material/dnn_as_a_classifier.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "dnn_as_a_classifier.ipynb",
      "version": "0.3.2",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bCKgjM9-xygH",
        "colab_type": "text"
      },
      "source": [
        "# 분류기로서의 DNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "etOyOaq6xxHv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import time"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OmBdLDqix7Jh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def build_square_data(all_low, all_high, square_low, square_high, zero_count, one_count):\n",
        "\n",
        "  x1 = np.random.rand(one_count)\n",
        "  x1 = x1*(square_high-square_low) + square_low\n",
        "\n",
        "  x0 = np.random.rand(zero_count*100)\n",
        "  x0 = x0[ (x0<square_low) | (square_high<x0) ]\n",
        "  x0 = x0[:zero_count]\n",
        "\n",
        "  x = np.append(x1, x0)\n",
        "\n",
        "  plt.hist(x)\n",
        "  plt.xlim(0, 1)\n",
        "  plt.show()\n",
        "\n",
        "  # 범위 안의 것을 1, 범위 밖의 것을 0으로 하고\n",
        "  all_data = np.ones((len(x),2))\n",
        "  all_data[:,0] = x\n",
        "\n",
        "  all_data[:len(x1),1] = 1\n",
        "  all_data[len(x1):,1] = 0\n",
        "\n",
        "  # 섞는다\n",
        "  np.random.shuffle(all_data)\n",
        "\n",
        "  # x, y로 분리하고\n",
        "  x = all_data[:,0]\n",
        "  y = all_data[:,1]\n",
        "\n",
        "  plt.xlim(0, 1)\n",
        "\n",
        "  plt.scatter(x, y)\n",
        "\n",
        "  return x, y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HJ5rov-EyCN4",
        "colab_type": "text"
      },
      "source": [
        "# 네모 함수의 학습\n",
        "\n",
        "특정 영역은 1, 이외의 영역은 0을 출력\n",
        "\n",
        "1개의 출력 노드를 갖는다."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "10wyiaaEx_Eo",
        "colab_type": "code",
        "outputId": "7fac232f-30aa-42fd-98fc-0c66a5f85200",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 521
        }
      },
      "source": [
        "all_low = 0\n",
        "all_high = 1\n",
        "square_low = 0.5\n",
        "square_high = 0.6\n",
        "zero_count = 10000\n",
        "one_count = 10000\n",
        "\n",
        "train_x, train_y = build_square_data(all_low, all_high, square_low, square_high, zero_count, one_count)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAD8CAYAAACGsIhGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nsG7mf/PzAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K7fuP8OIyJBP",
        "colab_type": "code",
        "outputId": "eb36ab46-b51a-47f6-e015-633336a42cca",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras import optimizers\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Dense(10, activation='relu', input_shape=(1,)))\n",
        "model.add(Dense(10, activation='relu'))\n",
        "model.add(Dense(10, activation='relu'))\n",
        "model.add(Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "\n",
        "model.compile(optimizer=\"SGD\", loss=\"mse\")\n",
        "model.fit(train_x, train_y, epochs=100, verbose=1, batch_size=32, validation_split=0.1)\n",
        "\n",
        "x_ = np.arange(all_low, all_high, 0.01)\n",
        "y_ = model.predict(x_)\n",
        "plt.scatter(x_, y_)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 18000 samples, validate on 2000 samples\n",
            "Epoch 1/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.2507 - val_loss: 0.2505\n",
            "Epoch 2/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.2499 - val_loss: 0.2494\n",
            "Epoch 3/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.2486 - val_loss: 0.2482\n",
            "Epoch 4/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.2444 - val_loss: 0.2405\n",
            "Epoch 5/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.2364 - val_loss: 0.2323\n",
            "Epoch 6/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.2291 - val_loss: 0.2254\n",
            "Epoch 7/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.2204 - val_loss: 0.2153\n",
            "Epoch 8/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.2100 - val_loss: 0.2043\n",
            "Epoch 9/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.1982 - val_loss: 0.1920\n",
            "Epoch 10/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.1851 - val_loss: 0.1782\n",
            "Epoch 11/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.1706 - val_loss: 0.1633\n",
            "Epoch 12/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.1544 - val_loss: 0.1459\n",
            "Epoch 13/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.1357 - val_loss: 0.1265\n",
            "Epoch 14/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.1159 - val_loss: 0.1073\n",
            "Epoch 15/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0978 - val_loss: 0.0910\n",
            "Epoch 16/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0831 - val_loss: 0.0783\n",
            "Epoch 17/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0717 - val_loss: 0.0683\n",
            "Epoch 18/100\n",
            "18000/18000 [==============================] - 0s 28us/sample - loss: 0.0629 - val_loss: 0.0604\n",
            "Epoch 19/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0559 - val_loss: 0.0542\n",
            "Epoch 20/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0504 - val_loss: 0.0492\n",
            "Epoch 21/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0459 - val_loss: 0.0453\n",
            "Epoch 22/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0421 - val_loss: 0.0416\n",
            "Epoch 23/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0389 - val_loss: 0.0386\n",
            "Epoch 24/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0362 - val_loss: 0.0361\n",
            "Epoch 25/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0338 - val_loss: 0.0339\n",
            "Epoch 26/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0318 - val_loss: 0.0319\n",
            "Epoch 27/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0301 - val_loss: 0.0303\n",
            "Epoch 28/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0285 - val_loss: 0.0287\n",
            "Epoch 29/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0272 - val_loss: 0.0273\n",
            "Epoch 30/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0259 - val_loss: 0.0262\n",
            "Epoch 31/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0248 - val_loss: 0.0260\n",
            "Epoch 32/100\n",
            "18000/18000 [==============================] - 0s 28us/sample - loss: 0.0239 - val_loss: 0.0240\n",
            "Epoch 33/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0229 - val_loss: 0.0232\n",
            "Epoch 34/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0221 - val_loss: 0.0223\n",
            "Epoch 35/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0213 - val_loss: 0.0216\n",
            "Epoch 36/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0206 - val_loss: 0.0211\n",
            "Epoch 37/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0200 - val_loss: 0.0201\n",
            "Epoch 38/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0194 - val_loss: 0.0200\n",
            "Epoch 39/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0188 - val_loss: 0.0190\n",
            "Epoch 40/100\n",
            "18000/18000 [==============================] - 1s 32us/sample - loss: 0.0183 - val_loss: 0.0184\n",
            "Epoch 41/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0178 - val_loss: 0.0179\n",
            "Epoch 42/100\n",
            "18000/18000 [==============================] - 0s 28us/sample - loss: 0.0174 - val_loss: 0.0174\n",
            "Epoch 43/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0169 - val_loss: 0.0171\n",
            "Epoch 44/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0165 - val_loss: 0.0165\n",
            "Epoch 45/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0161 - val_loss: 0.0164\n",
            "Epoch 46/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0158 - val_loss: 0.0158\n",
            "Epoch 47/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0154 - val_loss: 0.0154\n",
            "Epoch 48/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0151 - val_loss: 0.0151\n",
            "Epoch 49/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0148 - val_loss: 0.0152\n",
            "Epoch 50/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0145 - val_loss: 0.0147\n",
            "Epoch 51/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0143 - val_loss: 0.0146\n",
            "Epoch 52/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0139 - val_loss: 0.0140\n",
            "Epoch 53/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0137 - val_loss: 0.0137\n",
            "Epoch 54/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0135 - val_loss: 0.0135\n",
            "Epoch 55/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0132 - val_loss: 0.0133\n",
            "Epoch 56/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0130 - val_loss: 0.0130\n",
            "Epoch 57/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0127 - val_loss: 0.0132\n",
            "Epoch 58/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0126 - val_loss: 0.0129\n",
            "Epoch 59/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0124 - val_loss: 0.0126\n",
            "Epoch 60/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0122 - val_loss: 0.0130\n",
            "Epoch 61/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0121 - val_loss: 0.0120\n",
            "Epoch 62/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0119 - val_loss: 0.0123\n",
            "Epoch 63/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0116 - val_loss: 0.0136\n",
            "Epoch 64/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0116 - val_loss: 0.0118\n",
            "Epoch 65/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0114 - val_loss: 0.0118\n",
            "Epoch 66/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0112 - val_loss: 0.0127\n",
            "Epoch 67/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0111 - val_loss: 0.0111\n",
            "Epoch 68/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0109 - val_loss: 0.0110\n",
            "Epoch 69/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0108 - val_loss: 0.0110\n",
            "Epoch 70/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0107 - val_loss: 0.0112\n",
            "Epoch 71/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0105 - val_loss: 0.0106\n",
            "Epoch 72/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0104 - val_loss: 0.0106\n",
            "Epoch 73/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0103 - val_loss: 0.0103\n",
            "Epoch 74/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0101 - val_loss: 0.0104\n",
            "Epoch 75/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0101 - val_loss: 0.0101\n",
            "Epoch 76/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0100 - val_loss: 0.0102\n",
            "Epoch 77/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0099 - val_loss: 0.0099\n",
            "Epoch 78/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0098 - val_loss: 0.0104\n",
            "Epoch 79/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0097 - val_loss: 0.0101\n",
            "Epoch 80/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0096 - val_loss: 0.0097\n",
            "Epoch 81/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0095 - val_loss: 0.0098\n",
            "Epoch 82/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0093 - val_loss: 0.0094\n",
            "Epoch 83/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0093 - val_loss: 0.0093\n",
            "Epoch 84/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0092 - val_loss: 0.0095\n",
            "Epoch 85/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0091 - val_loss: 0.0092\n",
            "Epoch 86/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0090 - val_loss: 0.0092\n",
            "Epoch 87/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0089 - val_loss: 0.0096\n",
            "Epoch 88/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0089 - val_loss: 0.0089\n",
            "Epoch 89/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0087 - val_loss: 0.0089\n",
            "Epoch 90/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0087 - val_loss: 0.0087\n",
            "Epoch 91/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0086 - val_loss: 0.0087\n",
            "Epoch 92/100\n",
            "18000/18000 [==============================] - 0s 27us/sample - loss: 0.0086 - val_loss: 0.0096\n",
            "Epoch 93/100\n",
            "18000/18000 [==============================] - 1s 30us/sample - loss: 0.0085 - val_loss: 0.0099\n",
            "Epoch 94/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0084 - val_loss: 0.0090\n",
            "Epoch 95/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0083 - val_loss: 0.0086\n",
            "Epoch 96/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0083 - val_loss: 0.0084\n",
            "Epoch 97/100\n",
            "18000/18000 [==============================] - 0s 28us/sample - loss: 0.0082 - val_loss: 0.0084\n",
            "Epoch 98/100\n",
            "18000/18000 [==============================] - 1s 29us/sample - loss: 0.0082 - val_loss: 0.0084\n",
            "Epoch 99/100\n",
            "18000/18000 [==============================] - 1s 28us/sample - loss: 0.0081 - val_loss: 0.0082\n",
            "Epoch 100/100\n",
            "18000/18000 [==============================] - 1s 31us/sample - loss: 0.0080 - val_loss: 0.0082\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.collections.PathCollection at 0x7f279f7eca20>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rKbC_Lnyyavy",
        "colab_type": "text"
      },
      "source": [
        "특정 패턴인 경우 1, 다른 경우인 경우 0으로 분류한다.\n",
        "\n",
        "출력된 값이 0.5보다 크면 1로, 0.5보다 작으면 0으로 간주한다."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xjdw0YWYzOzf",
        "colab_type": "text"
      },
      "source": [
        "# 클래스 2개, 출력 노드 2개\n",
        "\n",
        "분류 카테고리의 개수 대로 출력노드를 갖고, one-hot-encoding된 값을 출력하면, 변별력이 커진다."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2LM2kPa60My1",
        "colab_type": "code",
        "outputId": "4ec48c69-9409-4a5f-c720-69d30cd39e6f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 521
        }
      },
      "source": [
        "all_low = 0\n",
        "all_high = 1\n",
        "square_low = 0.5\n",
        "square_high = 0.6\n",
        "zero_count = 10000\n",
        "one_count = 10000\n",
        "\n",
        "train_x, train_y = build_square_data(all_low, all_high, square_low, square_high, zero_count, one_count)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAD8CAYAAACGsIhGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PiFWYY4Z1B2D",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# ADD START\n",
        "reshaped_y = train_y.reshape((train_y.shape[0],1))\n",
        "train_y = np.append(reshaped_y, reshaped_y, axis=1)\n",
        "train_y[:,1] = 1-train_y[:,0]\n",
        "# ADD END"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1yBDVshEyMxb",
        "colab_type": "code",
        "outputId": "3240a046-5764-4f82-c9bb-c42dfb07a378",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras import optimizers\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Dense(10, activation='relu', input_shape=(1,)))\n",
        "model.add(Dense(10, activation='relu'))\n",
        "model.add(Dense(10, activation='relu'))\n",
        "# model.add(Dense(1))\n",
        "model.add(Dense(2))\n",
        "\n",
        "model.compile(optimizer=\"SGD\", loss=\"mse\")\n",
        "model.fit(train_x, train_y, epochs=100, verbose=1, batch_size=32, validation_split=0.1)\n",
        "\n",
        "x_ = np.arange(all_low, all_high, 0.01)\n",
        "y_ = model.predict(x_)\n",
        "plt.scatter(x_, y_[:,0])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 18000 samples, validate on 2000 samples\n",
            "Epoch 1/100\n",
            "18000/18000 [==============================] - 1s 50us/sample - loss: 0.2551 - val_loss: 0.2466\n",
            "Epoch 2/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.2430 - val_loss: 0.2389\n",
            "Epoch 3/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.2261 - val_loss: 0.2118\n",
            "Epoch 4/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.1948 - val_loss: 0.1705\n",
            "Epoch 5/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.1432 - val_loss: 0.1106\n",
            "Epoch 6/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0931 - val_loss: 0.0761\n",
            "Epoch 7/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0739 - val_loss: 0.0671\n",
            "Epoch 8/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0631 - val_loss: 0.0553\n",
            "Epoch 9/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0558 - val_loss: 0.0507\n",
            "Epoch 10/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0520 - val_loss: 0.0494\n",
            "Epoch 11/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0492 - val_loss: 0.0451\n",
            "Epoch 12/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0468 - val_loss: 0.0435\n",
            "Epoch 13/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0446 - val_loss: 0.0407\n",
            "Epoch 14/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0426 - val_loss: 0.0388\n",
            "Epoch 15/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0407 - val_loss: 0.0374\n",
            "Epoch 16/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0389 - val_loss: 0.0355\n",
            "Epoch 17/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0372 - val_loss: 0.0339\n",
            "Epoch 18/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0355 - val_loss: 0.0323\n",
            "Epoch 19/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0338 - val_loss: 0.0307\n",
            "Epoch 20/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0321 - val_loss: 0.0297\n",
            "Epoch 21/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0306 - val_loss: 0.0279\n",
            "Epoch 22/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0290 - val_loss: 0.0262\n",
            "Epoch 23/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0275 - val_loss: 0.0249\n",
            "Epoch 24/100\n",
            "18000/18000 [==============================] - 1s 38us/sample - loss: 0.0261 - val_loss: 0.0236\n",
            "Epoch 25/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0246 - val_loss: 0.0227\n",
            "Epoch 26/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0234 - val_loss: 0.0212\n",
            "Epoch 27/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0222 - val_loss: 0.0203\n",
            "Epoch 28/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0211 - val_loss: 0.0192\n",
            "Epoch 29/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0201 - val_loss: 0.0182\n",
            "Epoch 30/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0192 - val_loss: 0.0181\n",
            "Epoch 31/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0184 - val_loss: 0.0175\n",
            "Epoch 32/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0176 - val_loss: 0.0165\n",
            "Epoch 33/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0170 - val_loss: 0.0155\n",
            "Epoch 34/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0163 - val_loss: 0.0155\n",
            "Epoch 35/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0157 - val_loss: 0.0150\n",
            "Epoch 36/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0152 - val_loss: 0.0153\n",
            "Epoch 37/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0147 - val_loss: 0.0143\n",
            "Epoch 38/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0143 - val_loss: 0.0136\n",
            "Epoch 39/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0138 - val_loss: 0.0129\n",
            "Epoch 40/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0134 - val_loss: 0.0129\n",
            "Epoch 41/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0131 - val_loss: 0.0121\n",
            "Epoch 42/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0127 - val_loss: 0.0121\n",
            "Epoch 43/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0124 - val_loss: 0.0117\n",
            "Epoch 44/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0120 - val_loss: 0.0113\n",
            "Epoch 45/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0118 - val_loss: 0.0121\n",
            "Epoch 46/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0115 - val_loss: 0.0109\n",
            "Epoch 47/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0112 - val_loss: 0.0105\n",
            "Epoch 48/100\n",
            "18000/18000 [==============================] - 1s 38us/sample - loss: 0.0109 - val_loss: 0.0107\n",
            "Epoch 49/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0108 - val_loss: 0.0101\n",
            "Epoch 50/100\n",
            "18000/18000 [==============================] - 1s 38us/sample - loss: 0.0106 - val_loss: 0.0102\n",
            "Epoch 51/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0104 - val_loss: 0.0098\n",
            "Epoch 52/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0102 - val_loss: 0.0097\n",
            "Epoch 53/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0099 - val_loss: 0.0094\n",
            "Epoch 54/100\n",
            "18000/18000 [==============================] - 1s 38us/sample - loss: 0.0098 - val_loss: 0.0094\n",
            "Epoch 55/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0097 - val_loss: 0.0091\n",
            "Epoch 56/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0095 - val_loss: 0.0095\n",
            "Epoch 57/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0093 - val_loss: 0.0095\n",
            "Epoch 58/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0092 - val_loss: 0.0087\n",
            "Epoch 59/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0090 - val_loss: 0.0088\n",
            "Epoch 60/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0090 - val_loss: 0.0085\n",
            "Epoch 61/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0088 - val_loss: 0.0086\n",
            "Epoch 62/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0088 - val_loss: 0.0089\n",
            "Epoch 63/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0086 - val_loss: 0.0080\n",
            "Epoch 64/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0085 - val_loss: 0.0089\n",
            "Epoch 65/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0084 - val_loss: 0.0090\n",
            "Epoch 66/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0083 - val_loss: 0.0077\n",
            "Epoch 67/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0081 - val_loss: 0.0081\n",
            "Epoch 68/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0081 - val_loss: 0.0080\n",
            "Epoch 69/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0080 - val_loss: 0.0079\n",
            "Epoch 70/100\n",
            "18000/18000 [==============================] - 1s 38us/sample - loss: 0.0078 - val_loss: 0.0073\n",
            "Epoch 71/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0078 - val_loss: 0.0075\n",
            "Epoch 72/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0077 - val_loss: 0.0104\n",
            "Epoch 73/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0077 - val_loss: 0.0082\n",
            "Epoch 74/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0075 - val_loss: 0.0071\n",
            "Epoch 75/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0074 - val_loss: 0.0069\n",
            "Epoch 76/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0074 - val_loss: 0.0086\n",
            "Epoch 77/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0073 - val_loss: 0.0072\n",
            "Epoch 78/100\n",
            "18000/18000 [==============================] - 1s 38us/sample - loss: 0.0073 - val_loss: 0.0069\n",
            "Epoch 79/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0072 - val_loss: 0.0080\n",
            "Epoch 80/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0072 - val_loss: 0.0076\n",
            "Epoch 81/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0070 - val_loss: 0.0068\n",
            "Epoch 82/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0070 - val_loss: 0.0065\n",
            "Epoch 83/100\n",
            "18000/18000 [==============================] - 1s 37us/sample - loss: 0.0070 - val_loss: 0.0066\n",
            "Epoch 84/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0068 - val_loss: 0.0074\n",
            "Epoch 85/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0067 - val_loss: 0.0066\n",
            "Epoch 86/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0068 - val_loss: 0.0077\n",
            "Epoch 87/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0067 - val_loss: 0.0065\n",
            "Epoch 88/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0067 - val_loss: 0.0105\n",
            "Epoch 89/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0065 - val_loss: 0.0080\n",
            "Epoch 90/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0066 - val_loss: 0.0063\n",
            "Epoch 91/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0064 - val_loss: 0.0070\n",
            "Epoch 92/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0063 - val_loss: 0.0063\n",
            "Epoch 93/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0063 - val_loss: 0.0059\n",
            "Epoch 94/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0062 - val_loss: 0.0086\n",
            "Epoch 95/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0062 - val_loss: 0.0059\n",
            "Epoch 96/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0061 - val_loss: 0.0058\n",
            "Epoch 97/100\n",
            "18000/18000 [==============================] - 1s 34us/sample - loss: 0.0061 - val_loss: 0.0062\n",
            "Epoch 98/100\n",
            "18000/18000 [==============================] - 1s 35us/sample - loss: 0.0062 - val_loss: 0.0073\n",
            "Epoch 99/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0060 - val_loss: 0.0064\n",
            "Epoch 100/100\n",
            "18000/18000 [==============================] - 1s 36us/sample - loss: 0.0060 - val_loss: 0.0078\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.collections.PathCollection at 0x7f71a573b6a0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nuyQV6P8BkBRxCcensEUAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KzTg8ocJ11NI",
        "colab_type": "text"
      },
      "source": [
        "출력 노드의 값은 0과 1이 서로 상반된다.\n",
        "\n",
        "하지만 이렇게 카테고리 개수 만큼 출력 노드가 있는 경우, 카테고리를 2개 이상으로 해도 처리가 가능해 진다."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MckR3PraC79y",
        "colab_type": "text"
      },
      "source": [
        "# 클래스 4개, 출력 노드 4개"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "s5MGpN31104d",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 442
        },
        "outputId": "ed0b94e3-bacb-4dad-fca1-b058e238424a"
      },
      "source": [
        "train_x = np.arange(0, 1, 0.001)\n",
        "\n",
        "class_index = np.zeros_like(train_x, dtype=np.int)\n",
        "class_index[:250] = 0\n",
        "class_index[250:500] = 1\n",
        "class_index[500:750] = 2\n",
        "class_index[750:] = 3\n",
        "\n",
        "\n",
        "plt.scatter(train_x, class_index)\n",
        "plt.show()\n",
        "\n",
        "train_y = np.zeros((1000, 4))\n",
        "train_y[np.arange(1000), class_index] = 1 # one-hot 인코딩\n",
        "\n",
        "print(train_y[245:255])"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[[1. 0. 0. 0.]\n",
            " [1. 0. 0. 0.]\n",
            " [1. 0. 0. 0.]\n",
            " [1. 0. 0. 0.]\n",
            " [1. 0. 0. 0.]\n",
            " [0. 1. 0. 0.]\n",
            " [0. 1. 0. 0.]\n",
            " [0. 1. 0. 0.]\n",
            " [0. 1. 0. 0.]\n",
            " [0. 1. 0. 0.]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yueRB3TCGBVx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "f1436c76-46ea-4ed6-d7d7-6ea6a0f24d42"
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras import optimizers\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Dense(10, activation='relu', input_shape=(1,)))\n",
        "model.add(Dense(10, activation='relu'))\n",
        "model.add(Dense(10, activation='relu'))\n",
        "model.add(Dense(4))\n",
        "\n",
        "model.compile(optimizer=\"SGD\", loss=\"mse\")\n",
        "model.fit(train_x, train_y, epochs=100, verbose=1, batch_size=1, validation_split=0.1)\n",
        "\n",
        "x_ = np.arange(0, 1, 0.01)\n",
        "y_ = model.predict(x_)\n",
        "for i in range(4):\n",
        "  plt.scatter(x_, y_[:,i])\n",
        "  plt.show()\n"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 900 samples, validate on 100 samples\n",
            "Epoch 1/100\n",
            "900/900 [==============================] - 1s 992us/sample - loss: 0.1732 - val_loss: 0.2329\n",
            "Epoch 2/100\n",
            "900/900 [==============================] - 1s 891us/sample - loss: 0.1384 - val_loss: 0.1961\n",
            "Epoch 3/100\n",
            "900/900 [==============================] - 1s 884us/sample - loss: 0.1252 - val_loss: 0.1584\n",
            "Epoch 4/100\n",
            "900/900 [==============================] - 1s 907us/sample - loss: 0.1196 - val_loss: 0.1817\n",
            "Epoch 5/100\n",
            "900/900 [==============================] - 1s 907us/sample - loss: 0.1145 - val_loss: 0.1166\n",
            "Epoch 6/100\n",
            "900/900 [==============================] - 1s 990us/sample - loss: 0.1073 - val_loss: 0.1098\n",
            "Epoch 7/100\n",
            "900/900 [==============================] - 1s 916us/sample - loss: 0.0998 - val_loss: 0.0910\n",
            "Epoch 8/100\n",
            "900/900 [==============================] - 1s 931us/sample - loss: 0.0920 - val_loss: 0.0732\n",
            "Epoch 9/100\n",
            "900/900 [==============================] - 1s 870us/sample - loss: 0.0831 - val_loss: 0.0550\n",
            "Epoch 10/100\n",
            "900/900 [==============================] - 1s 902us/sample - loss: 0.0755 - val_loss: 0.0796\n",
            "Epoch 11/100\n",
            "900/900 [==============================] - 1s 926us/sample - loss: 0.0698 - val_loss: 0.0640\n",
            "Epoch 12/100\n",
            "900/900 [==============================] - 1s 926us/sample - loss: 0.0650 - val_loss: 0.0575\n",
            "Epoch 13/100\n",
            "900/900 [==============================] - 1s 977us/sample - loss: 0.0608 - val_loss: 0.0477\n",
            "Epoch 14/100\n",
            "900/900 [==============================] - 1s 970us/sample - loss: 0.0574 - val_loss: 0.0459\n",
            "Epoch 15/100\n",
            "900/900 [==============================] - 1s 989us/sample - loss: 0.0535 - val_loss: 0.0418\n",
            "Epoch 16/100\n",
            "900/900 [==============================] - 1s 979us/sample - loss: 0.0504 - val_loss: 0.0449\n",
            "Epoch 17/100\n",
            "900/900 [==============================] - 1s 932us/sample - loss: 0.0464 - val_loss: 0.0320\n",
            "Epoch 18/100\n",
            "900/900 [==============================] - 1s 942us/sample - loss: 0.0427 - val_loss: 0.0359\n",
            "Epoch 19/100\n",
            "900/900 [==============================] - 1s 927us/sample - loss: 0.0386 - val_loss: 0.0730\n",
            "Epoch 20/100\n",
            "900/900 [==============================] - 1s 934us/sample - loss: 0.0358 - val_loss: 0.0290\n",
            "Epoch 21/100\n",
            "900/900 [==============================] - 1s 903us/sample - loss: 0.0340 - val_loss: 0.0434\n",
            "Epoch 22/100\n",
            "900/900 [==============================] - 1s 932us/sample - loss: 0.0322 - val_loss: 0.0304\n",
            "Epoch 23/100\n",
            "900/900 [==============================] - 1s 907us/sample - loss: 0.0310 - val_loss: 0.0298\n",
            "Epoch 24/100\n",
            "900/900 [==============================] - 1s 997us/sample - loss: 0.0294 - val_loss: 0.0949\n",
            "Epoch 25/100\n",
            "900/900 [==============================] - 1s 974us/sample - loss: 0.0285 - val_loss: 0.0696\n",
            "Epoch 26/100\n",
            "900/900 [==============================] - 1s 942us/sample - loss: 0.0272 - val_loss: 0.0532\n",
            "Epoch 27/100\n",
            "900/900 [==============================] - 1s 980us/sample - loss: 0.0277 - val_loss: 0.0854\n",
            "Epoch 28/100\n",
            "900/900 [==============================] - 1s 985us/sample - loss: 0.0272 - val_loss: 0.0425\n",
            "Epoch 29/100\n",
            "900/900 [==============================] - 1s 977us/sample - loss: 0.0262 - val_loss: 0.2126\n",
            "Epoch 30/100\n",
            "900/900 [==============================] - 1s 963us/sample - loss: 0.0265 - val_loss: 0.0688\n",
            "Epoch 31/100\n",
            "900/900 [==============================] - 1s 1ms/sample - loss: 0.0263 - val_loss: 0.1437\n",
            "Epoch 32/100\n",
            "900/900 [==============================] - 1s 960us/sample - loss: 0.0253 - val_loss: 0.2082\n",
            "Epoch 33/100\n",
            "900/900 [==============================] - 1s 930us/sample - loss: 0.0263 - val_loss: 0.0393\n",
            "Epoch 34/100\n",
            "900/900 [==============================] - 1s 942us/sample - loss: 0.0250 - val_loss: 0.1769\n",
            "Epoch 35/100\n",
            "900/900 [==============================] - 1s 932us/sample - loss: 0.0247 - val_loss: 0.0873\n",
            "Epoch 36/100\n",
            "900/900 [==============================] - 1s 962us/sample - loss: 0.0239 - val_loss: 0.1873\n",
            "Epoch 37/100\n",
            "900/900 [==============================] - 1s 978us/sample - loss: 0.0241 - val_loss: 0.2234\n",
            "Epoch 38/100\n",
            "900/900 [==============================] - 1s 922us/sample - loss: 0.0233 - val_loss: 0.2150\n",
            "Epoch 39/100\n",
            "900/900 [==============================] - 1s 926us/sample - loss: 0.0227 - val_loss: 0.0802\n",
            "Epoch 40/100\n",
            "900/900 [==============================] - 1s 1ms/sample - loss: 0.0235 - val_loss: 0.2117\n",
            "Epoch 41/100\n",
            "900/900 [==============================] - 1s 967us/sample - loss: 0.0233 - val_loss: 0.0878\n",
            "Epoch 42/100\n",
            "900/900 [==============================] - 1s 944us/sample - loss: 0.0237 - val_loss: 0.0845\n",
            "Epoch 43/100\n",
            "900/900 [==============================] - 1s 944us/sample - loss: 0.0233 - val_loss: 0.1883\n",
            "Epoch 44/100\n",
            "900/900 [==============================] - 1s 908us/sample - loss: 0.0231 - val_loss: 0.0933\n",
            "Epoch 45/100\n",
            "900/900 [==============================] - 1s 960us/sample - loss: 0.0228 - val_loss: 0.0985\n",
            "Epoch 46/100\n",
            "900/900 [==============================] - 1s 945us/sample - loss: 0.0228 - val_loss: 0.1861\n",
            "Epoch 47/100\n",
            "900/900 [==============================] - 1s 971us/sample - loss: 0.0222 - val_loss: 0.1048\n",
            "Epoch 48/100\n",
            "900/900 [==============================] - 1s 955us/sample - loss: 0.0225 - val_loss: 0.1307\n",
            "Epoch 49/100\n",
            "900/900 [==============================] - 1s 954us/sample - loss: 0.0217 - val_loss: 0.1778\n",
            "Epoch 50/100\n",
            "900/900 [==============================] - 1s 944us/sample - loss: 0.0217 - val_loss: 0.1635\n",
            "Epoch 51/100\n",
            "900/900 [==============================] - 1s 900us/sample - loss: 0.0216 - val_loss: 0.1311\n",
            "Epoch 52/100\n",
            "900/900 [==============================] - 1s 943us/sample - loss: 0.0217 - val_loss: 0.1584\n",
            "Epoch 53/100\n",
            "900/900 [==============================] - 1s 903us/sample - loss: 0.0220 - val_loss: 0.0961\n",
            "Epoch 54/100\n",
            "900/900 [==============================] - 1s 948us/sample - loss: 0.0216 - val_loss: 0.0760\n",
            "Epoch 55/100\n",
            "900/900 [==============================] - 1s 909us/sample - loss: 0.0218 - val_loss: 0.1165\n",
            "Epoch 56/100\n",
            "900/900 [==============================] - 1s 920us/sample - loss: 0.0216 - val_loss: 0.0882\n",
            "Epoch 57/100\n",
            "900/900 [==============================] - 1s 938us/sample - loss: 0.0201 - val_loss: 0.0449\n",
            "Epoch 58/100\n",
            "900/900 [==============================] - 1s 904us/sample - loss: 0.0196 - val_loss: 0.1257\n",
            "Epoch 59/100\n",
            "900/900 [==============================] - 1s 939us/sample - loss: 0.0204 - val_loss: 0.0780\n",
            "Epoch 60/100\n",
            "900/900 [==============================] - 1s 946us/sample - loss: 0.0195 - val_loss: 0.0242\n",
            "Epoch 61/100\n",
            "900/900 [==============================] - 1s 924us/sample - loss: 0.0196 - val_loss: 0.0174\n",
            "Epoch 62/100\n",
            "900/900 [==============================] - 1s 937us/sample - loss: 0.0184 - val_loss: 0.0522\n",
            "Epoch 63/100\n",
            "900/900 [==============================] - 1s 946us/sample - loss: 0.0188 - val_loss: 0.0522\n",
            "Epoch 64/100\n",
            "900/900 [==============================] - 1s 962us/sample - loss: 0.0176 - val_loss: 0.0875\n",
            "Epoch 65/100\n",
            "900/900 [==============================] - 1s 946us/sample - loss: 0.0192 - val_loss: 0.0908\n",
            "Epoch 66/100\n",
            "900/900 [==============================] - 1s 908us/sample - loss: 0.0180 - val_loss: 0.1241\n",
            "Epoch 67/100\n",
            "900/900 [==============================] - 1s 905us/sample - loss: 0.0176 - val_loss: 0.0317\n",
            "Epoch 68/100\n",
            "900/900 [==============================] - 1s 925us/sample - loss: 0.0178 - val_loss: 0.0544\n",
            "Epoch 69/100\n",
            "900/900 [==============================] - 1s 922us/sample - loss: 0.0190 - val_loss: 0.0603\n",
            "Epoch 70/100\n",
            "900/900 [==============================] - 1s 936us/sample - loss: 0.0169 - val_loss: 0.0843\n",
            "Epoch 71/100\n",
            "900/900 [==============================] - 1s 954us/sample - loss: 0.0173 - val_loss: 0.0774\n",
            "Epoch 72/100\n",
            "900/900 [==============================] - 1s 964us/sample - loss: 0.0174 - val_loss: 0.0859\n",
            "Epoch 73/100\n",
            "900/900 [==============================] - 1s 919us/sample - loss: 0.0161 - val_loss: 0.0240\n",
            "Epoch 74/100\n",
            "900/900 [==============================] - 1s 949us/sample - loss: 0.0173 - val_loss: 0.0303\n",
            "Epoch 75/100\n",
            "900/900 [==============================] - 1s 954us/sample - loss: 0.0169 - val_loss: 0.0904\n",
            "Epoch 76/100\n",
            "900/900 [==============================] - 1s 975us/sample - loss: 0.0167 - val_loss: 0.0808\n",
            "Epoch 77/100\n",
            "900/900 [==============================] - 1s 1ms/sample - loss: 0.0178 - val_loss: 0.1506\n",
            "Epoch 78/100\n",
            "900/900 [==============================] - 1s 936us/sample - loss: 0.0168 - val_loss: 0.0117\n",
            "Epoch 79/100\n",
            "900/900 [==============================] - 1s 911us/sample - loss: 0.0168 - val_loss: 0.0369\n",
            "Epoch 80/100\n",
            "900/900 [==============================] - 1s 945us/sample - loss: 0.0164 - val_loss: 0.0630\n",
            "Epoch 81/100\n",
            "900/900 [==============================] - 1s 965us/sample - loss: 0.0160 - val_loss: 0.0895\n",
            "Epoch 82/100\n",
            "900/900 [==============================] - 1s 971us/sample - loss: 0.0166 - val_loss: 0.1010\n",
            "Epoch 83/100\n",
            "900/900 [==============================] - 1s 956us/sample - loss: 0.0168 - val_loss: 0.0612\n",
            "Epoch 84/100\n",
            "900/900 [==============================] - 1s 955us/sample - loss: 0.0170 - val_loss: 0.0700\n",
            "Epoch 85/100\n",
            "900/900 [==============================] - 1s 939us/sample - loss: 0.0166 - val_loss: 0.1315\n",
            "Epoch 86/100\n",
            "900/900 [==============================] - 1s 946us/sample - loss: 0.0161 - val_loss: 0.0827\n",
            "Epoch 87/100\n",
            "900/900 [==============================] - 1s 953us/sample - loss: 0.0158 - val_loss: 0.0492\n",
            "Epoch 88/100\n",
            "900/900 [==============================] - 1s 936us/sample - loss: 0.0170 - val_loss: 0.0673\n",
            "Epoch 89/100\n",
            "900/900 [==============================] - 1s 943us/sample - loss: 0.0158 - val_loss: 0.0805\n",
            "Epoch 90/100\n",
            "900/900 [==============================] - 1s 934us/sample - loss: 0.0165 - val_loss: 0.0309\n",
            "Epoch 91/100\n",
            "900/900 [==============================] - 1s 962us/sample - loss: 0.0148 - val_loss: 0.0352\n",
            "Epoch 92/100\n",
            "900/900 [==============================] - 1s 918us/sample - loss: 0.0161 - val_loss: 0.1317\n",
            "Epoch 93/100\n",
            "900/900 [==============================] - 1s 931us/sample - loss: 0.0154 - val_loss: 0.1267\n",
            "Epoch 94/100\n",
            "900/900 [==============================] - 1s 939us/sample - loss: 0.0158 - val_loss: 0.0363\n",
            "Epoch 95/100\n",
            "900/900 [==============================] - 1s 933us/sample - loss: 0.0160 - val_loss: 0.0569\n",
            "Epoch 96/100\n",
            "900/900 [==============================] - 1s 908us/sample - loss: 0.0151 - val_loss: 0.0624\n",
            "Epoch 97/100\n",
            "900/900 [==============================] - 1s 911us/sample - loss: 0.0160 - val_loss: 0.1878\n",
            "Epoch 98/100\n",
            "900/900 [==============================] - 1s 915us/sample - loss: 0.0146 - val_loss: 0.0484\n",
            "Epoch 99/100\n",
            "900/900 [==============================] - 1s 922us/sample - loss: 0.0156 - val_loss: 0.1114\n",
            "Epoch 100/100\n",
            "900/900 [==============================] - 1s 959us/sample - loss: 0.0158 - val_loss: 0.1289\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCIo7CLccSfsAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAD8CAYAAABzTgP2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCweDmZm1+P+QJOfnf7VkYgAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAD8CAYAAABw1c+bAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "82mnxtU0OqBH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}