0411_3_mnist.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "0411-3",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/iammiori/AI_19_SS_CLASS/blob/master/0411_3_mnist.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "SwxA1tKp7Ad4",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "**Mnist 데이터 셋으로 손글씨 데이터 분류**\n",
        "- 입력 데이터는 28x28 영상데이터를 784차원의 1차원 벡터로 사용\n",
        "- 출력데이터는 숫자 정답\n",
        "\n",
        "### 1) 손글씨 데이터 정확도 올릴려면,\n",
        "- parameters 인 num_epochs 와 batch_size 변경\n",
        "      - num_epochs = 50, batch_size = 100"
      ]
    },
    {
      "metadata": {
        "id": "VGllBEqa7tI4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "ed882f0f-a707-4ffe-b15d-1b9cbc5cfc9d"
      },
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "import os\n",
        "\n",
        "if os.path.exists('/content/gdrive')==False:\n",
        "    drive.mount('/content/gdrive')\n",
        "    print('Google Drive is mounted\\n')\n",
        "else:\n",
        "    print('Google Drive is already mounted\\n')"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Google Drive is already mounted\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "ijfpFEZX7Fop",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "import random"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "tJZ5MwFq7jy9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1300
        },
        "outputId": "022240e2-6fc8-4a25-83e0-51b0e1fa8f1a"
      },
      "cell_type": "code",
      "source": [
        "# random seed 설정\n",
        "tf.set_random_seed(777)\n",
        "\n",
        "\n",
        "# 데이터 가져오기\n",
        "from tensorflow.examples.tutorials.mnist import input_data\n",
        "mnist = input_data.read_data_sets(\"MNIST_data/\", one_hot=True)\n",
        "# one_hot : label 데이터셋을 만들 때, label 을 one-hot 방식으로 처리한 것인지\n",
        "# default : False 라서 사용할거면 True 로 꼭 설정\n",
        "# one_hot encoding : 어 집합의 크기를 벡터의 크기로 하고, 표현하고 싶은 단어의 인덱스에 1의 값을 부여\n",
        "#다른 인덱스에는 0을 부여하는 단어의 벡터 표현\n",
        "\n",
        "nb_classes = 10\n",
        "\n",
        "# MNIST data image of shape 28 * 28 = 784\n",
        "# 784 차원\n",
        "# placeholder\n",
        "X = tf.placeholder(tf.float32, [None, 784])\n",
        "\n",
        "# 0 - 9 digits recognition = 10 classes\n",
        "Y = tf.placeholder(tf.float32, [None, nb_classes])\n",
        "\n",
        "# Weight, bias\n",
        "W = tf.Variable(tf.random_normal([784, nb_classes]))\n",
        "b = tf.Variable(tf.random_normal([nb_classes]))\n",
        "\n",
        "\n",
        "\n",
        "# softmax 사용해서 hypothesis\n",
        "hypothesis = tf.nn.softmax(tf.matmul(X, W) + b)\n",
        "cost = tf.reduce_mean(-tf.reduce_sum(Y * tf.log(hypothesis), axis=1))\n",
        "train = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(cost)\n",
        "\n",
        "\n",
        "\n",
        "# Test model\n",
        "is_correct = tf.equal(tf.argmax(hypothesis, 1), tf.argmax(Y, 1))\n",
        "accuracy = tf.reduce_mean(tf.cast(is_correct, tf.float32))\n",
        "\n",
        "# parameters\n",
        "num_epochs = 50  #15\n",
        "batch_size = 100\n",
        "num_iterations = int(mnist.train.num_examples / batch_size)\n",
        "# num_iterations = 1 Epoch 달성위해 배치 몇번 돌았는지 \n",
        "\n",
        "# 배치란?\n",
        "# Batch : 방대한 양의 데이터를 나눠서 갖고오고 싶을 때, batch 사용\n",
        "# tf.train.batch()\n",
        "# Epoch : 전체 데이터를 학습하는 횟수.\n",
        "# 말인즉슨, 1 Epoch = 전체 데이터 학습\n",
        "# ex ) 훈련 데이터가 1000개 batch_size = 200 일때,\n",
        "# 1 Epoch 달성하려면 batch를 5번 반복 (Iteration) 해야함.\n",
        "\n",
        "with tf.Session() as sess:\n",
        "    sess.run(tf.global_variables_initializer())\n",
        "    # Training cycle\n",
        "    for epoch in range(num_epochs):\n",
        "        avg_cost = 0\n",
        "\n",
        "        for i in range(num_iterations):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "            _, cost_val = sess.run([train, cost], feed_dict={X: batch_xs, Y: batch_ys})\n",
        "            avg_cost += cost_val / num_iterations\n",
        "\n",
        "        print(\"Epoch: {:04d}, Cost: {:.9f}\".format(epoch + 1, avg_cost))\n",
        "\n",
        "    print(\"Learning finished\")\n",
        "    \n",
        "    # 학습끝\n",
        "    \n",
        "    # Test\n",
        "\n",
        "    # Test the model using test sets\n",
        "    print(\n",
        "        \"Accuracy: \",\n",
        "        accuracy.eval(\n",
        "            session=sess, feed_dict={X: mnist.test.images, Y: mnist.test.labels}\n",
        "        ),\n",
        "    )\n",
        "\n",
        "    # Get one and predict\n",
        "    r = random.randint(0, mnist.test.num_examples - 1)\n",
        "    print(\"Label: \", sess.run(tf.argmax(mnist.test.labels[r : r + 1], 1)))\n",
        "    print(\n",
        "        \"Prediction: \",\n",
        "        sess.run(tf.argmax(hypothesis, 1), feed_dict={X: mnist.test.images[r : r + 1]}),\n",
        "    )\n",
        "\n",
        "    plt.imshow(\n",
        "        mnist.test.images[r : r + 1].reshape(28, 28),\n",
        "        cmap=\"Greys\",\n",
        "        interpolation=\"nearest\",\n",
        "    )\n",
        "    plt.show()\n"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
            "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
            "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
            "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n",
            "Epoch: 0001, Cost: 2.818828986\n",
            "Epoch: 0002, Cost: 1.085257435\n",
            "Epoch: 0003, Cost: 0.864798745\n",
            "Epoch: 0004, Cost: 0.755770931\n",
            "Epoch: 0005, Cost: 0.686584523\n",
            "Epoch: 0006, Cost: 0.637953204\n",
            "Epoch: 0007, Cost: 0.600986018\n",
            "Epoch: 0008, Cost: 0.571490138\n",
            "Epoch: 0009, Cost: 0.547181054\n",
            "Epoch: 0010, Cost: 0.527145614\n",
            "Epoch: 0011, Cost: 0.510084291\n",
            "Epoch: 0012, Cost: 0.494643601\n",
            "Epoch: 0013, Cost: 0.481214442\n",
            "Epoch: 0014, Cost: 0.469831739\n",
            "Epoch: 0015, Cost: 0.458875789\n",
            "Epoch: 0016, Cost: 0.449458207\n",
            "Epoch: 0017, Cost: 0.441382795\n",
            "Epoch: 0018, Cost: 0.433171758\n",
            "Epoch: 0019, Cost: 0.426198458\n",
            "Epoch: 0020, Cost: 0.419485727\n",
            "Epoch: 0021, Cost: 0.413063932\n",
            "Epoch: 0022, Cost: 0.407621409\n",
            "Epoch: 0023, Cost: 0.402141436\n",
            "Epoch: 0024, Cost: 0.396709330\n",
            "Epoch: 0025, Cost: 0.392051933\n",
            "Epoch: 0026, Cost: 0.387775649\n",
            "Epoch: 0027, Cost: 0.383594061\n",
            "Epoch: 0028, Cost: 0.379419347\n",
            "Epoch: 0029, Cost: 0.375396779\n",
            "Epoch: 0030, Cost: 0.372197021\n",
            "Epoch: 0031, Cost: 0.368657762\n",
            "Epoch: 0032, Cost: 0.365397841\n",
            "Epoch: 0033, Cost: 0.362073182\n",
            "Epoch: 0034, Cost: 0.359214230\n",
            "Epoch: 0035, Cost: 0.356533955\n",
            "Epoch: 0036, Cost: 0.353747148\n",
            "Epoch: 0037, Cost: 0.351162789\n",
            "Epoch: 0038, Cost: 0.348794804\n",
            "Epoch: 0039, Cost: 0.345864348\n",
            "Epoch: 0040, Cost: 0.343771012\n",
            "Epoch: 0041, Cost: 0.341535829\n",
            "Epoch: 0042, Cost: 0.339390279\n",
            "Epoch: 0043, Cost: 0.337265284\n",
            "Epoch: 0044, Cost: 0.335304714\n",
            "Epoch: 0045, Cost: 0.333256269\n",
            "Epoch: 0046, Cost: 0.331430508\n",
            "Epoch: 0047, Cost: 0.329733489\n",
            "Epoch: 0048, Cost: 0.328065174\n",
            "Epoch: 0049, Cost: 0.326323766\n",
            "Epoch: 0050, Cost: 0.324703795\n",
            "Learning finished\n",
            "Accuracy:  0.9113\n",
            "Label:  [4]\n",
            "Prediction:  [4]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4QeCIvxAUIQfCIrwA0H9DZOaIEU0/vyCAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    }
  ]
}