Deep_Learning_keras_part4_train_callbacks.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.6"
    },
    "colab": {
      "name": "Deep_Learning_keras_part4_train_callbacks.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "include_colab_link": true
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/naomifridman/Introduction_to_deep_learning/blob/master/Deep_Learning_keras_part4_train_callbacks.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6TPJHm0iOKvN",
        "colab_type": "text"
      },
      "source": [
        "## Introduction to Deep Learning \n",
        "# Part 4 - Keras Train callbacks\n",
        "### Example Fushion Mnist data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kZhItv6nOKvR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "\n",
        "import numpy as np\n",
        "\n",
        "import matplotlib.gridspec as gridspec\n",
        "from sklearn.metrics import accuracy_score, log_loss\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "%matplotlib inline  "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bMjswk8pOKvp",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.metrics import accuracy_score\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.metrics import classification_report\n",
        "\n",
        "def print_stats(ytest, ypred):\n",
        "    \n",
        "    print(\"Accuracy: {:.5f}\".format(accuracy_score(ytest, ypred)))\n",
        "    print(\"Confusion Matrix:\")\n",
        "    print(confusion_matrix(ytest, ypred))\n",
        "    print(\"Classification Report:\")\n",
        "    print(classification_report(ytest, ypred))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Eq8sRWEbPLt7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def show_list_images(imgs, titles=None):\n",
        "    n_img = len(imgs)\n",
        "    fig, m_axs = plt.subplots(1, n_img, figsize = (n_img*2, 4))\n",
        "    i = 0\n",
        "    for (c_im) in m_axs.T:\n",
        "        c_im.imshow(imgs[i])\n",
        "        c_im.axis('off')\n",
        "        if (titles is not None):\n",
        "            c_im.set_title(titles[i])\n",
        "        i+=1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WpzRJ4c-PJp1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def drow_history(history, metric):\n",
        "    plt.plot(history.history[metric])\n",
        "    plt.plot(history.history['val_'+ metric])\n",
        "    plt.title('model '+metric)\n",
        "    plt.ylabel(metric)\n",
        "    plt.xlabel('epoch')\n",
        "    plt.legend(['train', 'test'], loc='upper left')\n",
        "    plt.show()    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zwQW3cBzOKwR",
        "colab_type": "text"
      },
      "source": [
        "## Keras summary\n",
        "### Sequential Model\n",
        "https://keras.io/models/sequential/\n",
        "\n",
        "### Dense layer\n",
        "https://keras.io/layers/core/\n",
        "\n",
        "### Dropout layer\n",
        "https://keras.io/layers/core/\n",
        "\n",
        "### Optimizer RMSprop, Adam, SGD\n",
        "https://keras.io/optimizers/"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u5qBIiSIOKwZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "1082521b-d285-4eeb-dedb-682696fa7f62"
      },
      "source": [
        "import keras\n",
        "# Keras has many build in data sets\n",
        "from keras.datasets import mnist\n",
        "# Sequential is the basic feed forward neural network (FFN)\n",
        "from keras.models import Sequential\n",
        "# Dense is fully connected layer\n",
        "# Dropout is a \"noising\" layer, to prevent over feet.\n",
        "from keras.layers import Dense, Dropout\n",
        "# Read about optimizers in keras documentation\n",
        "from keras.optimizers import RMSprop"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "s9u9foXWOKxC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.datasets import make_moons, make_circles, make_classification\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.model_selection import train_test_split\n",
        "from matplotlib.colors import ListedColormap\n",
        "from sklearn.metrics import accuracy_score"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X2BV6y33OKxg",
        "colab_type": "text"
      },
      "source": [
        "## Classification Exercise Fashion Mnist\n",
        "https://keras.io/datasets/\n",
        "Dataset of 60,000 28x28 grayscale images of 10 fashion categories, along with a test set of 10,000 images. This dataset can be used as a drop-in replacement for MNIST. The class labels are:\n",
        "```\n",
        "* Label\tDescription\n",
        "0\tT-shirt/top\n",
        "1\tTrouser\n",
        "2\tPullover\n",
        "3\tDress\n",
        "4\tCoat\n",
        "5\tSandal\n",
        "6\tShirt\n",
        "7\tSneaker\n",
        "8\tBag\n",
        "9\tAnkle boot\n",
        "```\n",
        "\n",
        "Build the best model for this data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7gZd_nFYOKxi",
        "colab_type": "text"
      },
      "source": [
        "## Load and Preprocess Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "azzLFarzOKxu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 163
        },
        "outputId": "3bab6606-18b9-4756-9457-7dc3268904ea"
      },
      "source": [
        "from keras.datasets import fashion_mnist\n",
        "\n",
        "(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 0s 3us/step\n",
            "Downloading data from http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz\n",
            "26427392/26421880 [==============================] - 2s 0us/step\n",
            "Downloading data from http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz\n",
            "8192/5148 [===============================================] - 0s 0us/step\n",
            "Downloading data from http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz\n",
            "4423680/4422102 [==============================] - 1s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l8PmCkn0OKyV",
        "colab_type": "code",
        "outputId": "a4905adb-6fe5-49ee-a263-b0ca52e609ad",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "x_train[0].shape, x_train.shape,x_test.shape"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((28, 28), (60000, 28, 28), (10000, 28, 28))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SFDnbiD2OKyv",
        "colab_type": "code",
        "outputId": "700f1340-c04c-4b19-f9e2-84ec1d1c0b23",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 178
        }
      },
      "source": [
        "imgs = [x_train[i].reshape(28,28) for i in range(100,110)]\n",
        "titles = [(\"lable {}\").format(y_train[i]) for i in range(20)]\n",
        "show_list_images(imgs, titles)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABH4AAACNCAYAAADB/L29AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 1440x288 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LYgmxAs0OKzZ",
        "colab_type": "code",
        "outputId": "33d86b01-8fed-48b8-e2f9-96a95f5cf214",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "x_train = x_train.reshape(60000, 784)\n",
        "x_test = x_test.reshape(10000, 784)\n",
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "x_train /= 255\n",
        "x_test /= 255\n",
        "print(x_train.shape[0], 'train samples')\n",
        "print(x_test.shape[0], 'test samples')"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "60000 train samples\n",
            "10000 test samples\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kM2I9gD9OKzv",
        "colab_type": "code",
        "outputId": "db4dc71f-b062-4bc4-f3d9-6342f608a2f2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 108
        }
      },
      "source": [
        "num_classes=10\n",
        "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
        "y_test = keras.utils.to_categorical(y_test, num_classes)\n",
        "y_train[0:5]"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
              "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
              "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SGypgNSlOK0d",
        "colab_type": "text"
      },
      "source": [
        "## Build Keras Basic Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KyLarqfkOK1G",
        "colab_type": "code",
        "outputId": "c41a5e9c-ee52-49db-c006-66b6aaa9c67b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 272
        }
      },
      "source": [
        "def get_model():\n",
        "    model = Sequential()\n",
        "    model.add(Dense(512, activation='relu', input_shape=(784,)))\n",
        "    model.add(Dense(512, activation='relu'))\n",
        "    model.add(Dense(num_classes, activation='softmax'))\n",
        "    return model\n",
        "\n",
        "model=get_model()\n",
        "model.summary()\n",
        "\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer=RMSprop(),\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_4 (Dense)              (None, 512)               401920    \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 512)               262656    \n",
            "_________________________________________________________________\n",
            "dense_6 (Dense)              (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 669,706\n",
            "Trainable params: 669,706\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pnHP2_-AOK1f",
        "colab_type": "code",
        "outputId": "8c810722-3a5e-4376-b777-3004ad56fa81",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 181
        }
      },
      "source": [
        "history = model.fit(x_train, y_train,\n",
        "                    batch_size=40,\n",
        "                    epochs=4,\n",
        "                    verbose=1,\n",
        "                    validation_data=(x_test, y_test))"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/4\n",
            "60000/60000 [==============================] - 8s 139us/step - loss: 0.5052 - acc: 0.8185 - val_loss: 0.4407 - val_acc: 0.8438\n",
            "Epoch 2/4\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.3997 - acc: 0.8598 - val_loss: 0.3968 - val_acc: 0.8598\n",
            "Epoch 3/4\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.3757 - acc: 0.8701 - val_loss: 0.4713 - val_acc: 0.8434\n",
            "Epoch 4/4\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.3617 - acc: 0.8768 - val_loss: 0.4344 - val_acc: 0.8592\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sF_8Yd-YOK1x",
        "colab_type": "code",
        "outputId": "89e1642c-6b23-48dc-8252-e0646274cf13",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "drow_history(history, 'loss')\n",
        "drow_history(history, 'acc')"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nU6H/B5t0jTlXg/P7AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGGMqZMnCGGNMhSxZGGOMqdD/ByU8Myp77AZRAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KKZX58JGOK2V",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pOpsmb6XOK2l",
        "colab_type": "code",
        "outputId": "843785c7-e36f-48a1-d89c-214d02efe3dd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "score = model.evaluate(x_test, y_test, verbose=0)\n",
        "print('Test loss:', score[0])\n",
        "print('Test accuracy:', score[1])"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test loss: 0.4344308448433876\n",
            "Test accuracy: 0.8592\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_P5zwQ0wOK3Q",
        "colab_type": "text"
      },
      "source": [
        "## Predict"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2eQ2E-fNOK3S",
        "colab_type": "code",
        "outputId": "53cf0332-171f-4331-d6f4-636ea05fda32",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 290
        }
      },
      "source": [
        "pred_test = model.predict(x_test)\n",
        "pred_test[0:5]"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[7.7835853e-09, 4.5823411e-11, 2.1200233e-08, 2.8843166e-09,\n",
              "        4.8735571e-09, 1.4115536e-02, 1.3823315e-07, 1.2286642e-02,\n",
              "        3.1321056e-07, 9.7359735e-01],\n",
              "       [1.2767391e-07, 4.9584055e-17, 9.9994624e-01, 1.6741096e-11,\n",
              "        2.2617141e-05, 3.4308292e-17, 3.1038602e-05, 1.2770224e-19,\n",
              "        4.5323784e-12, 3.0635242e-18],\n",
              "       [7.7411751e-32, 1.0000000e+00, 1.9398484e-36, 8.4954985e-25,\n",
              "        1.6125508e-38, 0.0000000e+00, 3.6318639e-36, 0.0000000e+00,\n",
              "        0.0000000e+00, 0.0000000e+00],\n",
              "       [3.6633983e-34, 1.0000000e+00, 0.0000000e+00, 2.7174607e-24,\n",
              "        0.0000000e+00, 0.0000000e+00, 0.0000000e+00, 0.0000000e+00,\n",
              "        0.0000000e+00, 0.0000000e+00],\n",
              "       [2.5374037e-01, 7.5791190e-06, 2.3574477e-01, 5.4945266e-03,\n",
              "        1.0437249e-02, 9.4554234e-06, 4.9092591e-01, 1.3413891e-06,\n",
              "        3.6380105e-03, 6.5928805e-07]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gGIX5KPnOK33",
        "colab_type": "code",
        "outputId": "3dee981c-1e52-4647-9dd5-13173c590584",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 217
        }
      },
      "source": [
        "y_pred_test=np.argmax(pred_test, axis=1)\n",
        "y_pred_test[0:10], y_test[0:10]"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(array([9, 2, 1, 1, 6, 1, 4, 6, 5, 7]),\n",
              " array([[0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
              "        [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
              "        [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "        [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
              "        [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
              "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
              "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
              "        [0., 0., 0., 0., 0., 0., 0., 1., 0., 0.]], dtype=float32))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dWdNdrX-OK4k",
        "colab_type": "code",
        "outputId": "8cbb1764-7287-46f9-a8f5-4eb2d4472116",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "pred_test = model.predict_classes(x_test)\n",
        "pred_test[0:5]"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([9, 2, 1, 1, 6])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "cXPlt3wFOK4_",
        "colab_type": "code",
        "outputId": "0466ccf5-b884-4027-c22d-a242e86eec1b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 563
        }
      },
      "source": [
        "print_stats(np.argmax(y_test, axis=1),y_pred_test )"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.85920\n",
            "Confusion Matrix:\n",
            "[[752   0  40  66   5   1 126   0  10   0]\n",
            " [  0 959  10  25   3   0   2   0   1   0]\n",
            " [  5   0 899  11  65   0  19   0   1   0]\n",
            " [ 16   8  24 895  38   0  13   0   6   0]\n",
            " [  0   0 215  29 735   0  21   0   0   0]\n",
            " [  0   0   0   2   0 967   0  11   2  18]\n",
            " [106   0 210  48  82   0 544   0  10   0]\n",
            " [  0   0   0   0   0  32   0 902   0  66]\n",
            " [  1   0  12   5   4   2   2   3 971   0]\n",
            " [  0   1   0   0   0   7   1  23   0 968]]\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.85      0.75      0.80      1000\n",
            "           1       0.99      0.96      0.97      1000\n",
            "           2       0.64      0.90      0.75      1000\n",
            "           3       0.83      0.90      0.86      1000\n",
            "           4       0.79      0.73      0.76      1000\n",
            "           5       0.96      0.97      0.96      1000\n",
            "           6       0.75      0.54      0.63      1000\n",
            "           7       0.96      0.90      0.93      1000\n",
            "           8       0.97      0.97      0.97      1000\n",
            "           9       0.92      0.97      0.94      1000\n",
            "\n",
            "    accuracy                           0.86     10000\n",
            "   macro avg       0.87      0.86      0.86     10000\n",
            "weighted avg       0.87      0.86      0.86     10000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zATAMIV2OK5i",
        "colab_type": "text"
      },
      "source": [
        "# Keras callbacks\n",
        "* https://keras.io/callbacks/"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QTSHVKudOK5v",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau\n",
        "\n",
        "\n",
        "earlystopper = EarlyStopping(patience=9, monitor='val_acc', verbose=1)\n",
        "\n",
        "filepath=\".\\\\model_acc_epoch_{epoch:03d}-acc_{acc:.3f}_loss_{loss:.3f}.hdf5\"\n",
        "checkpoint1 = ModelCheckpoint(filepath, monitor='acc', verbose=1, \n",
        "                              save_best_only=True, mode='max',\n",
        "                             save_weights_only=True)\n",
        "\n",
        "filepath=\".\\\\model_val_acc_epoch_{epoch:03d}_val_acc_{val_acc:.4f}_val_loss_{val_loss:.4f}.hdf5\"\n",
        "checkpoint2 = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, \n",
        "                              save_best_only=True, mode='max',\n",
        "                             save_weights_only=True)\n",
        "\n",
        "filepath=\".\\\\model_loss_epoch_{epoch:03d}_acc{acc:.3f}_loss{loss:.3f}.hdf5\"\n",
        "checkpoint3 = ModelCheckpoint(filepath, monitor='loss', verbose=1, \n",
        "                              save_best_only=True, mode='min',\n",
        "                             save_weights_only=True)\n",
        "\n",
        "filepath=\".\\\\model_val_loss_epoch_{epoch:03d}_val_acc{val_acc:.4f}_val_loss{val_loss:.4f}.hdf5\"\n",
        "checkpoint4 = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, \n",
        "                              save_best_only=True, mode='min',\n",
        "                             save_weights_only=True)\n",
        "\n",
        "filepath=\".\\\\model_last.hdf5\"\n",
        "checkpoint5 = ModelCheckpoint(filepath, monitor='val_loss', \n",
        "                              verbose=0, save_best_only=False,\n",
        "                             save_weights_only=True)\n",
        "\n",
        "\n",
        "reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5,\n",
        "                              patience=3, min_lr=0.000001, verbose=1,  cooldown=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NBdslbtaOK6E",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model=get_model()\n",
        "\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer=RMSprop(),\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jJ2Ymk-UOK6_",
        "colab_type": "code",
        "outputId": "309b93da-f824-4156-8da0-0c0d630509f6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "history = model.fit(x_train, y_train,\n",
        "                    batch_size=40,\n",
        "                    epochs=100,\n",
        "                    verbose=1,\n",
        "                    validation_data=(x_test, y_test),\n",
        "                   callbacks=[earlystopper, checkpoint1, checkpoint2, checkpoint3,\n",
        "                             checkpoint4, checkpoint5, reduce_lr])"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/100\n",
            "60000/60000 [==============================] - 9s 146us/step - loss: 0.5015 - acc: 0.8195 - val_loss: 0.4262 - val_acc: 0.8503\n",
            "\n",
            "Epoch 00001: acc improved from -inf to 0.81953, saving model to .\\model_acc_epoch_001-acc_0.820_loss_0.502.hdf5\n",
            "\n",
            "Epoch 00001: val_acc improved from -inf to 0.85030, saving model to .\\model_val_acc_epoch_001_val_acc_0.8503_val_loss_0.4262.hdf5\n",
            "\n",
            "Epoch 00001: loss improved from inf to 0.50151, saving model to .\\model_loss_epoch_001_acc0.820_loss0.502.hdf5\n",
            "\n",
            "Epoch 00001: val_loss improved from inf to 0.42620, saving model to .\\model_val_loss_epoch_001_val_acc0.8503_val_loss0.4262.hdf5\n",
            "Epoch 2/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.4017 - acc: 0.8603 - val_loss: 0.4169 - val_acc: 0.8666\n",
            "\n",
            "Epoch 00002: acc improved from 0.81953 to 0.86033, saving model to .\\model_acc_epoch_002-acc_0.860_loss_0.402.hdf5\n",
            "\n",
            "Epoch 00002: val_acc improved from 0.85030 to 0.86660, saving model to .\\model_val_acc_epoch_002_val_acc_0.8666_val_loss_0.4169.hdf5\n",
            "\n",
            "Epoch 00002: loss improved from 0.50151 to 0.40169, saving model to .\\model_loss_epoch_002_acc0.860_loss0.402.hdf5\n",
            "\n",
            "Epoch 00002: val_loss improved from 0.42620 to 0.41692, saving model to .\\model_val_loss_epoch_002_val_acc0.8666_val_loss0.4169.hdf5\n",
            "Epoch 3/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.3798 - acc: 0.8690 - val_loss: 0.4043 - val_acc: 0.8667\n",
            "\n",
            "Epoch 00003: acc improved from 0.86033 to 0.86897, saving model to .\\model_acc_epoch_003-acc_0.869_loss_0.380.hdf5\n",
            "\n",
            "Epoch 00003: val_acc improved from 0.86660 to 0.86670, saving model to .\\model_val_acc_epoch_003_val_acc_0.8667_val_loss_0.4043.hdf5\n",
            "\n",
            "Epoch 00003: loss improved from 0.40169 to 0.37977, saving model to .\\model_loss_epoch_003_acc0.869_loss0.380.hdf5\n",
            "\n",
            "Epoch 00003: val_loss improved from 0.41692 to 0.40430, saving model to .\\model_val_loss_epoch_003_val_acc0.8667_val_loss0.4043.hdf5\n",
            "Epoch 4/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.3649 - acc: 0.8768 - val_loss: 0.4066 - val_acc: 0.8712\n",
            "\n",
            "Epoch 00004: acc improved from 0.86897 to 0.87677, saving model to .\\model_acc_epoch_004-acc_0.877_loss_0.365.hdf5\n",
            "\n",
            "Epoch 00004: val_acc improved from 0.86670 to 0.87120, saving model to .\\model_val_acc_epoch_004_val_acc_0.8712_val_loss_0.4066.hdf5\n",
            "\n",
            "Epoch 00004: loss improved from 0.37977 to 0.36485, saving model to .\\model_loss_epoch_004_acc0.877_loss0.365.hdf5\n",
            "\n",
            "Epoch 00004: val_loss did not improve from 0.40430\n",
            "Epoch 5/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.3544 - acc: 0.8797 - val_loss: 0.4504 - val_acc: 0.8571\n",
            "\n",
            "Epoch 00005: acc improved from 0.87677 to 0.87972, saving model to .\\model_acc_epoch_005-acc_0.880_loss_0.354.hdf5\n",
            "\n",
            "Epoch 00005: val_acc did not improve from 0.87120\n",
            "\n",
            "Epoch 00005: loss improved from 0.36485 to 0.35438, saving model to .\\model_loss_epoch_005_acc0.880_loss0.354.hdf5\n",
            "\n",
            "Epoch 00005: val_loss did not improve from 0.40430\n",
            "Epoch 6/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.3457 - acc: 0.8840 - val_loss: 0.4602 - val_acc: 0.8639\n",
            "\n",
            "Epoch 00006: acc improved from 0.87972 to 0.88400, saving model to .\\model_acc_epoch_006-acc_0.884_loss_0.346.hdf5\n",
            "\n",
            "Epoch 00006: val_acc did not improve from 0.87120\n",
            "\n",
            "Epoch 00006: loss improved from 0.35438 to 0.34568, saving model to .\\model_loss_epoch_006_acc0.884_loss0.346.hdf5\n",
            "\n",
            "Epoch 00006: val_loss did not improve from 0.40430\n",
            "\n",
            "Epoch 00006: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
            "Epoch 7/100\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.2907 - acc: 0.9008 - val_loss: 0.3992 - val_acc: 0.8775\n",
            "\n",
            "Epoch 00007: acc improved from 0.88400 to 0.90080, saving model to .\\model_acc_epoch_007-acc_0.901_loss_0.291.hdf5\n",
            "\n",
            "Epoch 00007: val_acc improved from 0.87120 to 0.87750, saving model to .\\model_val_acc_epoch_007_val_acc_0.8775_val_loss_0.3992.hdf5\n",
            "\n",
            "Epoch 00007: loss improved from 0.34568 to 0.29071, saving model to .\\model_loss_epoch_007_acc0.901_loss0.291.hdf5\n",
            "\n",
            "Epoch 00007: val_loss improved from 0.40430 to 0.39922, saving model to .\\model_val_loss_epoch_007_val_acc0.8775_val_loss0.3992.hdf5\n",
            "Epoch 8/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.2832 - acc: 0.9047 - val_loss: 0.4381 - val_acc: 0.8761\n",
            "\n",
            "Epoch 00008: acc improved from 0.90080 to 0.90468, saving model to .\\model_acc_epoch_008-acc_0.905_loss_0.283.hdf5\n",
            "\n",
            "Epoch 00008: val_acc did not improve from 0.87750\n",
            "\n",
            "Epoch 00008: loss improved from 0.29071 to 0.28324, saving model to .\\model_loss_epoch_008_acc0.905_loss0.283.hdf5\n",
            "\n",
            "Epoch 00008: val_loss did not improve from 0.39922\n",
            "Epoch 9/100\n",
            "60000/60000 [==============================] - 9s 143us/step - loss: 0.2789 - acc: 0.9062 - val_loss: 0.4460 - val_acc: 0.8799\n",
            "\n",
            "Epoch 00009: acc improved from 0.90468 to 0.90620, saving model to .\\model_acc_epoch_009-acc_0.906_loss_0.279.hdf5\n",
            "\n",
            "Epoch 00009: val_acc improved from 0.87750 to 0.87990, saving model to .\\model_val_acc_epoch_009_val_acc_0.8799_val_loss_0.4460.hdf5\n",
            "\n",
            "Epoch 00009: loss improved from 0.28324 to 0.27892, saving model to .\\model_loss_epoch_009_acc0.906_loss0.279.hdf5\n",
            "\n",
            "Epoch 00009: val_loss did not improve from 0.39922\n",
            "Epoch 10/100\n",
            "60000/60000 [==============================] - 9s 144us/step - loss: 0.2761 - acc: 0.9073 - val_loss: 0.4104 - val_acc: 0.8844\n",
            "\n",
            "Epoch 00010: acc improved from 0.90620 to 0.90728, saving model to .\\model_acc_epoch_010-acc_0.907_loss_0.276.hdf5\n",
            "\n",
            "Epoch 00010: val_acc improved from 0.87990 to 0.88440, saving model to .\\model_val_acc_epoch_010_val_acc_0.8844_val_loss_0.4104.hdf5\n",
            "\n",
            "Epoch 00010: loss improved from 0.27892 to 0.27605, saving model to .\\model_loss_epoch_010_acc0.907_loss0.276.hdf5\n",
            "\n",
            "Epoch 00010: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00010: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
            "Epoch 11/100\n",
            "60000/60000 [==============================] - 8s 139us/step - loss: 0.2404 - acc: 0.9178 - val_loss: 0.4090 - val_acc: 0.8911\n",
            "\n",
            "Epoch 00011: acc improved from 0.90728 to 0.91782, saving model to .\\model_acc_epoch_011-acc_0.918_loss_0.240.hdf5\n",
            "\n",
            "Epoch 00011: val_acc improved from 0.88440 to 0.89110, saving model to .\\model_val_acc_epoch_011_val_acc_0.8911_val_loss_0.4090.hdf5\n",
            "\n",
            "Epoch 00011: loss improved from 0.27605 to 0.24036, saving model to .\\model_loss_epoch_011_acc0.918_loss0.240.hdf5\n",
            "\n",
            "Epoch 00011: val_loss did not improve from 0.39922\n",
            "Epoch 12/100\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.2359 - acc: 0.9200 - val_loss: 0.4300 - val_acc: 0.8861\n",
            "\n",
            "Epoch 00012: acc improved from 0.91782 to 0.91997, saving model to .\\model_acc_epoch_012-acc_0.920_loss_0.236.hdf5\n",
            "\n",
            "Epoch 00012: val_acc did not improve from 0.89110\n",
            "\n",
            "Epoch 00012: loss improved from 0.24036 to 0.23586, saving model to .\\model_loss_epoch_012_acc0.920_loss0.236.hdf5\n",
            "\n",
            "Epoch 00012: val_loss did not improve from 0.39922\n",
            "Epoch 13/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.2322 - acc: 0.9214 - val_loss: 0.4268 - val_acc: 0.8872\n",
            "\n",
            "Epoch 00013: acc improved from 0.91997 to 0.92137, saving model to .\\model_acc_epoch_013-acc_0.921_loss_0.232.hdf5\n",
            "\n",
            "Epoch 00013: val_acc did not improve from 0.89110\n",
            "\n",
            "Epoch 00013: loss improved from 0.23586 to 0.23216, saving model to .\\model_loss_epoch_013_acc0.921_loss0.232.hdf5\n",
            "\n",
            "Epoch 00013: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00013: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
            "Epoch 14/100\n",
            "60000/60000 [==============================] - 8s 138us/step - loss: 0.2138 - acc: 0.9282 - val_loss: 0.4273 - val_acc: 0.8914\n",
            "\n",
            "Epoch 00014: acc improved from 0.92137 to 0.92823, saving model to .\\model_acc_epoch_014-acc_0.928_loss_0.214.hdf5\n",
            "\n",
            "Epoch 00014: val_acc improved from 0.89110 to 0.89140, saving model to .\\model_val_acc_epoch_014_val_acc_0.8914_val_loss_0.4273.hdf5\n",
            "\n",
            "Epoch 00014: loss improved from 0.23216 to 0.21378, saving model to .\\model_loss_epoch_014_acc0.928_loss0.214.hdf5\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 0.39922\n",
            "Epoch 15/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.2099 - acc: 0.9292 - val_loss: 0.4283 - val_acc: 0.8921\n",
            "\n",
            "Epoch 00015: acc improved from 0.92823 to 0.92920, saving model to .\\model_acc_epoch_015-acc_0.929_loss_0.210.hdf5\n",
            "\n",
            "Epoch 00015: val_acc improved from 0.89140 to 0.89210, saving model to .\\model_val_acc_epoch_015_val_acc_0.8921_val_loss_0.4283.hdf5\n",
            "\n",
            "Epoch 00015: loss improved from 0.21378 to 0.20990, saving model to .\\model_loss_epoch_015_acc0.929_loss0.210.hdf5\n",
            "\n",
            "Epoch 00015: val_loss did not improve from 0.39922\n",
            "Epoch 16/100\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.2081 - acc: 0.9304 - val_loss: 0.4292 - val_acc: 0.8919\n",
            "\n",
            "Epoch 00016: acc improved from 0.92920 to 0.93038, saving model to .\\model_acc_epoch_016-acc_0.930_loss_0.208.hdf5\n",
            "\n",
            "Epoch 00016: val_acc did not improve from 0.89210\n",
            "\n",
            "Epoch 00016: loss improved from 0.20990 to 0.20815, saving model to .\\model_loss_epoch_016_acc0.930_loss0.208.hdf5\n",
            "\n",
            "Epoch 00016: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00016: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\n",
            "Epoch 17/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.1974 - acc: 0.9338 - val_loss: 0.4478 - val_acc: 0.8940\n",
            "\n",
            "Epoch 00017: acc improved from 0.93038 to 0.93378, saving model to .\\model_acc_epoch_017-acc_0.934_loss_0.197.hdf5\n",
            "\n",
            "Epoch 00017: val_acc improved from 0.89210 to 0.89400, saving model to .\\model_val_acc_epoch_017_val_acc_0.8940_val_loss_0.4478.hdf5\n",
            "\n",
            "Epoch 00017: loss improved from 0.20815 to 0.19735, saving model to .\\model_loss_epoch_017_acc0.934_loss0.197.hdf5\n",
            "\n",
            "Epoch 00017: val_loss did not improve from 0.39922\n",
            "Epoch 18/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1948 - acc: 0.9342 - val_loss: 0.4459 - val_acc: 0.8941\n",
            "\n",
            "Epoch 00018: acc improved from 0.93378 to 0.93418, saving model to .\\model_acc_epoch_018-acc_0.934_loss_0.195.hdf5\n",
            "\n",
            "Epoch 00018: val_acc improved from 0.89400 to 0.89410, saving model to .\\model_val_acc_epoch_018_val_acc_0.8941_val_loss_0.4459.hdf5\n",
            "\n",
            "Epoch 00018: loss improved from 0.19735 to 0.19481, saving model to .\\model_loss_epoch_018_acc0.934_loss0.195.hdf5\n",
            "\n",
            "Epoch 00018: val_loss did not improve from 0.39922\n",
            "Epoch 19/100\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.1929 - acc: 0.9363 - val_loss: 0.4487 - val_acc: 0.8932\n",
            "\n",
            "Epoch 00019: acc improved from 0.93418 to 0.93627, saving model to .\\model_acc_epoch_019-acc_0.936_loss_0.193.hdf5\n",
            "\n",
            "Epoch 00019: val_acc did not improve from 0.89410\n",
            "\n",
            "Epoch 00019: loss improved from 0.19481 to 0.19291, saving model to .\\model_loss_epoch_019_acc0.936_loss0.193.hdf5\n",
            "\n",
            "Epoch 00019: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00019: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\n",
            "Epoch 20/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.1885 - acc: 0.9376 - val_loss: 0.4529 - val_acc: 0.8948\n",
            "\n",
            "Epoch 00020: acc improved from 0.93627 to 0.93755, saving model to .\\model_acc_epoch_020-acc_0.938_loss_0.189.hdf5\n",
            "\n",
            "Epoch 00020: val_acc improved from 0.89410 to 0.89480, saving model to .\\model_val_acc_epoch_020_val_acc_0.8948_val_loss_0.4529.hdf5\n",
            "\n",
            "Epoch 00020: loss improved from 0.19291 to 0.18854, saving model to .\\model_loss_epoch_020_acc0.938_loss0.189.hdf5\n",
            "\n",
            "Epoch 00020: val_loss did not improve from 0.39922\n",
            "Epoch 21/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1869 - acc: 0.9380 - val_loss: 0.4522 - val_acc: 0.8945\n",
            "\n",
            "Epoch 00021: acc improved from 0.93755 to 0.93795, saving model to .\\model_acc_epoch_021-acc_0.938_loss_0.187.hdf5\n",
            "\n",
            "Epoch 00021: val_acc did not improve from 0.89480\n",
            "\n",
            "Epoch 00021: loss improved from 0.18854 to 0.18695, saving model to .\\model_loss_epoch_021_acc0.938_loss0.187.hdf5\n",
            "\n",
            "Epoch 00021: val_loss did not improve from 0.39922\n",
            "Epoch 22/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1861 - acc: 0.9381 - val_loss: 0.4528 - val_acc: 0.8947\n",
            "\n",
            "Epoch 00022: acc improved from 0.93795 to 0.93808, saving model to .\\model_acc_epoch_022-acc_0.938_loss_0.186.hdf5\n",
            "\n",
            "Epoch 00022: val_acc did not improve from 0.89480\n",
            "\n",
            "Epoch 00022: loss improved from 0.18695 to 0.18613, saving model to .\\model_loss_epoch_022_acc0.938_loss0.186.hdf5\n",
            "\n",
            "Epoch 00022: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00022: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\n",
            "Epoch 23/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.1833 - acc: 0.9386 - val_loss: 0.4499 - val_acc: 0.8949\n",
            "\n",
            "Epoch 00023: acc improved from 0.93808 to 0.93855, saving model to .\\model_acc_epoch_023-acc_0.939_loss_0.183.hdf5\n",
            "\n",
            "Epoch 00023: val_acc improved from 0.89480 to 0.89490, saving model to .\\model_val_acc_epoch_023_val_acc_0.8949_val_loss_0.4499.hdf5\n",
            "\n",
            "Epoch 00023: loss improved from 0.18613 to 0.18331, saving model to .\\model_loss_epoch_023_acc0.939_loss0.183.hdf5\n",
            "\n",
            "Epoch 00023: val_loss did not improve from 0.39922\n",
            "Epoch 24/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1821 - acc: 0.9394 - val_loss: 0.4569 - val_acc: 0.8950\n",
            "\n",
            "Epoch 00024: acc improved from 0.93855 to 0.93943, saving model to .\\model_acc_epoch_024-acc_0.939_loss_0.182.hdf5\n",
            "\n",
            "Epoch 00024: val_acc improved from 0.89490 to 0.89500, saving model to .\\model_val_acc_epoch_024_val_acc_0.8950_val_loss_0.4569.hdf5\n",
            "\n",
            "Epoch 00024: loss improved from 0.18331 to 0.18209, saving model to .\\model_loss_epoch_024_acc0.939_loss0.182.hdf5\n",
            "\n",
            "Epoch 00024: val_loss did not improve from 0.39922\n",
            "Epoch 25/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1821 - acc: 0.9398 - val_loss: 0.4590 - val_acc: 0.8948\n",
            "\n",
            "Epoch 00025: acc improved from 0.93943 to 0.93975, saving model to .\\model_acc_epoch_025-acc_0.940_loss_0.182.hdf5\n",
            "\n",
            "Epoch 00025: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00025: loss improved from 0.18209 to 0.18205, saving model to .\\model_loss_epoch_025_acc0.940_loss0.182.hdf5\n",
            "\n",
            "Epoch 00025: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00025: ReduceLROnPlateau reducing learning rate to 7.812500371073838e-06.\n",
            "Epoch 26/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.1806 - acc: 0.9403 - val_loss: 0.4567 - val_acc: 0.8943\n",
            "\n",
            "Epoch 00026: acc improved from 0.93975 to 0.94032, saving model to .\\model_acc_epoch_026-acc_0.940_loss_0.181.hdf5\n",
            "\n",
            "Epoch 00026: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00026: loss improved from 0.18205 to 0.18060, saving model to .\\model_loss_epoch_026_acc0.940_loss0.181.hdf5\n",
            "\n",
            "Epoch 00026: val_loss did not improve from 0.39922\n",
            "Epoch 27/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.1803 - acc: 0.9406 - val_loss: 0.4577 - val_acc: 0.8940\n",
            "\n",
            "Epoch 00027: acc improved from 0.94032 to 0.94057, saving model to .\\model_acc_epoch_027-acc_0.941_loss_0.180.hdf5\n",
            "\n",
            "Epoch 00027: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00027: loss improved from 0.18060 to 0.18028, saving model to .\\model_loss_epoch_027_acc0.941_loss0.180.hdf5\n",
            "\n",
            "Epoch 00027: val_loss did not improve from 0.39922\n",
            "Epoch 28/100\n",
            "60000/60000 [==============================] - 8s 135us/step - loss: 0.1803 - acc: 0.9404 - val_loss: 0.4570 - val_acc: 0.8947\n",
            "\n",
            "Epoch 00028: acc did not improve from 0.94057\n",
            "\n",
            "Epoch 00028: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00028: loss did not improve from 0.18028\n",
            "\n",
            "Epoch 00028: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00028: ReduceLROnPlateau reducing learning rate to 3.906250185536919e-06.\n",
            "Epoch 29/100\n",
            "60000/60000 [==============================] - 8s 133us/step - loss: 0.1791 - acc: 0.9409 - val_loss: 0.4576 - val_acc: 0.8938\n",
            "\n",
            "Epoch 00029: acc improved from 0.94057 to 0.94093, saving model to .\\model_acc_epoch_029-acc_0.941_loss_0.179.hdf5\n",
            "\n",
            "Epoch 00029: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00029: loss improved from 0.18028 to 0.17910, saving model to .\\model_loss_epoch_029_acc0.941_loss0.179.hdf5\n",
            "\n",
            "Epoch 00029: val_loss did not improve from 0.39922\n",
            "Epoch 30/100\n",
            "60000/60000 [==============================] - 8s 133us/step - loss: 0.1791 - acc: 0.9411 - val_loss: 0.4583 - val_acc: 0.8944\n",
            "\n",
            "Epoch 00030: acc improved from 0.94093 to 0.94105, saving model to .\\model_acc_epoch_030-acc_0.941_loss_0.179.hdf5\n",
            "\n",
            "Epoch 00030: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00030: loss improved from 0.17910 to 0.17907, saving model to .\\model_loss_epoch_030_acc0.941_loss0.179.hdf5\n",
            "\n",
            "Epoch 00030: val_loss did not improve from 0.39922\n",
            "Epoch 31/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1790 - acc: 0.9411 - val_loss: 0.4587 - val_acc: 0.8946\n",
            "\n",
            "Epoch 00031: acc improved from 0.94105 to 0.94112, saving model to .\\model_acc_epoch_031-acc_0.941_loss_0.179.hdf5\n",
            "\n",
            "Epoch 00031: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00031: loss improved from 0.17907 to 0.17900, saving model to .\\model_loss_epoch_031_acc0.941_loss0.179.hdf5\n",
            "\n",
            "Epoch 00031: val_loss did not improve from 0.39922\n",
            "\n",
            "Epoch 00031: ReduceLROnPlateau reducing learning rate to 1.9531250927684596e-06.\n",
            "Epoch 32/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1784 - acc: 0.9414 - val_loss: 0.4588 - val_acc: 0.8943\n",
            "\n",
            "Epoch 00032: acc improved from 0.94112 to 0.94138, saving model to .\\model_acc_epoch_032-acc_0.941_loss_0.178.hdf5\n",
            "\n",
            "Epoch 00032: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00032: loss improved from 0.17900 to 0.17845, saving model to .\\model_loss_epoch_032_acc0.941_loss0.178.hdf5\n",
            "\n",
            "Epoch 00032: val_loss did not improve from 0.39922\n",
            "Epoch 33/100\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 0.1785 - acc: 0.9412 - val_loss: 0.4588 - val_acc: 0.8944\n",
            "\n",
            "Epoch 00033: acc did not improve from 0.94138\n",
            "\n",
            "Epoch 00033: val_acc did not improve from 0.89500\n",
            "\n",
            "Epoch 00033: loss did not improve from 0.17845\n",
            "\n",
            "Epoch 00033: val_loss did not improve from 0.39922\n",
            "Epoch 00033: early stopping\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HIBGR5gxOK7q",
        "colab_type": "code",
        "outputId": "9e9fb55c-f25c-4ff3-c4b4-360f912b2a9f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        }
      },
      "source": [
        "drow_history(history, 'loss')\n",
        "drow_history(history, 'acc')"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nY4wxrbJkYYwxplX/P0UFIS0KBALoAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npkX/H2UIbIZqPWmFAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bidx8CQVQuAA",
        "colab_type": "code",
        "outputId": "db108e0b-af58-4bbe-b7be-e0316910a585",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 563
        }
      },
      "source": [
        "y_pred_test=model.predict_classes(x_test)\n",
        "print_stats(np.argmax(y_test, axis=1),y_pred_test )"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.89440\n",
            "Confusion Matrix:\n",
            "[[846   1  12  22   3   1 110   0   5   0]\n",
            " [  3 974   1  18   1   0   2   0   1   0]\n",
            " [ 17   0 807  14  97   0  63   1   1   0]\n",
            " [ 21   3  12 909  32   0  19   0   4   0]\n",
            " [  0   1  79  26 838   0  55   0   1   0]\n",
            " [  0   0   0   1   0 966   0  18   2  13]\n",
            " [108   0  86  30  60   1 707   0   8   0]\n",
            " [  0   0   0   0   0  11   0 970   0  19]\n",
            " [  4   0   4   4   5   4   4   4 970   1]\n",
            " [  0   0   0   0   0   8   1  34   0 957]]\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.85      0.85      0.85      1000\n",
            "           1       0.99      0.97      0.98      1000\n",
            "           2       0.81      0.81      0.81      1000\n",
            "           3       0.89      0.91      0.90      1000\n",
            "           4       0.81      0.84      0.82      1000\n",
            "           5       0.97      0.97      0.97      1000\n",
            "           6       0.74      0.71      0.72      1000\n",
            "           7       0.94      0.97      0.96      1000\n",
            "           8       0.98      0.97      0.97      1000\n",
            "           9       0.97      0.96      0.96      1000\n",
            "\n",
            "    accuracy                           0.89     10000\n",
            "   macro avg       0.89      0.89      0.89     10000\n",
            "weighted avg       0.89      0.89      0.89     10000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zPOw8ZAnQz0i",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model=get_model()\n",
        "model.load_weights('.\\model_val_acc_epoch_024_val_acc_0.8950_val_loss_0.4569.hdf5')\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer=RMSprop(),\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GzfbwUdgQ_D6",
        "colab_type": "code",
        "outputId": "11fceece-4e01-4dff-a198-2d2cabad2d4a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 563
        }
      },
      "source": [
        "y_pred_test=model.predict_classes(x_test)\n",
        "print_stats(np.argmax(y_test, axis=1),y_pred_test )"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.89500\n",
            "Confusion Matrix:\n",
            "[[841   1  14  20   3   1 115   0   5   0]\n",
            " [  2 977   2  15   1   0   2   0   1   0]\n",
            " [ 18   0 817  13  90   0  60   1   1   0]\n",
            " [ 20   4  14 908  31   0  19   0   4   0]\n",
            " [  0   1  83  27 836   0  52   0   1   0]\n",
            " [  0   0   0   1   0 967   0  18   2  12]\n",
            " [105   0  88  30  61   1 707   0   8   0]\n",
            " [  0   0   0   0   0  11   0 969   0  20]\n",
            " [  4   0   4   4   5   4   4   4 970   1]\n",
            " [  0   0   0   0   0   8   1  33   0 958]]\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.85      0.84      0.85      1000\n",
            "           1       0.99      0.98      0.99      1000\n",
            "           2       0.80      0.82      0.81      1000\n",
            "           3       0.89      0.91      0.90      1000\n",
            "           4       0.81      0.84      0.82      1000\n",
            "           5       0.97      0.97      0.97      1000\n",
            "           6       0.74      0.71      0.72      1000\n",
            "           7       0.95      0.97      0.96      1000\n",
            "           8       0.98      0.97      0.97      1000\n",
            "           9       0.97      0.96      0.96      1000\n",
            "\n",
            "    accuracy                           0.90     10000\n",
            "   macro avg       0.89      0.89      0.89     10000\n",
            "weighted avg       0.89      0.90      0.89     10000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wXrdkeQOOK8e",
        "colab_type": "text"
      },
      "source": [
        "### Learn more:\n",
        "* https://keras.io/\n",
        "* https://www.coursera.org/learn/neural-networks/lecture/kOQiH/why-object-recognition-is-difficult-5-min"
      ]
    }
  ]
}