very-simple-cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "waiting-cover",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:20.896624Z",
     "iopub.status.busy": "2021-05-14T09:47:20.894950Z",
     "iopub.status.idle": "2021-05-14T09:47:20.905025Z",
     "shell.execute_reply": "2021-05-14T09:47:20.904420Z"
    },
    "papermill": {
     "duration": 0.02639,
     "end_time": "2021-05-14T09:47:20.905183",
     "exception": false,
     "start_time": "2021-05-14T09:47:20.878793",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# This Python 3 environment comes with many helpful analytics libraries installed\n",
    "# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n",
    "# For example, here's several helpful packages to load\n",
    "\n",
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "\n",
    "# Input data files are available in the read-only \"../input/\" directory\n",
    "# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n",
    "\n",
    "import os\n",
    "for dirname, _, filenames in os.walk('/kaggle/input'):\n",
    "    for filename in filenames:\n",
    "        print(os.path.join(dirname, filename))\n",
    "\n",
    "# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n",
    "# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "native-fence",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:20.929787Z",
     "iopub.status.busy": "2021-05-14T09:47:20.929265Z",
     "iopub.status.idle": "2021-05-14T09:47:25.951492Z",
     "shell.execute_reply": "2021-05-14T09:47:25.950957Z"
    },
    "papermill": {
     "duration": 5.035078,
     "end_time": "2021-05-14T09:47:25.951625",
     "exception": false,
     "start_time": "2021-05-14T09:47:20.916547",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "narrative-proposal",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:25.978017Z",
     "iopub.status.busy": "2021-05-14T09:47:25.977504Z",
     "iopub.status.idle": "2021-05-14T09:47:26.731135Z",
     "shell.execute_reply": "2021-05-14T09:47:26.730500Z"
    },
    "papermill": {
     "duration": 0.769098,
     "end_time": "2021-05-14T09:47:26.731280",
     "exception": false,
     "start_time": "2021-05-14T09:47:25.962182",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
      "11493376/11490434 [==============================] - 0s 0us/step\n"
     ]
    }
   ],
   "source": [
    "(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "organic-danger",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:26.762740Z",
     "iopub.status.busy": "2021-05-14T09:47:26.761946Z",
     "iopub.status.idle": "2021-05-14T09:47:26.767798Z",
     "shell.execute_reply": "2021-05-14T09:47:26.767370Z"
    },
    "papermill": {
     "duration": 0.023756,
     "end_time": "2021-05-14T09:47:26.767907",
     "exception": false,
     "start_time": "2021-05-14T09:47:26.744151",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10000, 28, 28, 1)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train = x_train.reshape(-1,28,28,1)\n",
    "x_test = x_test.reshape(-1,28,28,1)\n",
    "\n",
    "x_train.shape\n",
    "x_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "damaged-rates",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:26.795511Z",
     "iopub.status.busy": "2021-05-14T09:47:26.794610Z",
     "iopub.status.idle": "2021-05-14T09:47:26.861355Z",
     "shell.execute_reply": "2021-05-14T09:47:26.860878Z"
    },
    "papermill": {
     "duration": 0.082325,
     "end_time": "2021-05-14T09:47:26.861489",
     "exception": false,
     "start_time": "2021-05-14T09:47:26.779164",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "x_train = np.pad(x_train, ((0,0),(2,2),(2,2),(0,0)), 'constant')\n",
    "x_test = np.pad(x_test, ((0,0),(2,2),(2,2),(0,0)), 'constant')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "assured-graph",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:26.895462Z",
     "iopub.status.busy": "2021-05-14T09:47:26.894722Z",
     "iopub.status.idle": "2021-05-14T09:47:26.897449Z",
     "shell.execute_reply": "2021-05-14T09:47:26.897024Z"
    },
    "papermill": {
     "duration": 0.018776,
     "end_time": "2021-05-14T09:47:26.897557",
     "exception": false,
     "start_time": "2021-05-14T09:47:26.878781",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "img_w = x_train[0].shape[0]\n",
    "img_h = x_test[0].shape[1]\n",
    "num_channels = 1\n",
    "seed = 98\n",
    "\n",
    "np.random.seed(seed)\n",
    "tf.random.set_seed(seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "commercial-tunisia",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:26.924609Z",
     "iopub.status.busy": "2021-05-14T09:47:26.923491Z",
     "iopub.status.idle": "2021-05-14T09:47:27.103313Z",
     "shell.execute_reply": "2021-05-14T09:47:27.102837Z"
    },
    "papermill": {
     "duration": 0.194276,
     "end_time": "2021-05-14T09:47:27.103446",
     "exception": false,
     "start_time": "2021-05-14T09:47:26.909170",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "x_train = x_train/255\n",
    "x_test = x_test/255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "personal-administrator",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:27.130860Z",
     "iopub.status.busy": "2021-05-14T09:47:27.130359Z",
     "iopub.status.idle": "2021-05-14T09:47:27.133602Z",
     "shell.execute_reply": "2021-05-14T09:47:27.134416Z"
    },
    "papermill": {
     "duration": 0.0192,
     "end_time": "2021-05-14T09:47:27.134543",
     "exception": false,
     "start_time": "2021-05-14T09:47:27.115343",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "batch_size = 100\n",
    "eval_size = 500\n",
    "\n",
    "epochs = 300\n",
    "eval_every = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "closing-physics",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:27.162364Z",
     "iopub.status.busy": "2021-05-14T09:47:27.161743Z",
     "iopub.status.idle": "2021-05-14T09:47:27.174832Z",
     "shell.execute_reply": "2021-05-14T09:47:27.174412Z"
    },
    "papermill": {
     "duration": 0.028461,
     "end_time": "2021-05-14T09:47:27.174932",
     "exception": false,
     "start_time": "2021-05-14T09:47:27.146471",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "input_data = tf.keras.Input(dtype=tf.float32, shape=(img_w, img_h, num_channels), name='Input')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "separated-cross",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:27.212197Z",
     "iopub.status.busy": "2021-05-14T09:47:27.211526Z",
     "iopub.status.idle": "2021-05-14T09:47:29.292305Z",
     "shell.execute_reply": "2021-05-14T09:47:29.291767Z"
    },
    "papermill": {
     "duration": 2.105729,
     "end_time": "2021-05-14T09:47:29.292436",
     "exception": false,
     "start_time": "2021-05-14T09:47:27.186707",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "#conv-relu-pooling\n",
    "\n",
    "conv1 = tf.keras.layers.Conv2D(filters = 6,\n",
    "                       kernel_size= 5,\n",
    "                       padding = 'VALID',\n",
    "                       activation = \"relu\",\n",
    "                       name = \"c1\"\n",
    "                      )(input_data)\n",
    "\n",
    "max_pool1 = tf.keras.layers.MaxPool2D(pool_size=2,\n",
    "                                     strides = 2,\n",
    "                                     padding = \"SAME\",\n",
    "                                     name = \"s1\")(conv1)\n",
    "\n",
    "conv2 = tf.keras.layers.Conv2D(filters = 16,\n",
    "                       kernel_size= 5,\n",
    "                       padding = 'VALID',\n",
    "                       strides = 1,        \n",
    "                       activation = \"relu\",\n",
    "                       name = \"c3\"\n",
    "                      )(max_pool1)\n",
    "\n",
    "max_pool2 = tf.keras.layers.MaxPool2D(pool_size=2,\n",
    "                                     strides = 2,\n",
    "                                     padding = \"SAME\",\n",
    "                                     name = \"s4\")(conv2)\n",
    "\n",
    "flatten = tf.keras.layers.Flatten(name=\"FLATTEN\")(max_pool2)\n",
    "\n",
    "fc1 = tf.keras.layers.Dense(units = 120,\n",
    "                           activation = \"relu\",\n",
    "                           name = \"f5\")(flatten)\n",
    "\n",
    "fc2 = tf.keras.layers.Dense(units = 84,\n",
    "                           activation = \"relu\",\n",
    "                           name=\"f6\")(fc1)\n",
    "\n",
    "final_layer = tf.keras.layers.Dense(units = 10,\n",
    "                                   activation = \"softmax\",\n",
    "                                   name = \"output\")(fc2)\n",
    "\n",
    "model = tf.keras.Model(inputs = input_data, outputs = final_layer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "mediterranean-violin",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:29.327650Z",
     "iopub.status.busy": "2021-05-14T09:47:29.326816Z",
     "iopub.status.idle": "2021-05-14T09:47:29.338775Z",
     "shell.execute_reply": "2021-05-14T09:47:29.337768Z"
    },
    "papermill": {
     "duration": 0.033629,
     "end_time": "2021-05-14T09:47:29.338894",
     "exception": false,
     "start_time": "2021-05-14T09:47:29.305265",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "Input (InputLayer)           [(None, 32, 32, 1)]       0         \n",
      "_________________________________________________________________\n",
      "c1 (Conv2D)                  (None, 28, 28, 6)         156       \n",
      "_________________________________________________________________\n",
      "s1 (MaxPooling2D)            (None, 14, 14, 6)         0         \n",
      "_________________________________________________________________\n",
      "c3 (Conv2D)                  (None, 10, 10, 16)        2416      \n",
      "_________________________________________________________________\n",
      "s4 (MaxPooling2D)            (None, 5, 5, 16)          0         \n",
      "_________________________________________________________________\n",
      "FLATTEN (Flatten)            (None, 400)               0         \n",
      "_________________________________________________________________\n",
      "f5 (Dense)                   (None, 120)               48120     \n",
      "_________________________________________________________________\n",
      "f6 (Dense)                   (None, 84)                10164     \n",
      "_________________________________________________________________\n",
      "output (Dense)               (None, 10)                850       \n",
      "=================================================================\n",
      "Total params: 61,706\n",
      "Trainable params: 61,706\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.compile(\n",
    "    optimizer = \"adam\",\n",
    "    loss = \"sparse_categorical_crossentropy\",\n",
    "    metrics = [\"accuracy\"]\n",
    ")\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "painful-confidentiality",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:29.371524Z",
     "iopub.status.busy": "2021-05-14T09:47:29.370704Z",
     "iopub.status.idle": "2021-05-14T09:47:40.803304Z",
     "shell.execute_reply": "2021-05-14T09:47:40.803855Z"
    },
    "papermill": {
     "duration": 11.452657,
     "end_time": "2021-05-14T09:47:40.804050",
     "exception": false,
     "start_time": "2021-05-14T09:47:29.351393",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16/16 [==============================] - 0s 3ms/step - loss: 2.1737 - accuracy: 0.3440\n",
      "Epoch # 5. Train Loss: 2.19. Train Acc (Test Acc): 0.23 (0.34)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 1.9759 - accuracy: 0.5800\n",
      "Epoch # 10. Train Loss: 2.01. Train Acc (Test Acc): 0.60 (0.58)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 1.6057 - accuracy: 0.7300\n",
      "Epoch # 15. Train Loss: 1.71. Train Acc (Test Acc): 0.74 (0.73)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 1.2110 - accuracy: 0.7740\n",
      "Epoch # 20. Train Loss: 1.32. Train Acc (Test Acc): 0.73 (0.77)\n",
      "16/16 [==============================] - 0s 4ms/step - loss: 0.8708 - accuracy: 0.7980\n",
      "Epoch # 25. Train Loss: 0.93. Train Acc (Test Acc): 0.76 (0.80)\n",
      "16/16 [==============================] - 0s 3ms/step - loss: 0.6734 - accuracy: 0.7880\n",
      "Epoch # 30. Train Loss: 0.82. Train Acc (Test Acc): 0.76 (0.79)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.5094 - accuracy: 0.8480\n",
      "Epoch # 35. Train Loss: 0.60. Train Acc (Test Acc): 0.76 (0.85)\n",
      "16/16 [==============================] - 0s 3ms/step - loss: 0.5499 - accuracy: 0.8120\n",
      "Epoch # 40. Train Loss: 0.40. Train Acc (Test Acc): 0.92 (0.81)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.4900 - accuracy: 0.8520\n",
      "Epoch # 45. Train Loss: 0.50. Train Acc (Test Acc): 0.81 (0.85)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.3852 - accuracy: 0.8740\n",
      "Epoch # 50. Train Loss: 0.49. Train Acc (Test Acc): 0.82 (0.87)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.3620 - accuracy: 0.8880\n",
      "Epoch # 55. Train Loss: 0.35. Train Acc (Test Acc): 0.87 (0.89)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.3570 - accuracy: 0.8760\n",
      "Epoch # 60. Train Loss: 0.37. Train Acc (Test Acc): 0.86 (0.88)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.3760 - accuracy: 0.8880\n",
      "Epoch # 65. Train Loss: 0.38. Train Acc (Test Acc): 0.90 (0.89)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.4148 - accuracy: 0.8980\n",
      "Epoch # 70. Train Loss: 0.49. Train Acc (Test Acc): 0.88 (0.90)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.3311 - accuracy: 0.9180\n",
      "Epoch # 75. Train Loss: 0.39. Train Acc (Test Acc): 0.87 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2846 - accuracy: 0.9060\n",
      "Epoch # 80. Train Loss: 0.45. Train Acc (Test Acc): 0.87 (0.91)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.3537 - accuracy: 0.9000\n",
      "Epoch # 85. Train Loss: 0.30. Train Acc (Test Acc): 0.94 (0.90)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2844 - accuracy: 0.9240\n",
      "Epoch # 90. Train Loss: 0.41. Train Acc (Test Acc): 0.86 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2715 - accuracy: 0.9280\n",
      "Epoch # 95. Train Loss: 0.28. Train Acc (Test Acc): 0.92 (0.93)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2842 - accuracy: 0.9200\n",
      "Epoch # 100. Train Loss: 0.20. Train Acc (Test Acc): 0.94 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2807 - accuracy: 0.9140\n",
      "Epoch # 105. Train Loss: 0.16. Train Acc (Test Acc): 0.94 (0.91)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2561 - accuracy: 0.9140\n",
      "Epoch # 110. Train Loss: 0.30. Train Acc (Test Acc): 0.93 (0.91)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1976 - accuracy: 0.9380\n",
      "Epoch # 115. Train Loss: 0.21. Train Acc (Test Acc): 0.92 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2853 - accuracy: 0.9200\n",
      "Epoch # 120. Train Loss: 0.44. Train Acc (Test Acc): 0.87 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2664 - accuracy: 0.9220\n",
      "Epoch # 125. Train Loss: 0.16. Train Acc (Test Acc): 0.96 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1774 - accuracy: 0.9360\n",
      "Epoch # 130. Train Loss: 0.28. Train Acc (Test Acc): 0.91 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2557 - accuracy: 0.9200\n",
      "Epoch # 135. Train Loss: 0.14. Train Acc (Test Acc): 0.96 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2306 - accuracy: 0.9360\n",
      "Epoch # 140. Train Loss: 0.20. Train Acc (Test Acc): 0.94 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2044 - accuracy: 0.9360\n",
      "Epoch # 145. Train Loss: 0.31. Train Acc (Test Acc): 0.93 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2767 - accuracy: 0.9140\n",
      "Epoch # 150. Train Loss: 0.28. Train Acc (Test Acc): 0.91 (0.91)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2212 - accuracy: 0.9300\n",
      "Epoch # 155. Train Loss: 0.32. Train Acc (Test Acc): 0.88 (0.93)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1589 - accuracy: 0.9520\n",
      "Epoch # 160. Train Loss: 0.18. Train Acc (Test Acc): 0.92 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2078 - accuracy: 0.9560\n",
      "Epoch # 165. Train Loss: 0.16. Train Acc (Test Acc): 0.96 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1819 - accuracy: 0.9440\n",
      "Epoch # 170. Train Loss: 0.25. Train Acc (Test Acc): 0.89 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2028 - accuracy: 0.9460\n",
      "Epoch # 175. Train Loss: 0.28. Train Acc (Test Acc): 0.92 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1784 - accuracy: 0.9560\n",
      "Epoch # 180. Train Loss: 0.24. Train Acc (Test Acc): 0.95 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1499 - accuracy: 0.9620\n",
      "Epoch # 185. Train Loss: 0.18. Train Acc (Test Acc): 0.95 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1948 - accuracy: 0.9400\n",
      "Epoch # 190. Train Loss: 0.28. Train Acc (Test Acc): 0.91 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1645 - accuracy: 0.9600\n",
      "Epoch # 195. Train Loss: 0.19. Train Acc (Test Acc): 0.95 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1949 - accuracy: 0.9380\n",
      "Epoch # 200. Train Loss: 0.23. Train Acc (Test Acc): 0.93 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2416 - accuracy: 0.9160\n",
      "Epoch # 205. Train Loss: 0.24. Train Acc (Test Acc): 0.93 (0.92)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1769 - accuracy: 0.9620\n",
      "Epoch # 210. Train Loss: 0.09. Train Acc (Test Acc): 0.97 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1679 - accuracy: 0.9520\n",
      "Epoch # 215. Train Loss: 0.39. Train Acc (Test Acc): 0.90 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1423 - accuracy: 0.9520\n",
      "Epoch # 220. Train Loss: 0.11. Train Acc (Test Acc): 0.97 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1777 - accuracy: 0.9460\n",
      "Epoch # 225. Train Loss: 0.14. Train Acc (Test Acc): 0.95 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1454 - accuracy: 0.9500\n",
      "Epoch # 230. Train Loss: 0.17. Train Acc (Test Acc): 0.93 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1873 - accuracy: 0.9480\n",
      "Epoch # 235. Train Loss: 0.23. Train Acc (Test Acc): 0.94 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1600 - accuracy: 0.9480\n",
      "Epoch # 240. Train Loss: 0.18. Train Acc (Test Acc): 0.96 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1422 - accuracy: 0.9560\n",
      "Epoch # 245. Train Loss: 0.14. Train Acc (Test Acc): 0.95 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1970 - accuracy: 0.9380\n",
      "Epoch # 250. Train Loss: 0.30. Train Acc (Test Acc): 0.92 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.2012 - accuracy: 0.9380\n",
      "Epoch # 255. Train Loss: 0.08. Train Acc (Test Acc): 0.98 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1566 - accuracy: 0.9500\n",
      "Epoch # 260. Train Loss: 0.12. Train Acc (Test Acc): 0.96 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1466 - accuracy: 0.9500\n",
      "Epoch # 265. Train Loss: 0.22. Train Acc (Test Acc): 0.91 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1519 - accuracy: 0.9600\n",
      "Epoch # 270. Train Loss: 0.21. Train Acc (Test Acc): 0.93 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1697 - accuracy: 0.9460\n",
      "Epoch # 275. Train Loss: 0.25. Train Acc (Test Acc): 0.90 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1447 - accuracy: 0.9680\n",
      "Epoch # 280. Train Loss: 0.14. Train Acc (Test Acc): 0.97 (0.97)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1342 - accuracy: 0.9540\n",
      "Epoch # 285. Train Loss: 0.14. Train Acc (Test Acc): 0.98 (0.95)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1816 - accuracy: 0.9420\n",
      "Epoch # 290. Train Loss: 0.17. Train Acc (Test Acc): 0.95 (0.94)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1471 - accuracy: 0.9560\n",
      "Epoch # 295. Train Loss: 0.12. Train Acc (Test Acc): 0.96 (0.96)\n",
      "16/16 [==============================] - 0s 2ms/step - loss: 0.1310 - accuracy: 0.9680\n",
      "Epoch # 300. Train Loss: 0.11. Train Acc (Test Acc): 0.97 (0.97)\n"
     ]
    }
   ],
   "source": [
    "train_loss = []\n",
    "train_acc = []\n",
    "test_acc = []\n",
    "\n",
    "for i in range(epochs):\n",
    "    rand_idx = np.random.choice(len(x_train), size=batch_size)\n",
    "    rand_x = x_train[rand_idx]\n",
    "    rand_y = y_train[rand_idx]\n",
    "    \n",
    "    history_train = model.train_on_batch(rand_x, rand_y)\n",
    "    \n",
    "    if((i+1)%eval_every == 0):\n",
    "        eval_index = np.random.choice(len(x_test), size=eval_size)\n",
    "        eval_x = x_test[eval_index]\n",
    "        eval_y = y_test[eval_index]\n",
    "        \n",
    "        history_eval = model.evaluate(eval_x, eval_y)\n",
    "        \n",
    "        train_loss.append(history_train[0])\n",
    "        train_acc.append(history_train[1])\n",
    "        \n",
    "        test_acc.append(history_eval[1])\n",
    "        \n",
    "        acc_and_loss = [(i+1), history_train[0], history_train[1], history_eval[1]]\n",
    "        acc_and_loss = [np.round(x,2) for x in acc_and_loss]\n",
    "        print('Epoch # {}. Train Loss: {:.2f}. Train Acc (Test Acc): {:.2f} ({:.2f})'.format(*acc_and_loss))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "about-doctrine",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:40.936777Z",
     "iopub.status.busy": "2021-05-14T09:47:40.935931Z",
     "iopub.status.idle": "2021-05-14T09:47:41.103998Z",
     "shell.execute_reply": "2021-05-14T09:47:41.104451Z"
    },
    "papermill": {
     "duration": 0.242027,
     "end_time": "2021-05-14T09:47:41.104629",
     "exception": false,
     "start_time": "2021-05-14T09:47:40.862602",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "range(0, 300, 5)\n"
     ]
    },
    {
     "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": [
    "eval_indices = range(0, epochs, eval_every)\n",
    "print(eval_indices)\n",
    "plt.plot(eval_indices, train_acc, 'k-', label='train_loss')\n",
    "plt.plot(eval_indices, test_acc, 'r--', label='test_loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "abstract-pastor",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-05-14T09:47:41.230565Z",
     "iopub.status.busy": "2021-05-14T09:47:41.229746Z",
     "iopub.status.idle": "2021-05-14T09:47:41.606493Z",
     "shell.execute_reply": "2021-05-14T09:47:41.605703Z"
    },
    "papermill": {
     "duration": 0.442151,
     "end_time": "2021-05-14T09:47:41.606618",
     "exception": false,
     "start_time": "2021-05-14T09:47:41.164467",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y = y_test[120:126]\n",
    "y_pred= model.predict(x_test[120:126])\n",
    "\n",
    "predictions = np.argmax(y_pred, axis=1)\n",
    "\n",
    "images = np.squeeze(x_test[120:126])\n",
    "\n",
    "rows = 3\n",
    "cols = 2\n",
    "\n",
    "for i in range(6):\n",
    "    plt.subplot(rows, cols, i+1)\n",
    "    plt.imshow(np.reshape(images[i], [32,32]))\n",
    "    plt.title('Actual: '+str(y[i])+' Pred: '+str(predictions[i]))\n",
    "    \n",
    "    frame = plt.gca()\n",
    "    frame.axes.get_xaxis().set_visible(False)\n",
    "    frame.axes.get_yaxis().set_visible(False)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aging-daniel",
   "metadata": {
    "papermill": {
     "duration": 0.060561,
     "end_time": "2021-05-14T09:47:41.728791",
     "exception": false,
     "start_time": "2021-05-14T09:47:41.668230",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.10"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 30.630851,
   "end_time": "2021-05-14T09:47:44.505157",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2021-05-14T09:47:13.874306",
   "version": "2.3.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}