ML_lab03.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "ML_lab03.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyM1zeXdRWcOyDw/IMrb5vsH",
      "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/Sinyoung3016/MachineLearning_Tutorial/blob/master/ML_lab03.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9HWAmv1McN_T",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "cb99a69f-b59d-47e3-ce0e-264f3658be99"
      },
      "source": [
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "#1.data set 설정\n",
        "x_train = [1, 2, 3, 4]\n",
        "y_train = [1, 2, 3, 4]\n",
        "\n",
        "#2.Model 구성\n",
        "tf.model = tf.keras.Sequential()\n",
        "  # units == output shape, input_dim == input shape\n",
        "tf.model.add(tf.keras.layers.Dense(units=1, input_dim=1))\n",
        "\n",
        "#3. Model 학습과정 설정\n",
        "  # SGD == standard gradient descendent, lr == learning rate\n",
        "sgd = tf.keras.optimizers.SGD(lr=0.1)\n",
        "  # mse == mean_squared_error, 1/m * sig (y'-y)^2\n",
        "tf.model.compile(loss='mse', optimizer=sgd)\n",
        "  # prints summary of the model to the terminal\n",
        "tf.model.summary()\n",
        "\n",
        "#4. Model 학습\n",
        "  # fit() trains the model and returns history of train\n",
        "history = tf.model.fit(x_train, y_train, epochs=100)\n",
        "\n",
        "#6. Model 사용하기\n",
        "  # predict() returns predicted value\n",
        "y_predict = tf.model.predict(np.array([5, 4]))\n",
        "print(y_predict)\n",
        "\n",
        "  # Plot training & validation loss values\n",
        "plt.plot(history.history['loss'])\n",
        "plt.title('Model loss')\n",
        "plt.ylabel('Loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.legend(['Train', 'Test'], loc='upper left')\n",
        "plt.show()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense (Dense)                (None, 1)                 2         \n",
            "=================================================================\n",
            "Total params: 2\n",
            "Trainable params: 2\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 18.6340\n",
            "Epoch 2/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 8.3853\n",
            "Epoch 3/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 3.7827\n",
            "Epoch 4/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 1.7152\n",
            "Epoch 5/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.7859\n",
            "Epoch 6/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.3677\n",
            "Epoch 7/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.1791\n",
            "Epoch 8/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0936\n",
            "Epoch 9/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0544\n",
            "Epoch 10/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0361\n",
            "Epoch 11/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0272\n",
            "Epoch 12/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0225\n",
            "Epoch 13/100\n",
            "1/1 [==============================] - 0s 4ms/step - loss: 0.0198\n",
            "Epoch 14/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0181\n",
            "Epoch 15/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0167\n",
            "Epoch 16/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0156\n",
            "Epoch 17/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0146\n",
            "Epoch 18/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0137\n",
            "Epoch 19/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0129\n",
            "Epoch 20/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0122\n",
            "Epoch 21/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0114\n",
            "Epoch 22/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0108\n",
            "Epoch 23/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0101\n",
            "Epoch 24/100\n",
            "1/1 [==============================] - 0s 3ms/step - loss: 0.0095\n",
            "Epoch 25/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0090\n",
            "Epoch 26/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0084\n",
            "Epoch 27/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0079\n",
            "Epoch 28/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0075\n",
            "Epoch 29/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0070\n",
            "Epoch 30/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0066\n",
            "Epoch 31/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0062\n",
            "Epoch 32/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0059\n",
            "Epoch 33/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0055\n",
            "Epoch 34/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0052\n",
            "Epoch 35/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0049\n",
            "Epoch 36/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0046\n",
            "Epoch 37/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0043\n",
            "Epoch 38/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0041\n",
            "Epoch 39/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0038\n",
            "Epoch 40/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0036\n",
            "Epoch 41/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0034\n",
            "Epoch 42/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0032\n",
            "Epoch 43/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0030\n",
            "Epoch 44/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0028\n",
            "Epoch 45/100\n",
            "1/1 [==============================] - 0s 4ms/step - loss: 0.0027\n",
            "Epoch 46/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0025\n",
            "Epoch 47/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0024\n",
            "Epoch 48/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0022\n",
            "Epoch 49/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0021\n",
            "Epoch 50/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0020\n",
            "Epoch 51/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0018\n",
            "Epoch 52/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0017\n",
            "Epoch 53/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0016\n",
            "Epoch 54/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0015\n",
            "Epoch 55/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 0.0014\n",
            "Epoch 56/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0014\n",
            "Epoch 57/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0013\n",
            "Epoch 58/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0012\n",
            "Epoch 59/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0011\n",
            "Epoch 60/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0011\n",
            "Epoch 61/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 0.0010\n",
            "Epoch 62/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 9.4550e-04\n",
            "Epoch 63/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 8.8973e-04\n",
            "Epoch 64/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 8.3726e-04\n",
            "Epoch 65/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 7.8787e-04\n",
            "Epoch 66/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 7.4140e-04\n",
            "Epoch 67/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.9767e-04\n",
            "Epoch 68/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.5652e-04\n",
            "Epoch 69/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 6.1780e-04\n",
            "Epoch 70/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.8136e-04\n",
            "Epoch 71/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 5.4707e-04\n",
            "Epoch 72/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 5.1481e-04\n",
            "Epoch 73/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 4.8444e-04\n",
            "Epoch 74/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 4.5587e-04\n",
            "Epoch 75/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 4.2898e-04\n",
            "Epoch 76/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 4.0368e-04\n",
            "Epoch 77/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 3.7987e-04\n",
            "Epoch 78/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 3.5746e-04\n",
            "Epoch 79/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 3.3638e-04\n",
            "Epoch 80/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 3.1654e-04\n",
            "Epoch 81/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 2.9787e-04\n",
            "Epoch 82/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.8030e-04\n",
            "Epoch 83/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.6377e-04\n",
            "Epoch 84/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.4821e-04\n",
            "Epoch 85/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.3357e-04\n",
            "Epoch 86/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.1979e-04\n",
            "Epoch 87/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 2.0683e-04\n",
            "Epoch 88/100\n",
            "1/1 [==============================] - 0s 3ms/step - loss: 1.9463e-04\n",
            "Epoch 89/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 1.8315e-04\n",
            "Epoch 90/100\n",
            "1/1 [==============================] - 0s 4ms/step - loss: 1.7235e-04\n",
            "Epoch 91/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 1.6218e-04\n",
            "Epoch 92/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 1.5262e-04\n",
            "Epoch 93/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 1.4362e-04\n",
            "Epoch 94/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 1.3515e-04\n",
            "Epoch 95/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 1.2717e-04\n",
            "Epoch 96/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 1.1967e-04\n",
            "Epoch 97/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 1.1261e-04\n",
            "Epoch 98/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 1.0597e-04\n",
            "Epoch 99/100\n",
            "1/1 [==============================] - 0s 2ms/step - loss: 9.9721e-05\n",
            "Epoch 100/100\n",
            "1/1 [==============================] - 0s 1ms/step - loss: 9.3840e-05\n",
            "[[4.98389  ]\n",
            " [3.9917111]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fAlClJQ3ZIRo",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "##### Version01 #####\n",
        "import tensorflow.compat.v1 as tf\n",
        "import matplotlib.pyplot as plt #Graph\n",
        "tf.disable_v2_behavior()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5A2-i0Wyxlox",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "##Draw Graph\n",
        "\n",
        "X = [1,2,3]\n",
        "Y = [1,2,3]\n",
        "W = tf.placeholder(tf.float32)\n",
        "\n",
        "hypothesis = W * X #H(x)\n",
        "\n",
        "cost = tf.reduce_mean(tf.square(hypothesis - Y))#cost(W)\n",
        "\n",
        "sess = tf.Session()\n",
        "sess.run(tf.global_variables_initializer())\n",
        "\n",
        "W_val = []\n",
        "cost_val = [] \n",
        "\n",
        "for i in range(-20, 50):\n",
        "  feed_W = i * 0.1 #-3~5, 0.1의 간격으로 이동\n",
        "  curr_cost, curr_W = sess.run([cost, W], feed_dict = {W: feed_W})\n",
        "  W_val.append(curr_W)\n",
        "  cost_val.append(curr_cost)\n",
        "\n",
        "plt.plot( W_val, cost_val) #plot(X축, Y축)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FsreXYER5R8j",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "##User Gradient Function\n",
        "\n",
        "X_data = [1,2,3]\n",
        "Y_data = [1,2,3]\n",
        "W = tf.Variable(tf.random_normal([1]), name = 'weight')\n",
        "\n",
        "X = tf.placeholder(tf.float32)\n",
        "Y = tf.placeholder(tf.float32)\n",
        "\n",
        "hypothesis = X * W\n",
        "cost = tf.reduce_mean(tf.square(hypothesis - Y))\n",
        "\n",
        "#Minimize : Gradient Descent\n",
        "#W -= learning_rate * derivative\n",
        "learning_rate = 0.1 #배우는 속도\n",
        "gradient = tf.reduce_mean((W*X-Y) * X)\n",
        "descent = W - learning_rate * gradient\n",
        "update = W.assign(descent) #tensorflow는 assign으로 할당\n",
        "\n",
        "sess = tf.Session()\n",
        "sess.run(tf.global_variables_initializer())\n",
        "\n",
        "for step in range(21):\n",
        "  sess.run(update, feed_dict = {X : X_data, Y : Y_data})\n",
        "  print(step, sess.run(cost, feed_dict = {X : X_data, Y : Y_data }), sess.run(W))\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mTM4h1o--Fxv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "##Use GradientDescentOptimizer\n",
        "X = [1,2,3]\n",
        "Y = [1,2,3]\n",
        "W = tf.Variable(10.0)\n",
        "\n",
        "hypothesis = W * X #H(x)\n",
        "\n",
        "cost = tf.reduce_mean(tf.square(hypothesis - Y))#cost(W)\n",
        "#minimize\n",
        "optimizer = tf.train.GradientDescentOptimizer(learning_rate = 0.1)\n",
        "train = optimizer.minimize(cost)\n",
        "\n",
        "sess = tf.Session()\n",
        "sess.run(tf.global_variables_initializer())\n",
        "\n",
        "for step in range(100):\n",
        "  print(step, sess.run(W))\n",
        "  sess.run(train)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZklcZx_YAJZV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "##Compare user_gradient and apply_gradient\n",
        "\n",
        "X = [1,2,3]\n",
        "Y = [1,2,3]\n",
        "W = tf.Variable(5.0)\n",
        "\n",
        "hypothesis = W * X \n",
        "\n",
        "user_gradients = tf.reduce_mean( (W*X-Y)*X )*2 \n",
        "#원래 미분할때 2로 나뉘지만 간략화 가능\n",
        "#그냥 씀 하지만 지금은 컴퓨팅과 비교하는 것이므로 2를 곱함\n",
        "\n",
        "cost = tf.reduce_mean(tf.square(hypothesis - Y))\n",
        "optimizer = tf.train.GradientDescentOptimizer(learning_rate = 0.1)\n",
        "gvs = optimizer.compute_gradients(cost,[W])\n",
        "apply_gradients = optimizer.apply_gradients(gvs)\n",
        "\n",
        "sess = tf.Session()\n",
        "sess.run(tf.global_variables_initializer())\n",
        "\n",
        "for step in range(100):\n",
        "  print(step, sess.run([user_gradients, W, gvs]))\n",
        "  sess.run(apply_gradients)\n",
        "\n",
        "  #conclusion : same \n"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}