DVK9D3lxDXu/TyxhhTDViCqGyufFg0CSJbw5WT/R2NMcb4zJqYKtuyfzgzo95m6yMYYwKLXUFUpqwUWPZPZ+nNjlf7OxpjjCkTnxKEiAwWkVQRSRORh0rY30hE3hWRDSKyUkS6e+2LEpEFIrJZRDaJSD/P9sdFJENE1nl+rvV6zcOec6WKyDUVUdEqV+R2mpbqNoRrnvJ3NMYYU2alNjGJSDDwPDAQSAdWicgiVU3xKjYZWKeqN4pIZ0/5AZ59U4FPVPVmEQkDIrxe9y9V/Xux83UFhgPdgJbA5yLSSVXd5auin6x6xZkx9aaXoV4Tf0djjDFl5ssVRB8gTVW3qWoBMBcYWqxMV+ALAFXdDMSLSIyINAQuB1717CtQ1cOlnG8oMFdV81V1O5DmiSFwHN4Nn//JuYv5vFv8HY0xxpSLLwkiFtjt9Tzds83beuAmABHpA7QB4oB2QDbwmoisFZFXRMR7GtF7Pc1SM0SkURnOh4iMF5EkEUnKzs72oRpVxO2CDx9wHl/3L5vmwhgTsHxJECV9wmmx51OARiKyDpgErAVcOE1YvYAXVfUC4Dhwsg/jRaA90BPIBP5RhvOhqtNVNVFVE6Oj/bwsptsF275y5kL6RwJsWQwD/gBRrf0blzHGnANfhrmmA628nscBe7wLqGoOcCeAiAiw3fMTAaSr6gpP0QV4EoSqZp18vYi8DHzg6/mqhSI37PzaWc950yI4ng2hEdDpGme6i87X+TtCY4w5J74kiFVARxFpC2TgdCCP9C4gIlFArqePYhyw1JM0ckRkt4gkqGoqTsd1iuc1LVQ103OIG4HvPY8XAbNF5J84ndQdgZXnUMeKU+SGnd84E+6lLILj+35MCl1vgI6DICyitKMYY0xAKDVBqKpLRO4FPgWCgRmqulFEJnr2TwO6ADNFxI2TAMZ6HWISMMszgmkbnisN4BkR6YnTfLQDmOA53kYRmec5jgu4p1qMYFr5Mnz1jJMUQsKdpNDtBk9SsNXZjDE1j6j+pHk/4CQmJmpSUlLlnWDT+/DWKGc96MS7nORgScEYE+BEZLWqJp5pv021UZqsjfDOBIhNdBbgCa3r74iMMaZK2FQbZ3P8AMwZAXUaONN0W3IwxtQidgVxJu5CmD8Gju6FOz+Chi38HZExxlQpSxBn8ukjsGMZ3DAN4s7YRGeMMTWWNTGVZM1MWPkSXHQP9Bzh72iMMcYvLEEUt2sFfPAAtLsSBj7h72iMMcZvLEF4O5LuDGeNjIObZ0CwtcAZY2ov+wQ8qfAEzL0NCnNhzPsQ0djfERljjF9ZggBQdRb3yVwPw2dDs87+jsgYY/zOmpgAvvk3JM+Hqx6BzteWXt4YY2oBSxBbPoPP/uhMtnfZb/0djTHGVBu1O0Hs3wILxkLz7nDDC7a4jzHGeKndCSKkDsT2cvodbPI9Y4w5Te3upI5qDaPf83cUxhhTLdXuKwhjjDFnZAnCGGNMiSxBGGOMKZElCGOMMSWyBGGMMaZEliCMMcaUyBKEMcaYElmCMMYYUyJRVX/HcM5EJBvYeQ6HaArsr6BwqgOrT/VX0+pU0+oDNa9OJdWnjapGn+kFNSJBnCsRSVLVGrPwtNWn+qtpdapp9YGaV6fy1MeamIwxxpTIEoQxxpgSWYJwTPd3ABXM6lP91bQ61bT6QM2rU5nrY30QxhhjSmRXEMYYY0pkCcIYY0yJanWCEJHBIpIqImki8pC/46kIIrJDRJJFZJ2IJPk7nrISkRkisk9Evvfa1lhEPhORLZ7fjfwZY1mdoU6Pi0iG531aJyLX+jPGshCRViKyREQ2ichGEfm1Z3tAvk9nqU8gv0d1RWSliKz31OlPnu1leo9qbR+EiAQDPwADgXRgFTBCVVP8Gtg5EpEdQKKqBuQNPiJyOXAMmKmq3T3bngEOquoUTyJvpKoP+jPOsjhDnR4Hjqnq3/0ZW3mISAughaquEZEGwGrgBuAOAvB9Okt9hhG475EA9VT1mIiEAsuBXwM3UYb3qDZfQfQB0lR1m6oWAHOBoX6OqdZT1aXAwWKbhwJveB6/gfPHGzDOUKeApaqZqrrG8/gosAmIJUDfp7PUJ2Cp45jnaajnRynje1SbE0QssNvreToB/p/CQ4HFIrJaRMb7O5gKEqOqmeD8MQPN/BxPRblXRDZ4mqACojmmOBGJBy4AVlAD3qdi9YEAfo9EJFhE1gH7gM9UtczvUW1OEFLCtprQ3naJqvYChgD3eJo3TPXzItAe6AlkAv/wazTlICL1gbeB+1U1x9/xnKsS6hPQ75GqulW1JxAH9BGR7mU9Rm1OEOlAK6/nccAeP8VSYVR1j+f3PuBdnKa0QJflaSc+2V68z8/xnDNVzfL8ARcBLxNg75OnXfttYJaqvuPZHLDvU0n1CfT36CRVPQx8CQymjO9RbU4Qq4COItJWRMKA4cAiP8d0TkSknqeTDRGpBwwCvj/7qwLCImCM5/EYYKEfY6kQJ/9IPW4kgN4nTwfoq8AmVf2n166AfJ/OVJ8Af4+iRSTK8zgcuBrYTBnfo1o7ignAM2ztWSAYmKGqf/FvROdGRNrhXDUAhACzA61OIjIH6I8zNXEW8EfgPWAe0BrYBdyiqgHT6XuGOvXHabpQYAcw4WTbcHUnIpcCy4BkoMizeTJOu33AvU9nqc8IAvc96oHTCR2McyEwT1WfEJEmlOE9qtUJwhhjzJnV5iYmY4wxZ2EJwhhjTIksQRhjjCmRJQhjjDElsgRhjDGmRJYgjDHGlMgShDHGmBL9P0mjnqxhlm6AAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(a[\"accuracy\"], label=\"train\")\n",
    "plt.plot(a[\"val_accuracy\"], label=\"valid\")\n",
    "plt.legend(fontsize='x-large')\n",
    "plt.show()\n",
    "# train is always good, valid : little bit flat (stop or not)\n",
    "# finally val_accuracy : 97%"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predictive analysis in Test data set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1f802108888>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[0,:,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 0., 0., 0., 1., 0., 0., 0., 0.])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train[0, :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1.1944930e-07, 1.9433463e-07, 2.7127862e-05, 1.3849593e-02,\n",
       "       1.8362323e-12, 9.8612261e-01, 3.4668182e-11, 4.4255081e-08,\n",
       "       3.5707373e-08, 4.0594156e-07], dtype=float32)"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.predict(x_train)\n",
    "y_pred[0,:]\n",
    "# 99% 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(50000, 10)\n"
     ]
    }
   ],
   "source": [
    "print(y_pred.shape) # dimension"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_class = np.argmax(y_pred, axis=1) # which is best"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_class[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1f87fdf1c08>"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_test[0,:,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([6.4372296e-05, 6.6478299e-07, 3.6091596e-04, 1.3369783e-03,\n",
       "       2.3672527e-07, 4.2162487e-06, 2.1588240e-10, 9.9818891e-01,\n",
       "       8.8055622e-06, 3.4845907e-05], dtype=float32)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.predict(x_test)\n",
    "y_pred[0,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_class = np.argmax(y_pred, axis=1) # axis not 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7\n"
     ]
    }
   ],
   "source": [
    "print(y_pred_class[0]) # test data ..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9718"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(y_test == y_pred_class)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Which figure went wrong?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "aa = (y_test == y_pred_class)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True,  True,  True,  True,  True,  True,  True,  True, False,\n",
       "        True])"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "aa[0:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "bb = np.arange(0, 10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "cc = bb[aa==False]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x27fafe98bc8>"
      ]
     },
     "execution_count": 158,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2, ax3, ax4) = plt.subplots(1, 4)\n",
    "fig.suptitle('Image of Wrong Classifications')\n",
    "ax1.imshow(x_test[cc[0],])\n",
    "ax2.imshow(x_test[cc[1],])\n",
    "ax3.imshow(x_test[cc[2],])\n",
    "ax4.imshow(x_test[cc[3],])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 첫번째 그림 maybe 6? 5의 확률도 15% 정도 존재"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[6 9 2 0] [5 4 4 6]\n"
     ]
    }
   ],
   "source": [
    "print(y_pred_class[cc[0:4],], y_test[cc[0:4]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Image of Correct Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x27fb2900dc8>"
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2, ax3, ax4) = plt.subplots(1, 4)\n",
    "fig.suptitle('Image of Correct Classifications')\n",
    "ax1.imshow(x_test[0,])\n",
    "ax2.imshow(x_test[1,])\n",
    "ax3.imshow(x_test[2,])\n",
    "ax4.imshow(x_test[3,])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[7 2 1 0] [7 2 1 0]\n"
     ]
    }
   ],
   "source": [
    "print(y_pred_class[0:4,], y_test[0:4])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Multinomial Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [],
   "source": [
    "m_logistic_model = tf.keras.models.Sequential([\n",
    "  tf.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tf.keras.layers.Dense(10, activation=\"softmax\")\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten_5 (Flatten)          (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 10)                7850      \n",
      "=================================================================\n",
      "Total params: 7,850\n",
      "Trainable params: 7,850\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "m_logistic_model.compile(loss=\"categorical_crossentropy\", optimizer=\"sgd\", metrics=[\"accuracy\"])\n",
    "m_logistic_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 1.3955 - accuracy: 0.6674 - val_loss: 0.8738 - val_accuracy: 0.8407\n",
      "Epoch 2/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.7832 - accuracy: 0.8318 - val_loss: 0.6371 - val_accuracy: 0.8656\n",
      "Epoch 3/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.6360 - accuracy: 0.8518 - val_loss: 0.5434 - val_accuracy: 0.8771\n",
      "Epoch 4/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.5653 - accuracy: 0.8628 - val_loss: 0.4923 - val_accuracy: 0.8834\n",
      "Epoch 5/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.5222 - accuracy: 0.8694 - val_loss: 0.4589 - val_accuracy: 0.8883\n",
      "Epoch 6/20\n",
      "50000/50000 [==============================] - 1s 22us/sample - loss: 0.4926 - accuracy: 0.8739 - val_loss: 0.4359 - val_accuracy: 0.8913\n",
      "Epoch 7/20\n",
      "50000/50000 [==============================] - 1s 22us/sample - loss: 0.4708 - accuracy: 0.8778 - val_loss: 0.4187 - val_accuracy: 0.8925\n",
      "Epoch 8/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.4538 - accuracy: 0.8805 - val_loss: 0.4048 - val_accuracy: 0.8963\n",
      "Epoch 9/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.4401 - accuracy: 0.8832 - val_loss: 0.3937 - val_accuracy: 0.8997\n",
      "Epoch 10/20\n",
      "50000/50000 [==============================] - 1s 18us/sample - loss: 0.4288 - accuracy: 0.8854 - val_loss: 0.3846 - val_accuracy: 0.9007\n",
      "Epoch 11/20\n",
      "50000/50000 [==============================] - 1s 17us/sample - loss: 0.4192 - accuracy: 0.8875 - val_loss: 0.3770 - val_accuracy: 0.9025\n",
      "Epoch 12/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.4109 - accuracy: 0.8887 - val_loss: 0.3704 - val_accuracy: 0.9038\n",
      "Epoch 13/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.4038 - accuracy: 0.8904 - val_loss: 0.3647 - val_accuracy: 0.9049\n",
      "Epoch 14/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.3974 - accuracy: 0.8917 - val_loss: 0.3595 - val_accuracy: 0.9066\n",
      "Epoch 15/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.3917 - accuracy: 0.8929 - val_loss: 0.3548 - val_accuracy: 0.9075\n",
      "Epoch 16/20\n",
      "50000/50000 [==============================] - 1s 22us/sample - loss: 0.3866 - accuracy: 0.8941 - val_loss: 0.3508 - val_accuracy: 0.9083\n",
      "Epoch 17/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.3821 - accuracy: 0.8946 - val_loss: 0.3471 - val_accuracy: 0.9085\n",
      "Epoch 18/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.3779 - accuracy: 0.8956 - val_loss: 0.3438 - val_accuracy: 0.9093\n",
      "Epoch 19/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.3740 - accuracy: 0.8965 - val_loss: 0.3409 - val_accuracy: 0.9092\n",
      "Epoch 20/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.3704 - accuracy: 0.8972 - val_loss: 0.3378 - val_accuracy: 0.9099\n"
     ]
    }
   ],
   "source": [
    "m_logistic_history =  m_logistic_model.fit(x_train, y_train, epochs=20, validation_data=(x_valid, y_valid), batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 더 이상 val_accuracy의 값이 증가하지 않음\n",
    "# 나쁘진 않지만 neural network보다는 성능이 떨어짐\n",
    "# tensorflow version of multivariate logistic regression\n",
    "# Why NN is working well?\n",
    "# How do we improve NN?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Hands-on Machine learning ... tensorflow 2nd edition\n",
    "# chapter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "a = m_logistic_history.history\n",
    "plt.plot(a[\"accuracy\"], label=\"train\")\n",
    "plt.plot(a[\"val_accuracy\"], label=\"valid\")\n",
    "plt.legend(fontsize='x-large')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([3.67651257e-04, 2.21080199e-06, 3.16101941e-04, 2.14435393e-03,\n",
       "       7.22445766e-05, 1.18915275e-04, 5.65385335e-06, 9.93860602e-01,\n",
       "       2.55093008e-04, 2.85719242e-03], dtype=float32)"
      ]
     },
     "execution_count": 172,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_logistic = m_logistic_model.predict(x_test)\n",
    "y_pred_logistic[0,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_logistic_class = np.argmax(y_pred_logistic, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9071"
      ]
     },
     "execution_count": 174,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(y_test == y_pred_logistic_class)"
   ]
  }
 ],
 "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}