code/ch14/ch14.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 14장. 텐서플로의 구조 자세히 알아보기"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**아래 링크를 통해 이 노트북을 주피터 노트북 뷰어(nbviewer.jupyter.org)로 보거나 구글 코랩(colab.research.google.com)에서 실행할 수 있습니다.**\n",
    "\n",
    "<table class=\"tfo-notebook-buttons\" align=\"left\">\n",
    "  <td>\n",
    "    <a target=\"_blank\" href=\"https://nbviewer.jupyter.org/github/rickiepark/python-machine-learning-book-2nd-edition/blob/master/code/ch14/ch14.ipynb\"><img src=\"https://jupyter.org/assets/main-logo.svg\" width=\"28\" />주피터 노트북 뷰어로 보기</a>\n",
    "  </td>\n",
    "  <td>\n",
    "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/rickiepark/python-machine-learning-book-2nd-edition/blob/master/code/ch14/ch14.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />구글 코랩(Colab)에서 실행하기</a>\n",
    "  </td>\n",
    "</table>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`watermark`는 주피터 노트북에 사용하는 파이썬 패키지를 출력하기 위한 유틸리티입니다. `watermark` 패키지를 설치하려면 다음 셀의 주석을 제거한 뒤 실행하세요."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!pip install watermark"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "last updated: 2020-08-10 \n",
      "\n",
      "CPython 3.6.9\n",
      "IPython 7.13.0\n",
      "\n",
      "numpy 1.18.5\n",
      "tensorflow 2.3.0\n",
      "matplotlib 3.1.1\n"
     ]
    }
   ],
   "source": [
    "%load_ext watermark\n",
    "%watermark -u -d -v -p numpy,tensorflow,matplotlib"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**이 노트북을 실행하려면 텐서플로 2.0.0-alpha0 버전 이상이 필요합니다.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 텐서플로의 랭크와 텐서"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**텐서의 랭크와 크기를 확인하는 방법**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "크기: () (4,) (2, 2)\n",
      "랭크: 0 1 2\n"
     ]
    }
   ],
   "source": [
    "## t1, t2, t3 텐서를 정의합니다.\n",
    "t1 = tf.constant(np.pi)\n",
    "t2 = tf.constant([1, 2, 3, 4])\n",
    "t3 = tf.constant([[1, 2], [3, 4]])\n",
    "\n",
    "## 랭크를 구합니다.\n",
    "r1 = tf.rank(t1)\n",
    "r2 = tf.rank(t2)\n",
    "r3 = tf.rank(t3)\n",
    "\n",
    "## 크기를 구합니다\n",
    "s1 = t1.get_shape()\n",
    "s2 = t2.get_shape()\n",
    "s3 = t3.get_shape()\n",
    "print('크기:', s1, s2, s3)\n",
    "\n",
    "print('랭크:', \n",
    "      r1.numpy(), \n",
    "      r2.numpy(), \n",
    "      r3.numpy())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 텐서를 다차원 배열로 변환하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tf.Tensor(\n",
      "[[1.  2.  3.  3.5]\n",
      " [4.  5.  6.  6.5]\n",
      " [7.  8.  9.  9.5]], shape=(3, 4), dtype=float64)\n",
      "T1의 크기: (3, 4)\n",
      "T1의 크기: (3, 4)\n",
      "<tf.Variable 'Variable:0' shape=(3, 4) dtype=float64, numpy=\n",
      "array([[ 1.45032437, -0.4560495 , -0.28348079,  2.26741791],\n",
      "       [ 0.2198861 , -1.57854223,  1.48609383, -0.6573271 ],\n",
      "       [ 0.4818955 ,  0.06341537,  0.50654947,  0.14419544]])>\n",
      "<tf.Variable 'Variable:0' shape=(3,) dtype=float64, numpy=array([-0.06310468,  0.45087586, -0.6786785 ])>\n"
     ]
    }
   ],
   "source": [
    "arr = np.array([[1., 2., 3., 3.5],\n",
    "                [4., 5., 6., 6.5],\n",
    "                [7., 8., 9., 9.5]])\n",
    "T1 = tf.constant(arr)\n",
    "print(T1)\n",
    "s = T1.get_shape()\n",
    "print('T1의 크기:', s)\n",
    "print('T1의 크기:', T1.shape)\n",
    "T2 = tf.Variable(np.random.normal(size=s))\n",
    "print(T2)\n",
    "T3 = tf.Variable(np.random.normal(size=s[0]))\n",
    "print(T3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tf.Tensor([[[1.  2.  3.  3.5 4.  5.  6.  6.5 7.  8.  9.  9.5]]], shape=(1, 1, 12), dtype=float64)\n",
      "tf.Tensor(\n",
      "[[[1.  2.  3.  3.5]\n",
      "  [4.  5.  6.  6.5]\n",
      "  [7.  8.  9.  9.5]]], shape=(1, 3, 4), dtype=float64)\n"
     ]
    }
   ],
   "source": [
    "T4 = tf.reshape(T1, shape=[1, 1, -1])\n",
    "print(T4)\n",
    "T5 = tf.reshape(T1, shape=[1, 3, -1])\n",
    "print(T5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tf.Tensor(\n",
      "[[[1. ]\n",
      "  [4. ]\n",
      "  [7. ]]\n",
      "\n",
      " [[2. ]\n",
      "  [5. ]\n",
      "  [8. ]]\n",
      "\n",
      " [[3. ]\n",
      "  [6. ]\n",
      "  [9. ]]\n",
      "\n",
      " [[3.5]\n",
      "  [6.5]\n",
      "  [9.5]]], shape=(4, 3, 1), dtype=float64)\n",
      "tf.Tensor(\n",
      "[[[1.  4.  7. ]\n",
      "  [2.  5.  8. ]\n",
      "  [3.  6.  9. ]\n",
      "  [3.5 6.5 9.5]]], shape=(1, 4, 3), dtype=float64)\n"
     ]
    }
   ],
   "source": [
    "T6 = tf.transpose(T5, perm=[2, 1, 0])\n",
    "print(T6)\n",
    "T7 = tf.transpose(T5, perm=[0, 2, 1])\n",
    "print(T7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[<tf.Tensor: shape=(1, 3, 2), dtype=float64, numpy=\n",
      "array([[[1., 2.],\n",
      "        [4., 5.],\n",
      "        [7., 8.]]])>, <tf.Tensor: shape=(1, 3, 2), dtype=float64, numpy=\n",
      "array([[[3. , 3.5],\n",
      "        [6. , 6.5],\n",
      "        [9. , 9.5]]])>]\n"
     ]
    }
   ],
   "source": [
    "t5_splt = tf.split(T5, \n",
    "                   num_or_size_splits=2, \n",
    "                   axis=2)\n",
    "print(t5_splt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tf.Tensor(\n",
      "[[1.]\n",
      " [1.]\n",
      " [1.]\n",
      " [1.]\n",
      " [1.]], shape=(5, 1), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.]\n",
      " [0.]\n",
      " [0.]\n",
      " [0.]\n",
      " [0.]], shape=(5, 1), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[1.]\n",
      " [1.]\n",
      " [1.]\n",
      " [1.]\n",
      " [1.]\n",
      " [0.]\n",
      " [0.]\n",
      " [0.]\n",
      " [0.]\n",
      " [0.]], shape=(10, 1), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[1. 0.]\n",
      " [1. 0.]\n",
      " [1. 0.]\n",
      " [1. 0.]\n",
      " [1. 0.]], shape=(5, 2), dtype=float32)\n"
     ]
    }
   ],
   "source": [
    "t1 = tf.ones(shape=(5, 1), dtype=tf.float32)\n",
    "t2 = tf.zeros(shape=(5, 1), dtype=tf.float32)\n",
    "print(t1)\n",
    "print(t2)\n",
    "\n",
    "t3 = tf.concat([t1, t2], axis=0)\n",
    "print(t3)\n",
    "t4 = tf.concat([t1, t2], axis=1)\n",
    "print(t4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 텐서플로의 계산 그래프 이해하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2*(a-b)+c =>  1\n"
     ]
    }
   ],
   "source": [
    "a = tf.constant(1)\n",
    "b = tf.constant(2) \n",
    "c = tf.constant(3) \n",
    "\n",
    "z = 2*(a-b) + c\n",
    "\n",
    "print('2*(a-b)+c => ', z.numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2*(a-b)+c =>  1\n"
     ]
    }
   ],
   "source": [
    "## 텐서플로 1.x 방식\n",
    "g = tf.Graph()\n",
    " \n",
    "## 그래프에 노드를 추가합니다.\n",
    "with g.as_default():\n",
    "    a = tf.constant(1, name='a')\n",
    "    b = tf.constant(2, name='b') \n",
    "    c = tf.constant(3, name='c') \n",
    "\n",
    "    z = 2*(a-b) + c\n",
    "    \n",
    "## 그래프를 실행합니다.\n",
    "with tf.compat.v1.Session(graph=g) as sess:\n",
    "    print('2*(a-b)+c => ', sess.run(z))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Operation 'a' type=Const>,\n",
       " <tf.Operation 'b' type=Const>,\n",
       " <tf.Operation 'c' type=Const>,\n",
       " <tf.Operation 'sub' type=Sub>,\n",
       " <tf.Operation 'mul/x' type=Const>,\n",
       " <tf.Operation 'mul' type=Mul>,\n",
       " <tf.Operation 'add' type=AddV2>]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "g.get_operations()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "node {\n",
       "  name: \"a\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 1\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"b\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 2\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"c\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 3\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"sub\"\n",
       "  op: \"Sub\"\n",
       "  input: \"a\"\n",
       "  input: \"b\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"mul/x\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 2\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"mul\"\n",
       "  op: \"Mul\"\n",
       "  input: \"mul/x\"\n",
       "  input: \"sub\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"add\"\n",
       "  op: \"AddV2\"\n",
       "  input: \"mul\"\n",
       "  input: \"c\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "versions {\n",
       "  producer: 440\n",
       "}"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "g.as_graph_def()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2*(a-b)+c =>  1\n"
     ]
    }
   ],
   "source": [
    "@tf.function\n",
    "def simple_func():\n",
    "    a = tf.constant(1)\n",
    "    b = tf.constant(2) \n",
    "    c = tf.constant(3) \n",
    "\n",
    "    z = 2*(a-b) + c\n",
    "    return z\n",
    "\n",
    "print('2*(a-b)+c => ', simple_func().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'tensorflow.python.eager.def_function.Function'>\n"
     ]
    }
   ],
   "source": [
    "print(simple_func.__class__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2*(a-b)+c =>  1\n"
     ]
    }
   ],
   "source": [
    "def simple_func():\n",
    "    a = tf.constant(1, name='a')\n",
    "    b = tf.constant(2, name='b') \n",
    "    c = tf.constant(3, name='c') \n",
    "\n",
    "    z = 2*(a-b) + c\n",
    "    return z\n",
    "\n",
    "simple_func = tf.function(simple_func)\n",
    "\n",
    "print('2*(a-b)+c => ', simple_func().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Operation 'a' type=Const>,\n",
       " <tf.Operation 'b' type=Const>,\n",
       " <tf.Operation 'c' type=Const>,\n",
       " <tf.Operation 'sub' type=Sub>,\n",
       " <tf.Operation 'mul/x' type=Const>,\n",
       " <tf.Operation 'mul' type=Mul>,\n",
       " <tf.Operation 'add' type=AddV2>,\n",
       " <tf.Operation 'Identity' type=Identity>]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "con_func = simple_func.get_concrete_function()\n",
    "con_func.graph.get_operations()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "node {\n",
       "  name: \"a\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 1\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"b\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 2\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"c\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 3\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"sub\"\n",
       "  op: \"Sub\"\n",
       "  input: \"a\"\n",
       "  input: \"b\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"mul/x\"\n",
       "  op: \"Const\"\n",
       "  attr {\n",
       "    key: \"dtype\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "  attr {\n",
       "    key: \"value\"\n",
       "    value {\n",
       "      tensor {\n",
       "        dtype: DT_INT32\n",
       "        tensor_shape {\n",
       "        }\n",
       "        int_val: 2\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"mul\"\n",
       "  op: \"Mul\"\n",
       "  input: \"mul/x\"\n",
       "  input: \"sub\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"add\"\n",
       "  op: \"AddV2\"\n",
       "  input: \"mul\"\n",
       "  input: \"c\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "node {\n",
       "  name: \"Identity\"\n",
       "  op: \"Identity\"\n",
       "  input: \"add\"\n",
       "  attr {\n",
       "    key: \"T\"\n",
       "    value {\n",
       "      type: DT_INT32\n",
       "    }\n",
       "  }\n",
       "}\n",
       "versions {\n",
       "  producer: 440\n",
       "}"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "con_func.graph.as_graph_def()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 텐서플로의 변수"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<tf.Variable 'w1:0' shape=(2, 4) dtype=int64>\n"
     ]
    }
   ],
   "source": [
    "g1 = tf.Graph()\n",
    "\n",
    "with g1.as_default():\n",
    "    w1 = tf.Variable(np.array([[1, 2, 3, 4],\n",
    "                              [5, 6, 7, 8]]), name='w1')\n",
    "print(w1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Operation 'w1/Initializer/initial_value' type=Const>,\n",
       " <tf.Operation 'w1' type=VarHandleOp>,\n",
       " <tf.Operation 'w1/IsInitialized/VarIsInitializedOp' type=VarIsInitializedOp>,\n",
       " <tf.Operation 'w1/Assign' type=AssignVariableOp>,\n",
       " <tf.Operation 'w1/Read/ReadVariableOp' type=ReadVariableOp>]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "g1.get_operations()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "name: \"init\"\n",
      "op: \"NoOp\"\n",
      "input: \"^w1/Assign\"\n",
      "\n"
     ]
    }
   ],
   "source": [
    "with g1.as_default():\n",
    "    init = tf.compat.v1.global_variables_initializer()\n",
    "    print(init.node_def)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"add:0\", shape=(2, 4), dtype=int64)\n",
      "[[2 3 4 5]\n",
      " [6 7 8 9]]\n",
      "[[2 3 4 5]\n",
      " [6 7 8 9]]\n"
     ]
    }
   ],
   "source": [
    "with g1.as_default():\n",
    "    w1 = w1 + 1\n",
    "    print(w1)\n",
    "    \n",
    "with tf.compat.v1.Session(graph=g1) as sess:\n",
    "    init = tf.compat.v1.global_variables_initializer()\n",
    "    sess.run(init)\n",
    "    print(sess.run(w1))\n",
    "    print(sess.run(w1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[2 3 4 5]\n",
      " [6 7 8 9]]\n",
      "[[ 3  4  5  6]\n",
      " [ 7  8  9 10]]\n"
     ]
    }
   ],
   "source": [
    "g2 = tf.Graph()\n",
    "\n",
    "with g2.as_default():\n",
    "    w1 = tf.Variable(np.array([[1, 2, 3, 4],\n",
    "                              [5, 6, 7, 8]]), name='w1')\n",
    "    w1 = w1.assign(w1 + 1)\n",
    "\n",
    "with tf.compat.v1.Session(graph=g2) as sess:\n",
    "    init = tf.compat.v1.global_variables_initializer()\n",
    "    sess.run(init)\n",
    "    print(sess.run(w1))\n",
    "    print(sess.run(w1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<tf.Variable 'w2:0' shape=(2, 4) dtype=int64, numpy=\n",
      "array([[1, 2, 3, 4],\n",
      "       [5, 6, 7, 8]])>\n"
     ]
    }
   ],
   "source": [
    "w2 = tf.Variable(np.array([[1, 2, 3, 4],\n",
    "                          [5, 6, 7, 8]]), name='w2')\n",
    "print(w2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tf.Tensor(\n",
      "[[2 3 4 5]\n",
      " [6 7 8 9]], shape=(2, 4), dtype=int64)\n"
     ]
    }
   ],
   "source": [
    "print(w2 + 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[2 3 4 5]\n",
      " [6 7 8 9]]\n",
      "[[ 3  4  5  6]\n",
      " [ 7  8  9 10]]\n"
     ]
    }
   ],
   "source": [
    "w2.assign(w2 + 1)\n",
    "print(w2.numpy())\n",
    "w2.assign(w2 + 1)\n",
    "print(w2.numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<tf.Variable 'w2:0' shape=(2, 4) dtype=int64, numpy=\n",
      "array([[ 3,  4,  5,  6],\n",
      "       [ 7,  8,  9, 10]])>\n"
     ]
    }
   ],
   "source": [
    "print(w2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## tf.keras API 자세히 배우기"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sequential 모델"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "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": [
    "## 랜덤한 회귀용 예제 데이터셋을 만듭니다\n",
    "\n",
    "np.random.seed(0)\n",
    "\n",
    "def make_random_data():\n",
    "    x = np.random.uniform(low=-2, high=2, size=200)\n",
    "    y = []\n",
    "    for t in x:\n",
    "        r = np.random.normal(loc=0.0, \n",
    "                             scale=(0.5 + t*t/3), \n",
    "                             size=None)\n",
    "        y.append(r)\n",
    "    return  x, 1.726*x -0.84 + np.array(y)\n",
    "\n",
    "\n",
    "x, y = make_random_data() \n",
    "\n",
    "plt.plot(x, y, 'o')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, y_train = x[:150], y[:150]\n",
    "x_test, y_test = x[150:], y[150:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.Sequential()\n",
    "\n",
    "model.add(tf.keras.layers.Dense(units=1, input_dim=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "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"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "4/4 [==============================] - 0s 44ms/step - loss: 10.8291 - val_loss: 6.4637\n",
      "Epoch 2/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 8.7347 - val_loss: 5.4838\n",
      "Epoch 3/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 7.2769 - val_loss: 4.6149\n",
      "Epoch 4/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 5.9967 - val_loss: 3.8855\n",
      "Epoch 5/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 4.9239 - val_loss: 3.3142\n",
      "Epoch 6/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 4.0648 - val_loss: 2.8162\n",
      "Epoch 7/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 3.3542 - val_loss: 2.4385\n",
      "Epoch 8/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 2.8431 - val_loss: 2.1606\n",
      "Epoch 9/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 2.4361 - val_loss: 1.9055\n",
      "Epoch 10/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 2.0927 - val_loss: 1.7239\n",
      "Epoch 11/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.8433 - val_loss: 1.5708\n",
      "Epoch 12/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 1.6153 - val_loss: 1.4501\n",
      "Epoch 13/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.4508 - val_loss: 1.3633\n",
      "Epoch 14/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 1.3311 - val_loss: 1.2883\n",
      "Epoch 15/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.2253 - val_loss: 1.2160\n",
      "Epoch 16/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.1226 - val_loss: 1.1603\n",
      "Epoch 17/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.0543 - val_loss: 1.1101\n",
      "Epoch 18/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.9934 - val_loss: 1.0669\n",
      "Epoch 19/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.9458 - val_loss: 1.0416\n",
      "Epoch 20/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.9101 - val_loss: 1.0116\n",
      "Epoch 21/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8777 - val_loss: 0.9937\n",
      "Epoch 22/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8597 - val_loss: 0.9853\n",
      "Epoch 23/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8452 - val_loss: 0.9734\n",
      "Epoch 24/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8311 - val_loss: 0.9551\n",
      "Epoch 25/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8130 - val_loss: 0.9415\n",
      "Epoch 26/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8005 - val_loss: 0.9311\n",
      "Epoch 27/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7874 - val_loss: 0.9229\n",
      "Epoch 28/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7789 - val_loss: 0.9203\n",
      "Epoch 29/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7751 - val_loss: 0.9164\n",
      "Epoch 30/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7718 - val_loss: 0.9135\n",
      "Epoch 31/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7701 - val_loss: 0.9070\n",
      "Epoch 32/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7662 - val_loss: 0.9039\n",
      "Epoch 33/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7646 - val_loss: 0.9069\n",
      "Epoch 34/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7635 - val_loss: 0.9044\n",
      "Epoch 35/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7630 - val_loss: 0.9021\n",
      "Epoch 36/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7622 - val_loss: 0.9026\n",
      "Epoch 37/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7643 - val_loss: 0.8994\n",
      "Epoch 38/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7605 - val_loss: 0.9070\n",
      "Epoch 39/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7599 - val_loss: 0.9094\n",
      "Epoch 40/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7602 - val_loss: 0.9074\n",
      "Epoch 41/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7586 - val_loss: 0.9022\n",
      "Epoch 42/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7586 - val_loss: 0.9028\n",
      "Epoch 43/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7580 - val_loss: 0.9005\n",
      "Epoch 44/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9015\n",
      "Epoch 45/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7571 - val_loss: 0.9008\n",
      "Epoch 46/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9030\n",
      "Epoch 47/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9010\n",
      "Epoch 48/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9024\n",
      "Epoch 49/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9019\n",
      "Epoch 50/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.9005\n",
      "Epoch 51/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.9036\n",
      "Epoch 52/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9042\n",
      "Epoch 53/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9033\n",
      "Epoch 54/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9049\n",
      "Epoch 55/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9093\n",
      "Epoch 56/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9101\n",
      "Epoch 57/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9090\n",
      "Epoch 58/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7578 - val_loss: 0.9024\n",
      "Epoch 59/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.9026\n",
      "Epoch 60/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9022\n",
      "Epoch 61/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7570 - val_loss: 0.9009\n",
      "Epoch 62/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9030\n",
      "Epoch 63/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9022\n",
      "Epoch 64/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9021\n",
      "Epoch 65/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8982\n",
      "Epoch 66/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8956\n",
      "Epoch 67/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.8965\n",
      "Epoch 68/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.8977\n",
      "Epoch 69/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.8985\n",
      "Epoch 70/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7578 - val_loss: 0.8992\n",
      "Epoch 71/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8961\n",
      "Epoch 72/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.8990\n",
      "Epoch 73/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.8983\n",
      "Epoch 74/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.8996\n",
      "Epoch 75/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.8991\n",
      "Epoch 76/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7588 - val_loss: 0.8994\n",
      "Epoch 77/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.8976\n",
      "Epoch 78/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9005\n",
      "Epoch 79/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.9003\n",
      "Epoch 80/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8991\n",
      "Epoch 81/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8963\n",
      "Epoch 82/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.8977\n",
      "Epoch 83/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7577 - val_loss: 0.8958\n",
      "Epoch 84/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.8955\n",
      "Epoch 85/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7576 - val_loss: 0.8958\n",
      "Epoch 86/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7575 - val_loss: 0.8949\n",
      "Epoch 87/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.8952\n",
      "Epoch 88/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.8964\n",
      "Epoch 89/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7563 - val_loss: 0.9003\n",
      "Epoch 90/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9065\n",
      "Epoch 91/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7572 - val_loss: 0.9033\n",
      "Epoch 92/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9002\n",
      "Epoch 93/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7578 - val_loss: 0.8975\n",
      "Epoch 94/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8978\n",
      "Epoch 95/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.8972\n",
      "Epoch 96/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7586 - val_loss: 0.8935\n",
      "Epoch 97/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.8922\n",
      "Epoch 98/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7576 - val_loss: 0.8893\n",
      "Epoch 99/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7579 - val_loss: 0.8902\n",
      "Epoch 100/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.8927\n",
      "Epoch 101/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8908\n",
      "Epoch 102/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7579 - val_loss: 0.8906\n",
      "Epoch 103/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7579 - val_loss: 0.8914\n",
      "Epoch 104/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7573 - val_loss: 0.8961\n",
      "Epoch 105/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.8966\n",
      "Epoch 106/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.8961\n",
      "Epoch 107/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7577 - val_loss: 0.8976\n",
      "Epoch 108/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.8988\n",
      "Epoch 109/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.8983\n",
      "Epoch 110/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9027\n",
      "Epoch 111/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7559 - val_loss: 0.9006\n",
      "Epoch 112/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9048\n",
      "Epoch 113/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9066\n",
      "Epoch 114/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9046\n",
      "Epoch 115/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9039\n",
      "Epoch 116/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9037\n",
      "Epoch 117/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.8994\n",
      "Epoch 118/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8994\n",
      "Epoch 119/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9026\n",
      "Epoch 120/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9004\n",
      "Epoch 121/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.8995\n",
      "Epoch 122/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8972\n",
      "Epoch 123/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9000\n",
      "Epoch 124/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.8985\n",
      "Epoch 125/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.8979\n",
      "Epoch 126/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8993\n",
      "Epoch 127/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9022\n",
      "Epoch 128/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9026\n",
      "Epoch 129/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7573 - val_loss: 0.9018\n",
      "Epoch 130/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9046\n",
      "Epoch 131/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9063\n",
      "Epoch 132/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9035\n",
      "Epoch 133/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.9009\n",
      "Epoch 134/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9006\n",
      "Epoch 135/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7580 - val_loss: 0.9016\n",
      "Epoch 136/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9012\n",
      "Epoch 137/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.8991\n",
      "Epoch 138/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9029\n",
      "Epoch 139/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.9007\n",
      "Epoch 140/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9008\n",
      "Epoch 141/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9014\n",
      "Epoch 142/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9011\n",
      "Epoch 143/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.8998\n",
      "Epoch 144/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9007\n",
      "Epoch 145/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9048\n",
      "Epoch 146/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.9094\n",
      "Epoch 147/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9058\n",
      "Epoch 148/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9059\n",
      "Epoch 149/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9082\n",
      "Epoch 150/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9081\n",
      "Epoch 151/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9053\n",
      "Epoch 152/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.9035\n",
      "Epoch 153/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9022\n",
      "Epoch 154/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9040\n",
      "Epoch 155/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9030\n",
      "Epoch 156/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9041\n",
      "Epoch 157/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7596 - val_loss: 0.8998\n",
      "Epoch 158/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7583 - val_loss: 0.9015\n",
      "Epoch 159/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7597 - val_loss: 0.9076\n",
      "Epoch 160/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9065\n",
      "Epoch 161/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7575 - val_loss: 0.9048\n",
      "Epoch 162/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.9047\n",
      "Epoch 163/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9065\n",
      "Epoch 164/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9092\n",
      "Epoch 165/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9086\n",
      "Epoch 166/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9106\n",
      "Epoch 167/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9073\n",
      "Epoch 168/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9074\n",
      "Epoch 169/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9063\n",
      "Epoch 170/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.9059\n",
      "Epoch 171/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9045\n",
      "Epoch 172/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9073\n",
      "Epoch 173/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9075\n",
      "Epoch 174/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9107\n",
      "Epoch 175/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9098\n",
      "Epoch 176/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9088\n",
      "Epoch 177/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9090\n",
      "Epoch 178/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9078\n",
      "Epoch 179/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7559 - val_loss: 0.9063\n",
      "Epoch 180/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9043\n",
      "Epoch 181/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7556 - val_loss: 0.9048\n",
      "Epoch 182/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9010\n",
      "Epoch 183/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9034\n",
      "Epoch 184/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7556 - val_loss: 0.9024\n",
      "Epoch 185/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7578 - val_loss: 0.9079\n",
      "Epoch 186/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9059\n",
      "Epoch 187/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9056\n",
      "Epoch 188/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9031\n",
      "Epoch 189/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7559 - val_loss: 0.9012\n",
      "Epoch 190/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7557 - val_loss: 0.9038\n",
      "Epoch 191/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9033\n",
      "Epoch 192/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9037\n",
      "Epoch 193/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9017\n",
      "Epoch 194/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9062\n",
      "Epoch 195/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9055\n",
      "Epoch 196/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9049\n",
      "Epoch 197/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9064\n",
      "Epoch 198/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7577 - val_loss: 0.9058\n",
      "Epoch 199/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9049\n",
      "Epoch 200/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9060\n",
      "Epoch 201/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9054\n",
      "Epoch 202/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.9034\n",
      "Epoch 203/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9005\n",
      "Epoch 204/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.9041\n",
      "Epoch 205/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.9046\n",
      "Epoch 206/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9054\n",
      "Epoch 207/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9069\n",
      "Epoch 208/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.9041\n",
      "Epoch 209/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9017\n",
      "Epoch 210/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9018\n",
      "Epoch 211/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9039\n",
      "Epoch 212/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.8996\n",
      "Epoch 213/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.8992\n",
      "Epoch 214/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.8998\n",
      "Epoch 215/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8986\n",
      "Epoch 216/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9017\n",
      "Epoch 217/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9025\n",
      "Epoch 218/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9026\n",
      "Epoch 219/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9031\n",
      "Epoch 220/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.8979\n",
      "Epoch 221/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.8992\n",
      "Epoch 222/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.8976\n",
      "Epoch 223/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.8979\n",
      "Epoch 224/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.9073\n",
      "Epoch 225/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9066\n",
      "Epoch 226/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.9060\n",
      "Epoch 227/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.9040\n",
      "Epoch 228/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9061\n",
      "Epoch 229/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7572 - val_loss: 0.9039\n",
      "Epoch 230/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9018\n",
      "Epoch 231/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.9001\n",
      "Epoch 232/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7587 - val_loss: 0.8984\n",
      "Epoch 233/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.8994\n",
      "Epoch 234/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9007\n",
      "Epoch 235/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.8976\n",
      "Epoch 236/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8992\n",
      "Epoch 237/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9040\n",
      "Epoch 238/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9042\n",
      "Epoch 239/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.9036\n",
      "Epoch 240/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9028\n",
      "Epoch 241/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9027\n",
      "Epoch 242/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9027\n",
      "Epoch 243/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9093\n",
      "Epoch 244/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7581 - val_loss: 0.9091\n",
      "Epoch 245/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.9084\n",
      "Epoch 246/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9079\n",
      "Epoch 247/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9050\n",
      "Epoch 248/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9015\n",
      "Epoch 249/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7557 - val_loss: 0.9007\n",
      "Epoch 250/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9011\n",
      "Epoch 251/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9037\n",
      "Epoch 252/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7575 - val_loss: 0.9036\n",
      "Epoch 253/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9041\n",
      "Epoch 254/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7573 - val_loss: 0.9044\n",
      "Epoch 255/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.9056\n",
      "Epoch 256/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7573 - val_loss: 0.9076\n",
      "Epoch 257/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9065\n",
      "Epoch 258/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9048\n",
      "Epoch 259/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9081\n",
      "Epoch 260/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9080\n",
      "Epoch 261/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9082\n",
      "Epoch 262/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9037\n",
      "Epoch 263/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9070\n",
      "Epoch 264/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7559 - val_loss: 0.9044\n",
      "Epoch 265/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9036\n",
      "Epoch 266/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9044\n",
      "Epoch 267/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7555 - val_loss: 0.9053\n",
      "Epoch 268/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7557 - val_loss: 0.9070\n",
      "Epoch 269/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9045\n",
      "Epoch 270/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9036\n",
      "Epoch 271/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9031\n",
      "Epoch 272/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9018\n",
      "Epoch 273/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9058\n",
      "Epoch 274/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7556 - val_loss: 0.9081\n",
      "Epoch 275/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7559 - val_loss: 0.9063\n",
      "Epoch 276/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7571 - val_loss: 0.9072\n",
      "Epoch 277/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9063\n",
      "Epoch 278/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.9036\n",
      "Epoch 279/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7581 - val_loss: 0.9041\n",
      "Epoch 280/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7559 - val_loss: 0.9077\n",
      "Epoch 281/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9056\n",
      "Epoch 282/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9106\n",
      "Epoch 283/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9117\n",
      "Epoch 284/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9106\n",
      "Epoch 285/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9120\n",
      "Epoch 286/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9116\n",
      "Epoch 287/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9119\n",
      "Epoch 288/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.9203\n",
      "Epoch 289/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7580 - val_loss: 0.9217\n",
      "Epoch 290/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7582 - val_loss: 0.9235\n",
      "Epoch 291/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.9286\n",
      "Epoch 292/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7579 - val_loss: 0.9213\n",
      "Epoch 293/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.9173\n",
      "Epoch 294/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9151\n",
      "Epoch 295/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7577 - val_loss: 0.9145\n",
      "Epoch 296/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.9127\n",
      "Epoch 297/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9093\n",
      "Epoch 298/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9107\n",
      "Epoch 299/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9082\n",
      "Epoch 300/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9071\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='sgd', loss='mse')\n",
    "history = model.fit(x_train, y_train, epochs=300, \n",
    "                    validation_split=0.3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "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": [
    "epochs = np.arange(1, 300+1)\n",
    "plt.plot(epochs, history.history['loss'], label='Training loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 함수형 API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"functional_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 1)]               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 2         \n",
      "=================================================================\n",
      "Total params: 2\n",
      "Trainable params: 2\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input = tf.keras.Input(shape=(1,))\n",
    "output = tf.keras.layers.Dense(1)(input)\n",
    "\n",
    "model = tf.keras.Model(input, output)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "4/4 [==============================] - 0s 28ms/step - loss: 1.7650 - val_loss: 1.3711\n",
      "Epoch 2/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.6087 - val_loss: 1.2814\n",
      "Epoch 3/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.4741 - val_loss: 1.2025\n",
      "Epoch 4/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.3592 - val_loss: 1.1410\n",
      "Epoch 5/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 1.2661 - val_loss: 1.0851\n",
      "Epoch 6/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.1784 - val_loss: 1.0379\n",
      "Epoch 7/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.1037 - val_loss: 1.0073\n",
      "Epoch 8/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.0573 - val_loss: 0.9804\n",
      "Epoch 9/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 1.0102 - val_loss: 0.9566\n",
      "Epoch 10/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.9691 - val_loss: 0.9394\n",
      "Epoch 11/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.9372 - val_loss: 0.9244\n",
      "Epoch 12/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.9081 - val_loss: 0.9098\n",
      "Epoch 13/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8813 - val_loss: 0.9030\n",
      "Epoch 14/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8646 - val_loss: 0.8953\n",
      "Epoch 15/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8460 - val_loss: 0.8896\n",
      "Epoch 16/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8358 - val_loss: 0.8846\n",
      "Epoch 17/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8238 - val_loss: 0.8821\n",
      "Epoch 18/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8143 - val_loss: 0.8804\n",
      "Epoch 19/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8072 - val_loss: 0.8762\n",
      "Epoch 20/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7980 - val_loss: 0.8750\n",
      "Epoch 21/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7887 - val_loss: 0.8727\n",
      "Epoch 22/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7852 - val_loss: 0.8720\n",
      "Epoch 23/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7800 - val_loss: 0.8731\n",
      "Epoch 24/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7758 - val_loss: 0.8760\n",
      "Epoch 25/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7722 - val_loss: 0.8757\n",
      "Epoch 26/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7691 - val_loss: 0.8752\n",
      "Epoch 27/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7665 - val_loss: 0.8769\n",
      "Epoch 28/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7643 - val_loss: 0.8754\n",
      "Epoch 29/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7649 - val_loss: 0.8765\n",
      "Epoch 30/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7637 - val_loss: 0.8757\n",
      "Epoch 31/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7629 - val_loss: 0.8756\n",
      "Epoch 32/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7632 - val_loss: 0.8746\n",
      "Epoch 33/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7633 - val_loss: 0.8745\n",
      "Epoch 34/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7625 - val_loss: 0.8778\n",
      "Epoch 35/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7609 - val_loss: 0.8840\n",
      "Epoch 36/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7588 - val_loss: 0.8866\n",
      "Epoch 37/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7598 - val_loss: 0.8869\n",
      "Epoch 38/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7579 - val_loss: 0.8860\n",
      "Epoch 39/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7576 - val_loss: 0.8884\n",
      "Epoch 40/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7588 - val_loss: 0.8877\n",
      "Epoch 41/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7577 - val_loss: 0.8879\n",
      "Epoch 42/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7580 - val_loss: 0.8912\n",
      "Epoch 43/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7592 - val_loss: 0.8930\n",
      "Epoch 44/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7566 - val_loss: 0.8944\n",
      "Epoch 45/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7577 - val_loss: 0.8957\n",
      "Epoch 46/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9024\n",
      "Epoch 47/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9018\n",
      "Epoch 48/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7559 - val_loss: 0.8983\n",
      "Epoch 49/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7593 - val_loss: 0.9011\n",
      "Epoch 50/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.8999\n",
      "Epoch 51/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.8983\n",
      "Epoch 52/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.8978\n",
      "Epoch 53/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8982\n",
      "Epoch 54/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.8964\n",
      "Epoch 55/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.8970\n",
      "Epoch 56/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.8974\n",
      "Epoch 57/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9028\n",
      "Epoch 58/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9008\n",
      "Epoch 59/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.8989\n",
      "Epoch 60/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7576 - val_loss: 0.9002\n",
      "Epoch 61/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9040\n",
      "Epoch 62/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9056\n",
      "Epoch 63/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7579 - val_loss: 0.9021\n",
      "Epoch 64/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.9004\n",
      "Epoch 65/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7572 - val_loss: 0.9025\n",
      "Epoch 66/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9046\n",
      "Epoch 67/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9044\n",
      "Epoch 68/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9020\n",
      "Epoch 69/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9004\n",
      "Epoch 70/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8993\n",
      "Epoch 71/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9008\n",
      "Epoch 72/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9031\n",
      "Epoch 73/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9046\n",
      "Epoch 74/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9039\n",
      "Epoch 75/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.9028\n",
      "Epoch 76/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7578 - val_loss: 0.9032\n",
      "Epoch 77/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7570 - val_loss: 0.9020\n",
      "Epoch 78/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.9018\n",
      "Epoch 79/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7570 - val_loss: 0.9017\n",
      "Epoch 80/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9028\n",
      "Epoch 81/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9035\n",
      "Epoch 82/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9046\n",
      "Epoch 83/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7578 - val_loss: 0.9094\n",
      "Epoch 84/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.9105\n",
      "Epoch 85/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.9086\n",
      "Epoch 86/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7598 - val_loss: 0.9067\n",
      "Epoch 87/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.9040\n",
      "Epoch 88/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.9038\n",
      "Epoch 89/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.8998\n",
      "Epoch 90/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9036\n",
      "Epoch 91/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9078\n",
      "Epoch 92/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9053\n",
      "Epoch 93/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9063\n",
      "Epoch 94/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.9061\n",
      "Epoch 95/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9107\n",
      "Epoch 96/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9085\n",
      "Epoch 97/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9070\n",
      "Epoch 98/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9043\n",
      "Epoch 99/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9112\n",
      "Epoch 100/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9066\n",
      "Epoch 101/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9060\n",
      "Epoch 102/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7580 - val_loss: 0.8989\n",
      "Epoch 103/300\n",
      "4/4 [==============================] - 0s 11ms/step - loss: 0.7559 - val_loss: 0.9010\n",
      "Epoch 104/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9000\n",
      "Epoch 105/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9010\n",
      "Epoch 106/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7560 - val_loss: 0.8987\n",
      "Epoch 107/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.8965\n",
      "Epoch 108/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7569 - val_loss: 0.8989\n",
      "Epoch 109/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9000\n",
      "Epoch 110/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.8984\n",
      "Epoch 111/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.8998\n",
      "Epoch 112/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9004\n",
      "Epoch 113/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9004\n",
      "Epoch 114/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.8995\n",
      "Epoch 115/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9010\n",
      "Epoch 116/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9002\n",
      "Epoch 117/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.8994\n",
      "Epoch 118/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9008\n",
      "Epoch 119/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.8966\n",
      "Epoch 120/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8980\n",
      "Epoch 121/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9015\n",
      "Epoch 122/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7568 - val_loss: 0.9056\n",
      "Epoch 123/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9062\n",
      "Epoch 124/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9037\n",
      "Epoch 125/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7577 - val_loss: 0.9050\n",
      "Epoch 126/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9035\n",
      "Epoch 127/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.9042\n",
      "Epoch 128/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9064\n",
      "Epoch 129/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9039\n",
      "Epoch 130/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9035\n",
      "Epoch 131/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7562 - val_loss: 0.9030\n",
      "Epoch 132/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9024\n",
      "Epoch 133/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.9041\n",
      "Epoch 134/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.9004\n",
      "Epoch 135/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9007\n",
      "Epoch 136/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9025\n",
      "Epoch 137/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9005\n",
      "Epoch 138/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7571 - val_loss: 0.9006\n",
      "Epoch 139/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.8994\n",
      "Epoch 140/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9021\n",
      "Epoch 141/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7556 - val_loss: 0.9021\n",
      "Epoch 142/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9027\n",
      "Epoch 143/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9015\n",
      "Epoch 144/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9021\n",
      "Epoch 145/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.8984\n",
      "Epoch 146/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.8989\n",
      "Epoch 147/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7577 - val_loss: 0.8988\n",
      "Epoch 148/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.8999\n",
      "Epoch 149/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.8975\n",
      "Epoch 150/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9001\n",
      "Epoch 151/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9061\n",
      "Epoch 152/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7587 - val_loss: 0.9058\n",
      "Epoch 153/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7556 - val_loss: 0.9038\n",
      "Epoch 154/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9050\n",
      "Epoch 155/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9041\n",
      "Epoch 156/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9026\n",
      "Epoch 157/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9031\n",
      "Epoch 158/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7585 - val_loss: 0.9022\n",
      "Epoch 159/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9043\n",
      "Epoch 160/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7596 - val_loss: 0.9019\n",
      "Epoch 161/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9022\n",
      "Epoch 162/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9031\n",
      "Epoch 163/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9042\n",
      "Epoch 164/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9028\n",
      "Epoch 165/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9018\n",
      "Epoch 166/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9030\n",
      "Epoch 167/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7559 - val_loss: 0.9024\n",
      "Epoch 168/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9012\n",
      "Epoch 169/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9024\n",
      "Epoch 170/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9016\n",
      "Epoch 171/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.8988\n",
      "Epoch 172/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8976\n",
      "Epoch 173/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.8964\n",
      "Epoch 174/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.9006\n",
      "Epoch 175/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7586 - val_loss: 0.8998\n",
      "Epoch 176/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7575 - val_loss: 0.9006\n",
      "Epoch 177/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7573 - val_loss: 0.9005\n",
      "Epoch 178/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8994\n",
      "Epoch 179/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9030\n",
      "Epoch 180/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7570 - val_loss: 0.9032\n",
      "Epoch 181/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9057\n",
      "Epoch 182/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9044\n",
      "Epoch 183/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9020\n",
      "Epoch 184/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9010\n",
      "Epoch 185/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9016\n",
      "Epoch 186/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.9059\n",
      "Epoch 187/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9044\n",
      "Epoch 188/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9015\n",
      "Epoch 189/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9007\n",
      "Epoch 190/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8992\n",
      "Epoch 191/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.8973\n",
      "Epoch 192/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7576 - val_loss: 0.9013\n",
      "Epoch 193/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9038\n",
      "Epoch 194/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9044\n",
      "Epoch 195/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9042\n",
      "Epoch 196/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8998\n",
      "Epoch 197/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.8991\n",
      "Epoch 198/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.8987\n",
      "Epoch 199/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7575 - val_loss: 0.8955\n",
      "Epoch 200/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9001\n",
      "Epoch 201/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7601 - val_loss: 0.9010\n",
      "Epoch 202/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7575 - val_loss: 0.8982\n",
      "Epoch 203/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8967\n",
      "Epoch 204/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.8962\n",
      "Epoch 205/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.8971\n",
      "Epoch 206/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.8975\n",
      "Epoch 207/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.8983\n",
      "Epoch 208/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.8995\n",
      "Epoch 209/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9003\n",
      "Epoch 210/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.9008\n",
      "Epoch 211/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7559 - val_loss: 0.9026\n",
      "Epoch 212/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9000\n",
      "Epoch 213/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7570 - val_loss: 0.9021\n",
      "Epoch 214/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9016\n",
      "Epoch 215/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7577 - val_loss: 0.9005\n",
      "Epoch 216/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.9014\n",
      "Epoch 217/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8988\n",
      "Epoch 218/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9013\n",
      "Epoch 219/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7564 - val_loss: 0.9006\n",
      "Epoch 220/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.8989\n",
      "Epoch 221/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8944\n",
      "Epoch 222/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.8945\n",
      "Epoch 223/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8939\n",
      "Epoch 224/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.8948\n",
      "Epoch 225/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.8932\n",
      "Epoch 226/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7575 - val_loss: 0.8936\n",
      "Epoch 227/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8957\n",
      "Epoch 228/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8962\n",
      "Epoch 229/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.8975\n",
      "Epoch 230/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.8966\n",
      "Epoch 231/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7572 - val_loss: 0.8985\n",
      "Epoch 232/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9007\n",
      "Epoch 233/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.8993\n",
      "Epoch 234/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.8992\n",
      "Epoch 235/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9013\n",
      "Epoch 236/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9035\n",
      "Epoch 237/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9093\n",
      "Epoch 238/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9061\n",
      "Epoch 239/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9049\n",
      "Epoch 240/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9050\n",
      "Epoch 241/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9040\n",
      "Epoch 242/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7570 - val_loss: 0.9073\n",
      "Epoch 243/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9084\n",
      "Epoch 244/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9149\n",
      "Epoch 245/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.9136\n",
      "Epoch 246/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9156\n",
      "Epoch 247/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7566 - val_loss: 0.9164\n",
      "Epoch 248/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7582 - val_loss: 0.9126\n",
      "Epoch 249/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9103\n",
      "Epoch 250/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7573 - val_loss: 0.9021\n",
      "Epoch 251/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.8992\n",
      "Epoch 252/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7570 - val_loss: 0.9016\n",
      "Epoch 253/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7582 - val_loss: 0.8984\n",
      "Epoch 254/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8985\n",
      "Epoch 255/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9028\n",
      "Epoch 256/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9053\n",
      "Epoch 257/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7572 - val_loss: 0.9032\n",
      "Epoch 258/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9003\n",
      "Epoch 259/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7581 - val_loss: 0.9090\n",
      "Epoch 260/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9094\n",
      "Epoch 261/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9080\n",
      "Epoch 262/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9089\n",
      "Epoch 263/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9087\n",
      "Epoch 264/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9119\n",
      "Epoch 265/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9136\n",
      "Epoch 266/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9103\n",
      "Epoch 267/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.9083\n",
      "Epoch 268/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9093\n",
      "Epoch 269/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9042\n",
      "Epoch 270/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9000\n",
      "Epoch 271/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.8992\n",
      "Epoch 272/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9012\n",
      "Epoch 273/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.8995\n",
      "Epoch 274/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9009\n",
      "Epoch 275/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9004\n",
      "Epoch 276/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.9010\n",
      "Epoch 277/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8986\n",
      "Epoch 278/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7567 - val_loss: 0.8977\n",
      "Epoch 279/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7563 - val_loss: 0.8978\n",
      "Epoch 280/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8983\n",
      "Epoch 281/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7578 - val_loss: 0.9042\n",
      "Epoch 282/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9034\n",
      "Epoch 283/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7562 - val_loss: 0.9044\n",
      "Epoch 284/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9070\n",
      "Epoch 285/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7561 - val_loss: 0.9043\n",
      "Epoch 286/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9065\n",
      "Epoch 287/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9121\n",
      "Epoch 288/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9085\n",
      "Epoch 289/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9069\n",
      "Epoch 290/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9054\n",
      "Epoch 291/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7560 - val_loss: 0.9076\n",
      "Epoch 292/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7558 - val_loss: 0.9062\n",
      "Epoch 293/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9050\n",
      "Epoch 294/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7576 - val_loss: 0.9037\n",
      "Epoch 295/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.9032\n",
      "Epoch 296/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9064\n",
      "Epoch 297/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9057\n",
      "Epoch 298/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7569 - val_loss: 0.9077\n",
      "Epoch 299/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7564 - val_loss: 0.9078\n",
      "Epoch 300/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9096\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='sgd', loss='mse')\n",
    "history = model.fit(x_train, y_train, epochs=300, \n",
    "                    validation_split=0.3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": false
   },
   "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": [
    "plt.plot(epochs, history.history['loss'], label='Training loss')\n",
    "plt.plot(epochs, history.history['val_loss'], label='Validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## tf.keras 모델의 저장과 복원"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save_weights('simple_weights.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.Sequential()\n",
    "model.add(tf.keras.layers.Dense(units=1, input_dim=1))\n",
    "model.compile(optimizer='sgd', loss='mse')\n",
    "\n",
    "model.load_weights('simple_weights.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 [==============================] - 0s 2ms/step - loss: 0.9453\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.9453349113464355"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(x_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('simple_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 [==============================] - 0s 2ms/step - loss: 0.9453\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.9453349113464355"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = tf.keras.models.load_model('simple_model.h5')\n",
    "model.evaluate(x_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "4/4 [==============================] - 0s 35ms/step - loss: 5.8028 - val_loss: 3.0364\n",
      "Epoch 2/300\n",
      "4/4 [==============================] - 0s 13ms/step - loss: 4.8878 - val_loss: 2.5510\n",
      "Epoch 3/300\n",
      "4/4 [==============================] - 0s 17ms/step - loss: 4.1228 - val_loss: 2.1622\n",
      "Epoch 4/300\n",
      "4/4 [==============================] - 0s 13ms/step - loss: 3.5009 - val_loss: 1.8568\n",
      "Epoch 5/300\n",
      "4/4 [==============================] - 0s 16ms/step - loss: 3.0004 - val_loss: 1.6178\n",
      "Epoch 6/300\n",
      "4/4 [==============================] - 0s 15ms/step - loss: 2.5971 - val_loss: 1.4237\n",
      "Epoch 7/300\n",
      "4/4 [==============================] - 0s 17ms/step - loss: 2.2644 - val_loss: 1.2706\n",
      "Epoch 8/300\n",
      "4/4 [==============================] - 0s 17ms/step - loss: 1.9939 - val_loss: 1.1538\n",
      "Epoch 9/300\n",
      "4/4 [==============================] - 0s 14ms/step - loss: 1.7756 - val_loss: 1.0602\n",
      "Epoch 10/300\n",
      "4/4 [==============================] - 0s 17ms/step - loss: 1.5949 - val_loss: 0.9871\n",
      "Epoch 11/300\n",
      "4/4 [==============================] - 0s 16ms/step - loss: 1.4477 - val_loss: 0.9292\n",
      "Epoch 12/300\n",
      "4/4 [==============================] - 0s 15ms/step - loss: 1.3285 - val_loss: 0.8864\n",
      "Epoch 13/300\n",
      "4/4 [==============================] - 0s 13ms/step - loss: 1.2294 - val_loss: 0.8530\n",
      "Epoch 14/300\n",
      "4/4 [==============================] - 0s 16ms/step - loss: 1.1484 - val_loss: 0.8284\n",
      "Epoch 15/300\n",
      "4/4 [==============================] - 0s 16ms/step - loss: 1.0848 - val_loss: 0.8096\n",
      "Epoch 16/300\n",
      "4/4 [==============================] - 0s 15ms/step - loss: 1.0298 - val_loss: 0.7970\n",
      "Epoch 17/300\n",
      "4/4 [==============================] - 0s 17ms/step - loss: 0.9875 - val_loss: 0.7880\n",
      "Epoch 18/300\n",
      "4/4 [==============================] - 0s 18ms/step - loss: 0.9496 - val_loss: 0.7821\n",
      "Epoch 19/300\n",
      "4/4 [==============================] - 0s 21ms/step - loss: 0.9226 - val_loss: 0.7789\n",
      "Epoch 20/300\n",
      "4/4 [==============================] - 0s 17ms/step - loss: 0.8995 - val_loss: 0.7773\n",
      "Epoch 21/300\n",
      "4/4 [==============================] - 0s 14ms/step - loss: 0.8783 - val_loss: 0.7771\n",
      "Epoch 22/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8631 - val_loss: 0.7778\n",
      "Epoch 23/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8494 - val_loss: 0.7795\n",
      "Epoch 24/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8386 - val_loss: 0.7816\n",
      "Epoch 25/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8299 - val_loss: 0.7839\n",
      "Epoch 26/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.8224 - val_loss: 0.7866\n"
     ]
    }
   ],
   "source": [
    "model = tf.keras.Sequential()\n",
    "model.add(tf.keras.layers.Dense(units=1, input_dim=1))\n",
    "model.compile(optimizer='sgd', loss='mse')\n",
    "\n",
    "callback_list = [tf.keras.callbacks.ModelCheckpoint(filepath='my_model.h5', \n",
    "                                 monitor='val_loss', save_best_only=True),\n",
    "                 tf.keras.callbacks.EarlyStopping(patience=5)]\n",
    "history = model.fit(x_train, y_train, epochs=300, \n",
    "                    validation_split=0.2, callbacks=callback_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "scrolled": false
   },
   "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": [
    "epochs = np.arange(1, len(history.history['loss'])+1)\n",
    "plt.plot(epochs, history.history['loss'], label='Training loss')\n",
    "plt.plot(epochs, history.history['val_loss'], label='Validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "저장된 모델 복원:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 [==============================] - 0s 2ms/step - loss: 1.0001\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "1.0001060962677002"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = tf.keras.models.load_model('simple_model.h5')\n",
    "model.load_weights('my_model.h5')\n",
    "model.evaluate(x_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "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": [
    "x_arr = np.arange(-2, 2, 0.1)\n",
    "y_arr = model.predict(x_arr)\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(x_train, y_train, 'bo')\n",
    "plt.plot(x_test, y_test, 'bo', alpha=0.3)\n",
    "plt.plot(x_arr, y_arr, '-r', lw=3)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## 계산 그래프 시각화하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.keras.backend.clear_session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.Sequential()\n",
    "model.add(tf.keras.layers.Dense(units=1, input_dim=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "1/4 [======>.......................] - ETA: 0s - loss: 4.7190WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/summary_ops_v2.py:1277: stop (from tensorflow.python.eager.profiler) is deprecated and will be removed after 2020-07-01.\n",
      "Instructions for updating:\n",
      "use `tf.profiler.experimental.stop` instead.\n",
      "2/4 [==============>...............] - ETA: 0s - loss: 4.7298WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0023s vs `on_train_batch_end` time: 0.0906s). Check your callbacks.\n",
      "4/4 [==============================] - 0s 58ms/step - loss: 4.6915 - val_loss: 2.8168\n",
      "Epoch 2/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 3.8818 - val_loss: 2.4785\n",
      "Epoch 3/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 3.3211 - val_loss: 2.1914\n",
      "Epoch 4/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 2.8498 - val_loss: 1.9437\n",
      "Epoch 5/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 2.4581 - val_loss: 1.7506\n",
      "Epoch 6/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 2.1344 - val_loss: 1.5639\n",
      "Epoch 7/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.8434 - val_loss: 1.4327\n",
      "Epoch 8/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.6300 - val_loss: 1.3238\n",
      "Epoch 9/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 1.4578 - val_loss: 1.2450\n",
      "Epoch 10/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 1.3307 - val_loss: 1.1512\n",
      "Epoch 11/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.1891 - val_loss: 1.0955\n",
      "Epoch 12/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.0996 - val_loss: 1.0606\n",
      "Epoch 13/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 1.0429 - val_loss: 1.0300\n",
      "Epoch 14/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.9932 - val_loss: 1.0007\n",
      "Epoch 15/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.9477 - val_loss: 0.9791\n",
      "Epoch 16/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.9075 - val_loss: 0.9595\n",
      "Epoch 17/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8727 - val_loss: 0.9415\n",
      "Epoch 18/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8438 - val_loss: 0.9341\n",
      "Epoch 19/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.8348 - val_loss: 0.9293\n",
      "Epoch 20/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8261 - val_loss: 0.9189\n",
      "Epoch 21/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8105 - val_loss: 0.9096\n",
      "Epoch 22/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.8002 - val_loss: 0.9053\n",
      "Epoch 23/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7962 - val_loss: 0.9023\n",
      "Epoch 24/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7898 - val_loss: 0.9011\n",
      "Epoch 25/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7829 - val_loss: 0.9028\n",
      "Epoch 26/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7805 - val_loss: 0.9011\n",
      "Epoch 27/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7745 - val_loss: 0.9001\n",
      "Epoch 28/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7729 - val_loss: 0.8972\n",
      "Epoch 29/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7698 - val_loss: 0.8988\n",
      "Epoch 30/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7682 - val_loss: 0.8994\n",
      "Epoch 31/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7663 - val_loss: 0.9040\n",
      "Epoch 32/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7667 - val_loss: 0.9033\n",
      "Epoch 33/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7647 - val_loss: 0.9028\n",
      "Epoch 34/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7629 - val_loss: 0.8990\n",
      "Epoch 35/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7617 - val_loss: 0.9001\n",
      "Epoch 36/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7610 - val_loss: 0.8969\n",
      "Epoch 37/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7598 - val_loss: 0.8965\n",
      "Epoch 38/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7581 - val_loss: 0.8982\n",
      "Epoch 39/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7582 - val_loss: 0.9006\n",
      "Epoch 40/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7615 - val_loss: 0.8989\n",
      "Epoch 41/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7580 - val_loss: 0.8998\n",
      "Epoch 42/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.8997\n",
      "Epoch 43/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9009\n",
      "Epoch 44/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7576 - val_loss: 0.8990\n",
      "Epoch 45/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7585 - val_loss: 0.8957\n",
      "Epoch 46/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.8939\n",
      "Epoch 47/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8934\n",
      "Epoch 48/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.8946\n",
      "Epoch 49/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.8937\n",
      "Epoch 50/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8942\n",
      "Epoch 51/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.8936\n",
      "Epoch 52/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.8967\n",
      "Epoch 53/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9002\n",
      "Epoch 54/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9009\n",
      "Epoch 55/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9056\n",
      "Epoch 56/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7575 - val_loss: 0.9033\n",
      "Epoch 57/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8996\n",
      "Epoch 58/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.8994\n",
      "Epoch 59/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7569 - val_loss: 0.9028\n",
      "Epoch 60/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9019\n",
      "Epoch 61/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8999\n",
      "Epoch 62/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9004\n",
      "Epoch 63/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9010\n",
      "Epoch 64/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9009\n",
      "Epoch 65/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.8992\n",
      "Epoch 66/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8978\n",
      "Epoch 67/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8975\n",
      "Epoch 68/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.8952\n",
      "Epoch 69/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9017\n",
      "Epoch 70/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9002\n",
      "Epoch 71/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9007\n",
      "Epoch 72/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.8989\n",
      "Epoch 73/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9013\n",
      "Epoch 74/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9043\n",
      "Epoch 75/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9057\n",
      "Epoch 76/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9039\n",
      "Epoch 77/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7559 - val_loss: 0.9074\n",
      "Epoch 78/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9057\n",
      "Epoch 79/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9068\n",
      "Epoch 80/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7579 - val_loss: 0.9080\n",
      "Epoch 81/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9170\n",
      "Epoch 82/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9169\n",
      "Epoch 83/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9168\n",
      "Epoch 84/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7576 - val_loss: 0.9155\n",
      "Epoch 85/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9222\n",
      "Epoch 86/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9190\n",
      "Epoch 87/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9165\n",
      "Epoch 88/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9184\n",
      "Epoch 89/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9193\n",
      "Epoch 90/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9160\n",
      "Epoch 91/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9129\n",
      "Epoch 92/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9131\n",
      "Epoch 93/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9110\n",
      "Epoch 94/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9117\n",
      "Epoch 95/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9166\n",
      "Epoch 96/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7573 - val_loss: 0.9140\n",
      "Epoch 97/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9167\n",
      "Epoch 98/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.9148\n",
      "Epoch 99/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9183\n",
      "Epoch 100/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9173\n",
      "Epoch 101/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9127\n",
      "Epoch 102/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9094\n",
      "Epoch 103/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9059\n",
      "Epoch 104/300\n",
      "4/4 [==============================] - 0s 12ms/step - loss: 0.7558 - val_loss: 0.9074\n",
      "Epoch 105/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9075\n",
      "Epoch 106/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.9076\n",
      "Epoch 107/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9086\n",
      "Epoch 108/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9073\n",
      "Epoch 109/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.9053\n",
      "Epoch 110/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9075\n",
      "Epoch 111/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9043\n",
      "Epoch 112/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9014\n",
      "Epoch 113/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.8999\n",
      "Epoch 114/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.8995\n",
      "Epoch 115/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8957\n",
      "Epoch 116/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.8931\n",
      "Epoch 117/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.8913\n",
      "Epoch 118/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8929\n",
      "Epoch 119/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.8948\n",
      "Epoch 120/300\n",
      "4/4 [==============================] - 0s 7ms/step - loss: 0.7565 - val_loss: 0.8975\n",
      "Epoch 121/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.8964\n",
      "Epoch 122/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.8960\n",
      "Epoch 123/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8939\n",
      "Epoch 124/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8944\n",
      "Epoch 125/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.8953\n",
      "Epoch 126/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7582 - val_loss: 0.8980\n",
      "Epoch 127/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.8938\n",
      "Epoch 128/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7587 - val_loss: 0.8990\n",
      "Epoch 129/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9006\n",
      "Epoch 130/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7578 - val_loss: 0.9009\n",
      "Epoch 131/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.8966\n",
      "Epoch 132/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.8973\n",
      "Epoch 133/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9050\n",
      "Epoch 134/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9054\n",
      "Epoch 135/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9047\n",
      "Epoch 136/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9044\n",
      "Epoch 137/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9037\n",
      "Epoch 138/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9006\n",
      "Epoch 139/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.8981\n",
      "Epoch 140/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7573 - val_loss: 0.8994\n",
      "Epoch 141/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.8979\n",
      "Epoch 142/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7573 - val_loss: 0.9059\n",
      "Epoch 143/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9031\n",
      "Epoch 144/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9024\n",
      "Epoch 145/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9053\n",
      "Epoch 146/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7577 - val_loss: 0.9076\n",
      "Epoch 147/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.9051\n",
      "Epoch 148/300\n",
      "4/4 [==============================] - 0s 11ms/step - loss: 0.7570 - val_loss: 0.9045\n",
      "Epoch 149/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7577 - val_loss: 0.9053\n",
      "Epoch 150/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7557 - val_loss: 0.9065\n",
      "Epoch 151/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9048\n",
      "Epoch 152/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7575 - val_loss: 0.9057\n",
      "Epoch 153/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9048\n",
      "Epoch 154/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9054\n",
      "Epoch 155/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9061\n",
      "Epoch 156/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7561 - val_loss: 0.9014\n",
      "Epoch 157/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9042\n",
      "Epoch 158/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9038\n",
      "Epoch 159/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9036\n",
      "Epoch 160/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7556 - val_loss: 0.9053\n",
      "Epoch 161/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9062\n",
      "Epoch 162/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9041\n",
      "Epoch 163/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7566 - val_loss: 0.9019\n",
      "Epoch 164/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9077\n",
      "Epoch 165/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9072\n",
      "Epoch 166/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9106\n",
      "Epoch 167/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7556 - val_loss: 0.9123\n",
      "Epoch 168/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9118\n",
      "Epoch 169/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.9090\n",
      "Epoch 170/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9113\n",
      "Epoch 171/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.9076\n",
      "Epoch 172/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9062\n",
      "Epoch 173/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7561 - val_loss: 0.9065\n",
      "Epoch 174/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9045\n",
      "Epoch 175/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9102\n",
      "Epoch 176/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7572 - val_loss: 0.9182\n",
      "Epoch 177/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7581 - val_loss: 0.9165\n",
      "Epoch 178/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7583 - val_loss: 0.9172\n",
      "Epoch 179/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7580 - val_loss: 0.9125\n",
      "Epoch 180/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9113\n",
      "Epoch 181/300\n",
      "4/4 [==============================] - 0s 11ms/step - loss: 0.7563 - val_loss: 0.9086\n",
      "Epoch 182/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7557 - val_loss: 0.9070\n",
      "Epoch 183/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7577 - val_loss: 0.9123\n",
      "Epoch 184/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9086\n",
      "Epoch 185/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.9080\n",
      "Epoch 186/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9056\n",
      "Epoch 187/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7568 - val_loss: 0.9050\n",
      "Epoch 188/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7559 - val_loss: 0.9082\n",
      "Epoch 189/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9098\n",
      "Epoch 190/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9125\n",
      "Epoch 191/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7579 - val_loss: 0.9112\n",
      "Epoch 192/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9086\n",
      "Epoch 193/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9087\n",
      "Epoch 194/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9068\n",
      "Epoch 195/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9067\n",
      "Epoch 196/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9070\n",
      "Epoch 197/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9075\n",
      "Epoch 198/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9051\n",
      "Epoch 199/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7580 - val_loss: 0.9016\n",
      "Epoch 200/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9002\n",
      "Epoch 201/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.9005\n",
      "Epoch 202/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9045\n",
      "Epoch 203/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9031\n",
      "Epoch 204/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9003\n",
      "Epoch 205/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8969\n",
      "Epoch 206/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.8972\n",
      "Epoch 207/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8966\n",
      "Epoch 208/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7591 - val_loss: 0.8986\n",
      "Epoch 209/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8978\n",
      "Epoch 210/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.8994\n",
      "Epoch 211/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.8992\n",
      "Epoch 212/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7578 - val_loss: 0.9036\n",
      "Epoch 213/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7569 - val_loss: 0.9055\n",
      "Epoch 214/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9086\n",
      "Epoch 215/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7578 - val_loss: 0.9082\n",
      "Epoch 216/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7560 - val_loss: 0.9065\n",
      "Epoch 217/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7559 - val_loss: 0.9057\n",
      "Epoch 218/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7575 - val_loss: 0.9061\n",
      "Epoch 219/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7558 - val_loss: 0.9076\n",
      "Epoch 220/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9054\n",
      "Epoch 221/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7573 - val_loss: 0.9081\n",
      "Epoch 222/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9045\n",
      "Epoch 223/300\n",
      "4/4 [==============================] - 0s 11ms/step - loss: 0.7562 - val_loss: 0.9018\n",
      "Epoch 224/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9015\n",
      "Epoch 225/300\n",
      "4/4 [==============================] - 0s 12ms/step - loss: 0.7571 - val_loss: 0.9068\n",
      "Epoch 226/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9084\n",
      "Epoch 227/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9098\n",
      "Epoch 228/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9134\n",
      "Epoch 229/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9111\n",
      "Epoch 230/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9155\n",
      "Epoch 231/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.9163\n",
      "Epoch 232/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7580 - val_loss: 0.9175\n",
      "Epoch 233/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9165\n",
      "Epoch 234/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.9125\n",
      "Epoch 235/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9103\n",
      "Epoch 236/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9094\n",
      "Epoch 237/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.9102\n",
      "Epoch 238/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9118\n",
      "Epoch 239/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7585 - val_loss: 0.9099\n",
      "Epoch 240/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.9036\n",
      "Epoch 241/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9011\n",
      "Epoch 242/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8999\n",
      "Epoch 243/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9050\n",
      "Epoch 244/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7558 - val_loss: 0.9030\n",
      "Epoch 245/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9036\n",
      "Epoch 246/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9028\n",
      "Epoch 247/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7564 - val_loss: 0.8995\n",
      "Epoch 248/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9000\n",
      "Epoch 249/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7571 - val_loss: 0.8980\n",
      "Epoch 250/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7579 - val_loss: 0.8923\n",
      "Epoch 251/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.8921\n",
      "Epoch 252/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.8961\n",
      "Epoch 253/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7583 - val_loss: 0.8949\n",
      "Epoch 254/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7574 - val_loss: 0.8953\n",
      "Epoch 255/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.8949\n",
      "Epoch 256/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.8932\n",
      "Epoch 257/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.8933\n",
      "Epoch 258/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8932\n",
      "Epoch 259/300\n",
      "4/4 [==============================] - 0s 11ms/step - loss: 0.7575 - val_loss: 0.8941\n",
      "Epoch 260/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.8954\n",
      "Epoch 261/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8996\n",
      "Epoch 262/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7567 - val_loss: 0.9004\n",
      "Epoch 263/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9003\n",
      "Epoch 264/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7563 - val_loss: 0.9030\n",
      "Epoch 265/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7567 - val_loss: 0.9069\n",
      "Epoch 266/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7565 - val_loss: 0.9039\n",
      "Epoch 267/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9019\n",
      "Epoch 268/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.8995\n",
      "Epoch 269/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7571 - val_loss: 0.8991\n",
      "Epoch 270/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8961\n",
      "Epoch 271/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7568 - val_loss: 0.8936\n",
      "Epoch 272/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7578 - val_loss: 0.8958\n",
      "Epoch 273/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7563 - val_loss: 0.8964\n",
      "Epoch 274/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7565 - val_loss: 0.8967\n",
      "Epoch 275/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7569 - val_loss: 0.8969\n",
      "Epoch 276/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.8966\n",
      "Epoch 277/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7562 - val_loss: 0.9002\n",
      "Epoch 278/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7561 - val_loss: 0.9024\n",
      "Epoch 279/300\n",
      "4/4 [==============================] - 0s 12ms/step - loss: 0.7561 - val_loss: 0.9075\n",
      "Epoch 280/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9071\n",
      "Epoch 281/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9052\n",
      "Epoch 282/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9040\n",
      "Epoch 283/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7574 - val_loss: 0.9025\n",
      "Epoch 284/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7570 - val_loss: 0.9073\n",
      "Epoch 285/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7566 - val_loss: 0.9099\n",
      "Epoch 286/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7572 - val_loss: 0.9068\n",
      "Epoch 287/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7580 - val_loss: 0.9077\n",
      "Epoch 288/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7561 - val_loss: 0.9030\n",
      "Epoch 289/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7561 - val_loss: 0.9042\n",
      "Epoch 290/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7564 - val_loss: 0.9041\n",
      "Epoch 291/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9015\n",
      "Epoch 292/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7556 - val_loss: 0.9005\n",
      "Epoch 293/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.8996\n",
      "Epoch 294/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7559 - val_loss: 0.9021\n",
      "Epoch 295/300\n",
      "4/4 [==============================] - 0s 10ms/step - loss: 0.7559 - val_loss: 0.9021\n",
      "Epoch 296/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7560 - val_loss: 0.9045\n",
      "Epoch 297/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7575 - val_loss: 0.9125\n",
      "Epoch 298/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7570 - val_loss: 0.9143\n",
      "Epoch 299/300\n",
      "4/4 [==============================] - 0s 8ms/step - loss: 0.7568 - val_loss: 0.9144\n",
      "Epoch 300/300\n",
      "4/4 [==============================] - 0s 9ms/step - loss: 0.7568 - val_loss: 0.9079\n"
     ]
    }
   ],
   "source": [
    "callback_list = [tf.keras.callbacks.TensorBoard(log_dir='logs')]\n",
    "model.compile(optimizer='sgd', loss='mse')\n",
    "history = model.fit(x_train, y_train, epochs=300, \n",
    "                    callbacks=callback_list, validation_split=0.3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "코랩을 사용하거나 주피터 노트북을 로컬 컴퓨터에서 실행하는 경우엔 아래 두 개의 코드 셀의 주석을 삭제하고 실행하면 텐서보드를 노트북에 임베딩시킬 수 있습니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load_ext tensorboard.notebook"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %tensorboard --logdir logs --port 6006"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"tensorboard-1.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"tensorboard-2.png\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.keras.Input(shape=(784,))\n",
    "hidden = tf.keras.layers.Dense(100)(input)\n",
    "output = tf.keras.layers.Dense(10)(hidden)\n",
    "\n",
    "model = tf.keras.Model(input, output)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.keras.utils.plot_model(model, to_file='model_1.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.keras.utils.plot_model(model, show_shapes=True, to_file='model_2.png')"
   ]
  }
 ],
 "metadata": {
  "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}