Day_070-Keras_Mnist_MLP_HW.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 作業目標:¶\n",
    "利用手寫辨識資料說明MLP 中, 神經網路層的多寡跟預測結果的差異"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 作業重點:\n",
    "[練習]請嘗試新增一層或兩層看看最後預測結果是否有差異\n",
    "model.add(Dense(units=256, input_dim=784, kernel_initializer='normal', activation='relu'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "zfI5zCjRPlnc"
   },
   "source": [
    "# 資料預處理"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 20601,
     "status": "ok",
     "timestamp": 1533441753568,
     "user": {
      "displayName": "Edward Cheng",
      "photoUrl": "//lh4.googleusercontent.com/-ANnwjGu3IBk/AAAAAAAAAAI/AAAAAAAAAAc/qXXg6Jek9xw/s50-c-k-no/photo.jpg",
      "userId": "112554017642991017343"
     },
     "user_tz": -480
    },
    "id": "GyIQ7tpGPlnf",
    "outputId": "f3642b99-3ef2-43fe-ee42-e17b1ef50cbb"
   },
   "outputs": [],
   "source": [
    "from keras.utils import np_utils\n",
    "import numpy as np\n",
    "np.random.seed(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 5937,
     "status": "ok",
     "timestamp": 1533441759521,
     "user": {
      "displayName": "Edward Cheng",
      "photoUrl": "//lh4.googleusercontent.com/-ANnwjGu3IBk/AAAAAAAAAAI/AAAAAAAAAAc/qXXg6Jek9xw/s50-c-k-no/photo.jpg",
      "userId": "112554017642991017343"
     },
     "user_tz": -480
    },
    "id": "OLzoKlJRPlno",
    "outputId": "c75f10a4-951d-4b3b-ad5f-c8fc4dff0e96"
   },
   "outputs": [],
   "source": [
    "#載入手寫辨識的資料集\n",
    "from keras.datasets import mnist\n",
    "(x_train_image,y_train_label),\\\n",
    "(x_test_image,y_test_label)= mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "QEuQwa4nPlnu"
   },
   "outputs": [],
   "source": [
    "#指定測試集與訓練資料集\n",
    "x_Train =x_train_image.reshape(60000, 784).astype('float32')\n",
    "x_Test = x_test_image.reshape(10000, 784).astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "YisRq-SxPln0"
   },
   "outputs": [],
   "source": [
    "# normalize inputs from 0-255 to 0-1\n",
    "x_Train_normalize = x_Train / 255\n",
    "x_Test_normalize = x_Test / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "AHtfmzpfPln4"
   },
   "outputs": [],
   "source": [
    "#把LABEL轉成NUMERICAL Categorical \n",
    "y_Train_OneHot = np_utils.to_categorical(y_train_label)\n",
    "y_Test_OneHot = np_utils.to_categorical(y_test_label)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "0s8PjvqJPln7"
   },
   "source": [
    "# 建立模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "N10v1eX2Pln8"
   },
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "jmaxrsXwPln-"
   },
   "outputs": [],
   "source": [
    "#建立模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "\n",
    "#宣告採用序列模型\n",
    "model = Sequential()\n",
    "\n",
    "#建構輸入層\n",
    "model.add(Dense(units=256, \n",
    "                input_dim=784, \n",
    "                kernel_initializer='normal', \n",
    "                activation='relu'))\n",
    "\n",
    "#建構中間層\n",
    "model.add(Dense(units=512, \n",
    "#                 input_dim=784, \n",
    "                kernel_initializer='normal', \n",
    "                activation='relu'))\n",
    "\n",
    "model.add(Dense(units=512, \n",
    "#                 input_dim=784, \n",
    "                kernel_initializer='normal', \n",
    "                activation='relu'))\n",
    "\n",
    "#建構輸出層\n",
    "model.add(Dense(units=10, \n",
    "                kernel_initializer='normal', \n",
    "                activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_4 (Dense)              (None, 256)               200960    \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 512)               131584    \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 600,330\n",
      "Trainable params: 600,330\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "jdoAZgv1PloK"
   },
   "source": [
    "# 訓練模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "5d5wrEL3PloK"
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy', \n",
    "              optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 374
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 15426,
     "status": "ok",
     "timestamp": 1533441786624,
     "user": {
      "displayName": "Edward Cheng",
      "photoUrl": "//lh4.googleusercontent.com/-ANnwjGu3IBk/AAAAAAAAAAI/AAAAAAAAAAc/qXXg6Jek9xw/s50-c-k-no/photo.jpg",
      "userId": "112554017642991017343"
     },
     "user_tz": -480
    },
    "id": "VlpjoqQ9PloM",
    "outputId": "b91454c9-af9a-40fe-9736-b4a1c97462cd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.2236 - accuracy: 0.9321 - val_loss: 0.1259 - val_accuracy: 0.9630\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0992 - accuracy: 0.9703 - val_loss: 0.1105 - val_accuracy: 0.9682\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.0698 - accuracy: 0.9785 - val_loss: 0.1115 - val_accuracy: 0.9689\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0568 - accuracy: 0.9823 - val_loss: 0.1021 - val_accuracy: 0.9718\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0467 - accuracy: 0.9860 - val_loss: 0.1157 - val_accuracy: 0.9678\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.0381 - accuracy: 0.9880 - val_loss: 0.1001 - val_accuracy: 0.9769\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.0311 - accuracy: 0.9904 - val_loss: 0.1233 - val_accuracy: 0.9730\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.0322 - accuracy: 0.9899 - val_loss: 0.1247 - val_accuracy: 0.9752\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.0273 - accuracy: 0.9914 - val_loss: 0.1225 - val_accuracy: 0.9760\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.0252 - accuracy: 0.9924 - val_loss: 0.1097 - val_accuracy: 0.9760\n"
     ]
    }
   ],
   "source": [
    "train_history =model.fit(x=x_Train_normalize,\n",
    "                         y=y_Train_OneHot,validation_split=0.2, \n",
    "                         epochs=10, batch_size=32,verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "pV1Ftdb0PloP"
   },
   "source": [
    "# 以圖形顯示訓練過程"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "EEmuoa1dPloQ"
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "def show_train_history(train_history,train,validation):\n",
    "    plt.plot(train_history.history[train])\n",
    "    plt.plot(train_history.history[validation])\n",
    "    plt.title('Train History')\n",
    "    plt.ylabel(train)\n",
    "    plt.xlabel('Epoch')\n",
    "    plt.legend(['train', 'validation'], loc='upper left')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 376
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1451,
     "status": "ok",
     "timestamp": 1533441789030,
     "user": {
      "displayName": "Edward Cheng",
      "photoUrl": "//lh4.googleusercontent.com/-ANnwjGu3IBk/AAAAAAAAAAI/AAAAAAAAAAc/qXXg6Jek9xw/s50-c-k-no/photo.jpg",
      "userId": "112554017642991017343"
     },
     "user_tz": -480
    },
    "id": "fvE2WsejPloT",
    "outputId": "bb56e170-8850-409e-bbb2-eb6b5b89e528"
   },
   "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": [
    "show_train_history(train_history,'accuracy','val_accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 376
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1937,
     "status": "ok",
     "timestamp": 1533441791049,
     "user": {
      "displayName": "Edward Cheng",
      "photoUrl": "//lh4.googleusercontent.com/-ANnwjGu3IBk/AAAAAAAAAAI/AAAAAAAAAAc/qXXg6Jek9xw/s50-c-k-no/photo.jpg",
      "userId": "112554017642991017343"
     },
     "user_tz": -480
    },
    "id": "zIYyrnFWPloW",
    "outputId": "4bdbd1fa-6f25-49af-88d0-8d35f6e2ecaa"
   },
   "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": [
    "show_train_history(train_history,'loss','val_loss')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "h1z2y0g-PloZ"
   },
   "source": [
    "# 評估模型準確率"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1425,
     "status": "ok",
     "timestamp": 1533441792584,
     "user": {
      "displayName": "Edward Cheng",
      "photoUrl": "//lh4.googleusercontent.com/-ANnwjGu3IBk/AAAAAAAAAAI/AAAAAAAAAAc/qXXg6Jek9xw/s50-c-k-no/photo.jpg",
      "userId": "112554017642991017343"
     },
     "user_tz": -480
    },
    "id": "vMawurJqPloZ",
    "outputId": "31465f8d-cfad-49b8-dc9e-90d643e6b4a8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "313/313 [==============================] - 0s 811us/step - loss: 0.0994 - accuracy: 0.9786\n",
      "\n",
      "accuracy= 0.978600025177002\n"
     ]
    }
   ],
   "source": [
    "scores = model.evaluate(x_Test_normalize, y_Test_OneHot)\n",
    "print()\n",
    "print('accuracy=',scores[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "再新增兩層下準確度反而下降了"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "default_view": {},
   "name": "Keras_Mnist_MLP_h256.ipynb",
   "provenance": [],
   "version": "0.3.2",
   "views": {}
  },
  "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}