7-1_Artificial_Neural_Network.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FuoCYvFP3rz2"
      },
      "source": [
        "# 인공 신경망"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Abp3s6mi3rz9"
      },
      "source": [
        "<table align=\"left\">\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/rickiepark/hg-mldl/blob/master/7-1.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />구글 코랩에서 실행하기</a>\n",
        "  </td>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u5msa4prlifV"
      },
      "source": [
        "## 패션 MNIST"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "fMTyGYIMqUE9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "606f9e81-2f74-43df-df8e-60ca9d7afb9f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 0s 0us/step\n",
            "40960/29515 [=========================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
            "26427392/26421880 [==============================] - 0s 0us/step\n",
            "26435584/26421880 [==============================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "16384/5148 [===============================================================================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
            "4423680/4422102 [==============================] - 0s 0us/step\n",
            "4431872/4422102 [==============================] - 0s 0us/step\n"
          ]
        }
      ],
      "source": [
        "from tensorflow import keras\n",
        "(train_input, train_target), (test_input, test_target) = keras.datasets.fashion_mnist.load_data()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2-r-EgKgqYkK",
        "outputId": "bbe8d753-b4a4-46af-ad4e-378c9657288d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(60000, 28, 28) (60000,)\n"
          ]
        }
      ],
      "source": [
        "print(train_input.shape, train_target.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LAJGYTkFrS6p",
        "outputId": "0ccb800e-3f66-462e-8ee4-9f4288010dc5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(10000, 28, 28) (10000,)\n"
          ]
        }
      ],
      "source": [
        "print(test_input.shape, test_target.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 78
        },
        "id": "lAYiLboglZIz",
        "outputId": "6acc93ce-25e0-47c1-85a6-98666def9886"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjwAAAA9CAYAAACpzLMWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO1daYydV3l+7r7v987uWWyPlxnbieNg7JAQsjaBQqioEsoidUEVSNBKFaC2P4qoSlVVlApUVAlES9OKpiBIIQlhacHKYhs78TLeZsYZzz5z931f+2P0HJ/7+c7q8dxJeh/Jsn3vd+/9znfOec+7PO/7qmq1GlpooYUWWmihhRbeyVA3+wZaaKGFFlpooYUW7jRaCk8LLbTQQgsttPCOR0vhaaGFFlpooYUW3vFoKTwttNBCCy200MI7Hi2Fp4UWWmihhRZaeMejpfC00EILLbTQQgvveGhXef/tnrOuWsM1ax5jrVZDOp1GJpNBNBpFIpFAKpVCMBjE4uIinn/+ebS1taGvrw8ejwcmkwlarRYqlQpqtRqxWAyJRAKTk5Mol8vweDzo6+vD4OAg7HY7jEYj3G43bDYbOjs7YTQaodfrt3SM8lir1SoWFxcRDAbxk5/8BMViEUeOHIHVaoXNZoNKpRJjq9VqqFQqqFarqFarmJqaQi6XQ7VaxcDAAIaGhuDz+WAymdZ7K8DqY1zT+OLxOLLZLGZnZ5HL5ZBMJrGwsIBQKIRr164hn8/DYDCgra0NPT098Hg80Ov1SKVSCIfDGB8fRzAYRDweh8lkgsPhwKFDh9Db24udO3fC4XDAYrFgcHAQWu1qW2td41vzGGVEo1FcvXpVzIlWq4XFYsGePXug1+uh0+nqrq9WqygWi7hx4wbm5+cRj8ehVqvh8/nQ1dWFgYEBqNVqqFRrud1bsGljLJfLKBaLmJ+fRyaTQTKZRDabRS6Xw+TkJILBIM6cOYNKpYJarQan0wmj0QitVotKpYJisYhoNIpUKoXBwUF0d3fj0Ucfhclkgk6ng8PhgMlkQk9PD3Q6HTQazZaPcRvjtvdirVarW0Ncd5Spc3NziMfjuH79OkwmE+x2O2w22y2yMJVKIZPJYH5+Hnq9HgcPHoTX60VXVxc6OjpgNpvvxPiAd/4cAu/QMapWqcPzjhy0AmseY6lUwtjYGKampvD666/jwoULmJiYgFqtRrVaRTabRblcRqlUQjqdRrFYvPWGVCr09vbCYDAIJSgYDMLhcMDpdOLYsWMYHh7GU089ha6uLng8ni0dI1GpVJDP5/Hmm2/i6tWr+P73v49wOAytVguz2Qyr1QqtVgu1Wi0Oknw+j1KphHK5jFAoBJPJhAceeAA9PT0YGBjAe97zHnR2dq73VoBNUnjOnTuHGzdu4Hvf+x5yuRz0ej0qlQoqlQr0ej1qtZpQZhcXFxGPx1GpVNDZ2QmNRgOtVgubzQar1Qqr1Qpg6fBNpVJIpVI4dOgQdu/ejc9//vOw2+2bOb41j1HGa6+9hi996UvIZrMoFApwOp3YtWsXvvjFL8Lr9cLlctVdn8/nEYvF8I1vfAM/+tGPUCgUoNFo0N7ejt/93d/Fpz/9aRgMhvUoADI2bYypVArRaBT/+Z//iampKVy8eBHlchmVSgUGgwEqlQrlchmRSAShUAi5XA7lcnnpJlQqaLVaOJ1OWK1WeL1eaDQaFAoFGAwGWK1WDA8Po7e3F8888wycTicsFsuWj3EbY9MVnnw+j0gkglOnTmFkZATPPfccFhcXkU6n13RDarUaarUaGo0GDzzwAD74wQ/igx/8IAYGBm75XQCrKeytOVzCO3KM6zJD/z+iUqmgVCrh17/+NSYnJzE+Po54PC4sYI1GA51OB5VKBYPBIDwjDocDlUoFhUIBKpVKXKNSqWCz2aDRaKBSqeByuWA0GsX3zM7OIp1Ow+/3o6+vD93d3XjiiSfgdru3dNxqtRp6vR5GoxEWiwXve9/7EAwG8eabbyKbzSKdTiOfzwuvDgUOD5TOzk60t7djaGgIFovldjwDt418Po90Oo0rV65gYmICLpcLdrtdeD7o/VCpVLDb7TCZTDAYDCgUCqjVarBarUKoms1m8b5avRQR9nq9AACTyYRkMomXX34Z/f39ePe7392U8cqo1WrI5/PIZDLI5XJIp9P4+7//e1gsFlitVgwMDMBkMuHq1avIZrPI5/O4fPkystkstFqtUAaq1Sr0er0YczPAeXzhhRcwPj6O+fl5pNPpOq+hvM+cTif6+vqQy+VQqVTEGtTpdNBqtdBqtWK/cmyVSgUzMzOIx+Mol8vo6urCwYMH0d3dLea5hY1BqXCk02k8++yzCAQCCAQCWFhYQDAYRCKRgFqtFnu0XC4LD53BYEClUkE2m60zVoAlT9HExASef/55XLt2TXjbBwYG8NBDD4nfVSpc2wVrVMhWRCQSQSqVQldXlzCoTSaTMNDezpidncWvf/1r7NmzBzt27EBbWxt0Oh0qlYqINKyGbaPw1Go11Gq1pgrURigWi0in0zh16hTOnj2LGzduCM8NwwIUnjzwKTgrlQrS6TQ0Go0QqBTGhF6vh91uR7FYRKVSQSgUQiAQwOXLl9Hf34++vj4cP34cLpdrSzcpDwadTgej0YjDhw8jEAjgypUryOVyIjyXz+dRKBSg1+thsVhgNBphNBpx4MAB9PX1oa+vD7VaraG3a6uQy+UQDocxNTWFqakpuN1u1Go1odBQAFJZ0+v1de5zzqlKpYLRaBReDpVKBY1GIzwD0WgUuVwOp0+fRiaTwb333ttURQ9Y2lelUkl4H4PBIK5duwatVguj0YgHH3wQLpcLL7zwAjKZDCqVCrRarZh7tVqNcrmMWq223jDdpiOXyyEUCuHEiRM4d+4c3G63OAQ1Go1QSmVl3WQyoVAooFwuiznT6/UolUqoVquIx+MolUpCcS+VSohEIgiHw1hYWEB/fz/0ej3MZnNL4dlElEolxGIx/PCHP8TMzAwWFxdRLpdRrVZhMBjE+qTBabVaodPpYLPZhCyhzKRXr1AoIBgMYnZ2FhcuXIDVasW9996L48eP47777hMyGth+Sg/lELGee6tWq+Lz0WgUgUAAdrsdGo0GgUAAbrcbFotlW413reC4isUiZmdn8fLLLyOXy0Gj0cDlconowlq9zk1XeOgFmZubQzQaxcGDB29xIVerVQBoijL0+uuv47/+678wNjaGeDwuLGOGsYCbViUAwWXhhmJoQ15sjcKIPGB0Op14PxKJIB6PY2JiAgaDAZ2dnVu+aGlhaTQa+Hw+/M7v/A6uXLmCM2fOQK/Xo1gsolgsQqfTwWKxoLOzE21tbXj88cfhdDoRj8eh1+thMpmapsyOjY3hZz/7GZLJJBwOB4CbHKVSqSTmioqrxWKB2WwWgrdUKgFYmkNZMFHh0Wg0qFQqcDqdqFQqWFhYwOTkJE6ePIndu3dvNIx32yAfJZFICMWsWq0inU6jUCggGo3iySefxOHDhzExMYHZ2VksLCwIpZVjJL+s2RgZGcHzzz+PaDQKt9sNh8Mh5kTmk2k0mjolW61WQ6fTCSW2UqkIHhAVewpMPiMeIouLi/jRj34kOD1Go/FteXBsB8jP7Zvf/CZOnjyJkZERlMtlWK1WMS9UrKkAqdVqpNNpVKtVRKPROmteo9GI/WkwGIRySs/miRMnMDk5icXFRXzkIx/B/fffv+3nbz33VyqVcO3aNcRiMSwuLiIQCCASieB73/ueoBg89thj+L3f+z3hxd7ukD02fr8fCwsL+PrXv47JyUlcvnxZGN5f+MIX0NPTs64xbbnCI2vWxWIR2WwWoVAICwsLiMVi6O3tFdfIQmqrUa1WUSgU4Pf7MTo6imQyiVKpBKPRKKx2enEKhYL4DMdIyBaF/N1KIc3FSCsUWLJo8/k8/H4/vF4vOjo6tnzB8h6r1So0Gg26urqQTCYRCoXEwZnP56HT6WC320Uoy+PxwGg0Ip1Oi7E1a7NRCQUgxsH1VavVhJDlvPJ+5bAV54mHIUGPgkzeJpotXDQaDcxmsxgnQzharRaFQgGlUglutxvd3d1wuVwIh8PigNHpdEIZ3A5eqkqlgng8junpaZRKJej1ejGP8tzK+5L7jJY9eT7kmzFUp5wzzin5eMFgEOFwGLFYDD6f7xaydwtrR6FQQDabxejoKC5duoR8Pi/kH0ElVjZ0qfzI65FzVS6XxfqU5VW5XEYul0MgEMCFCxdw9OhRJJNJYbBuJ8gGswyGmU0mE2q1GnK5HLLZLFKpFLRaLYrFIkZGRpBMJhEOh0XoOpvNolarCSL+doPScFSOv1QqoVQq4fr163jrrbdw6dIlBAIB5PN5APVOhvWcLVuq8Mjhg0qlgsXFRdy4cQO/+tWvRNiHrHy9Xo/29nbs2LGj4eJUKheyS2szPEKFQgHj4+OYmppCIBCAxWIRvBwK2FKpJLILePBzjHSx81qlB4j3rNfrYTAYhIJAYSs/s1dffRWxWAwHDhzY8o3KkFY2mxVeHp/Ph/vvvx+RSASZTAbxeBxGoxFtbW2wWq1C0WGcvZEiuJXo7+/Hk08+iRdffBHBYBBOpxN6vR4ulwuFQgHFYrFOKaMATqfTIpyj0+nqsu5k8PpsNotisQi3242dO3fi+PHjTRWsRqMRnZ2dIlzAsdGjwflQq9Voa2tDIBBAOp2u82AWi0Xk83lB+m0GKpUKIpEIIpEIotGo4FDREtRoNCgWi3WHItcc9yGvY5YWBarsFZLDlgxt8nuvX78Om82GRx99FE6ns2nP4u2O+fl5XLp0SSR8eDweIbvp4QHqDzHOCXBThsiyhIYGDU/OtU6ng8/nQ6lUwtmzZ3HXXXehra0NR48ehc1m27IxrxWNDu6rV6/irbfewv79+1Gr1XDp0iWcP38er732GtxuN9RqNa5fvw6dToeOjg709PSgvb0dH/vYx9DV1YXdu3cLz3UzISs4/DeNR6U8Yijuxo0b+OpXv4qzZ88CWDqLent78dRTT+HDH/4w2trabnESrIYtfwqyFTY/P4+pqSlMTEzAZDLBYrFgampKhLT8fj/8fj96e3sFmZRQDpDCC8CmWKSFQgGzs7OIRqPC0lUe3uQD3H333SL8Qf5HsVgU8WguZN4TLU1u0nw+j0AggFwuJ8YoW62Tk5OwWCx1noWtguy5oNeLY+GYyaHQ6/WoVqvigGyWgqOE1WpFd3c3urq6kM1mEY1GBRmbm69UKgnuB8NUBoNB8FjkUAkVIM4fNxxDf3fffTf6+/s3ms20adDr9fB6vYJMT1AAqtVqkfFEy1kWjPwMn0mzwDVFy5brkX+4l+QQJaGUCfTokcMjK7n8UyqVoNFoYDKZUC6XRcbX9PS0OFRbWBuUXJmJiQn8+Mc/RiQSgVarraMFyHMkf04mnPM7lZBDm41+X6vVYmRkBJVKBTt27BCGXKN7bCZqtRoSiYRQ3BkepzI/NTWFSqWCXbt2YceOHbDZbHjXu95VZ2jTcaDT6ZDP54UC30zIz1cZhpYV20qlglwuhytXruDll1/G7OysMGzcbjeOHj2KwcFBuFwuIZO2bUhL1vLK5TJu3LiB69evY2xsDE6nE3a7XSgNdONZLBY8/vjjGBgYuCXEwL8bHcS3a1kXCgW89dZb8Pv9wvKQD3+Ow2g04n3vex86Ozvh8Xhgs9ngdDqFouT1esUE04okSTKXywmS3QsvvCDquyjd7OPj44Jb0AxQ2DD1njCZTDCbzYK7UigURAqwRqOpO1CaGdKy2Wyw2Wzo7+9HNpvF+Pg4LBYLvF6vECQ85Jgxx3AQAKHIySEQnU6HarWKTCYjNm+pVILJZMKDDz645Vl1jWA0GtHR0SHc4TLJOpfLQafTiRR8KtgysZPhv0Y1e7YS1WoVuVwOmUwG6XRalBAgaVXe+1S0uf9lxYj7kGMF6j2vfD4A6vZ5tVpFIBAQxNgW1gal4gIAV65cwb/+678KA5b8P3q3GXqUvwO4yesAbvXc0xvJ1+W551ybTCacPn0ap06dwlNPPSU86pzfZodtCSau8IyYnp7G9evXMT8/j2q1ilQqBZ/Ph6NHj2J4eBidnZ3Yt28f0uk0JiYmMD09jUgkIs6LcDgs6lA1G0qlB8AthhRD17/5zW/wrW99CyaTSZwlHR0d+PCHP4zh4eH1lv0Q2FKFR469lstlTE5OYnJyUkxwLpcTNTJ27dqFbDaLYDCIl19+GVarFbt27YLdbkdbW5uojzE7Oyti+y6XC+3t7ejt7YXNZtto4SkASymTp0+fxtzcnNhQskuOdTt0Oh2OHz+Ojo4OTE9PI5VKIRKJiHTsN998U4SCKIxpsZhMJuzduxeHDx/Gm2++idnZWaHU0GOl1+sRCARE1ojX693wZG8UcqxVPgjIg5Bj7o3CV7IV3ky8613vQkdHB1588UVks1nh/SBZnN4qhhmV98usEb5erVaFF6BQKGBgYABtbW2w2+0wGAzNGGId9Hq9KIApF4csl8vQ6/VwOp34t3/7N/zgBz9AKBRCoVAQxRZlnpnFYmkalw5YMj4oyJn5yJII9LpSUeOz93q9IsGAHi1eQ28CcNN7RG/yzMyM8H4xsy2fz8Nut4usvhbWBtnIWVxcxLPPPotTp07BYrGINRmNRlGpVEQWEb2MVEKUxt9yaKT0yJ4E7ulKpYKJiQnY7Xa4XK46Dl6zwD05NzcnsnqZpTY8PAyXy4WTJ08inU7DYrFgaGgIDz/8sAhP8yzp7u5GNBpFOBzG2NiYkGuDg4NbnmG4nNcsk8kI7i7P/FAohHg8LmrX5fN5XL9+HVarFQ6HAzqdTsjr3bt3i8STjWDL/Vy0hDnQWCwmrE2NRiOquzLMkE6nkU6nhYXpcrlEhoXVahXZXdPT00gkEigWiyIl73YUnmKxKL5TtgKVf1QqFTo7O+H1enHx4kUkk0mkUins2bMHWq0Ws7Ozgnydy+VQKBRQqVSg0+ng8Xjgcrmwc+dOWK1W4UXhwSRb2iwKZzabt1ThoSIjL2DZUl5OGCljtmsVXHcS7e3tUKvVsFqtKJVKKBQK0Ol0wiNIz45cYgCA8O7xdYZF6LVj2MPj8aCnp2dTPIybAY1GA4vFckuxQB4MRqMR4+PjKBaLIjOL/B3ZmqYC0EyQqCx75QCI9HN6oTweDywWCzo6OoS3lSE5yhiWiCDhNZfLweFwwGAwIJ/PI5fLiexDFtIsFArCw9fC+pFOp3Hy5ElMT0+LQ5oZVrKSQq+w7Glb6zOXvf9Kz51sbIdCIYTD4TpDrpmgPEkkEkgkEoIrWi6XhTJDY8zr9aK7uxs7d+5EKpUS17AgLNc2i6bWajW43e66sgxbMR5lSJLef5YyodGSTCaxuLiIcDiMUCgkzj3WD5IrpLO6/e0Yk00J7I2NjWFychJXrlxBOBwWikk2m8X8/Dz8fj/Gx8fh9XrR2dkJn88Hq9WKYrGIUCiEaDQq2Otkovf392NmZgZnzpxBNptFX18fHnjggQ3fYz6fF1WUbTZb3QHPjVIoFETtktnZWfzZn/2ZCI1cunQJXq8X//Ef/wG73Y53vetdGB0dxfT0NMrlMsxmM3bv3o2nn34an/nMZ4QAlol33KgU4leuXIFKpUJHR8ftT8IaQeW0WCwKjovSDSzXgSAYj6Xi2szDQr5fl8uFr33taxgdHcVPfvITcQ2VAlqXrC3E8QEQm1Hm8si/MTw8jOHh4aZ6Q2TIng45hs8xqFQqdHV1AUCdQkpuFsnczYbdbsejjz6KI0eO4Omnn8apU6cwOzuLqakp9Pb24rd/+7eF11dJUGYdF5KVKXzlonUmkwk+n0+ENr/85S/j9ddfRywWA7DEL4nH48JYa2H9yOVyuHTpEjKZDADUpf4r65M1Cn3wM6sp3o0UHeBm1g9bpzidzqYYYLIywHtlnSy+Nj09jXA4jNHRUZw5cwZvvfUW/vzP/xwHDhzAkSNHRHiKxQRVKhWKxaJYrzabDYVCAfF4HK+88grS6TS6urrQ3d29JQUIZXlbLBbh9/tx8uRJ/PSnP0U6nUalUoHRaEQqlcL8/DzMZjOMRiOmpqag1Wpx4MAB5PN5+Hw+4cGlkrxapIDPd7l10hRpxnz6WCwm0uc4MGWmDMmhTMPT6XQwm81CYJGYNTc3J/5oNJrbyqTIZDLIZDJ19wKg7uDj4cfNajKZcOjQIdRqNej1emFhHjt2TPQvotuRn7XZbMJbQ0uVaXfATaWHfBESuLcSsndG/sP3+LfyNfnzjV7favBZajQadHZ2IhqNCjc6yw0oGf+yF0AO5ykzlkhwNZlM20bZASCsQ3Jg6OGQPTjy+gYghIWsFDXbAi4UCpicnEQ2m0U2m4XNZhOVZDs7O2G32xEMBjE1NQXgZpYOwfUp87XkGi/kbun1emi1WnR0dODo0aNC5tDSZIjs7YDbIeLeKc+H0tPL75cNIXqPgZs8Kv4t783liMvKe+b6ls+W+fl5tLe3bxtvneyBIl8sl8vB5XJhaGgInZ2dorowa0/xcwSjACTj22w2VCoV0b+Q2ailUumO8/HkeVCr1TAYDHC73RgYGBDnPM+6np4eUfOsr68PmUxGjD+TyQiKAekq5Hopsdbz5Y4qPMqb4GJ+5ZVXcOLECfj9/ls4HxRAWq0W4XAY4XC4rkqxw+HArl274PP54HQ6odFokEwm8fOf/xyxWAzRaBRerxcHDx7c8D0vLi4iFArBarXeQoCkslMsFkWNE5VqqT/WP/zDP4j7j8ViqFar+NSnPiWsZB42tVpNtDrYsWMHgCVt3el0IhAIiGtlhadSqWB0dBR9fX0bGtdGUa1Wb/FsUXAphY88l3L9IdmaawaU1qPD4YDX60V7e7toIkmyslyPhx4eKqhMt8/n82J+yOPZbsoOAFEbiRVoSQCklw64qbwTTLvn/MoZXc1CJBLBv//7vwvu2/DwMHbt2oXBwUER33/uuefwi1/8AsDSHFutVpH0QFe/sqQA92Emk8Hk5CRsNht8Ph/++I//GM8884xQdGw2Wx2Zu5lQKguNPAarEXFXU4bkshmbpfRwPzGM1Ug5oexU3uNyHh+lPJHnh2OQPdLc3+fPn6/z3G4lGo2be5LG17Vr1+D1evHggw9i165d6O7uXvV7S6WSaKZaKBSwY8cOdHZ2ivIu5XJZOBfW0J/xtiAbHFqtFu3t7Xjve9+LI0eONGwCK+PChQv44Ac/KEqC9PX1weVy4fjx4zh48KCokq88f2T9YSVZdUcVnuU2C1n59FwwRseDkSEFKggMpzCNdmJiQlSEBZbCTzMzM7BYLOjv799oR24ASxvl9OnTuHTpEorF4i2pbzJ3gweB3++HVqsVZerlxcXxyRwQFj5LpVKYnZ1FJpNBMBgUBfzIUeJvkGcwOzsr3JZbBS4m3rvS2yM/G4KbmmNuNv+jEXjvDLdRwMq1XeRaLjz4ZctFTqnlobidoNVqGxKo5ZIBykND5o7JGS7NVOaMRiP27dsn4vxsETIzMwOHw4GhoSH09vbi6aefFgUVGToBUNcDTKkMsB7L0NAQzGYzbDYb5ubmEAwGhXVptVqxuLiIZDKJZ555Bh0dHU3LelmLAiN7DIh0Oo2RkRHcuHEDY2Nj+PjHP459+/Y1/I2NpPsuh1qthnA4DL/fj1QqJTwMsmyUvahyqjpDx3KGllyvRckt5Gvyc2BIPZfLiXlvNklZCRpTLABKA4ohqLXwb8gBAiAUdZ1Oh6GhIaFkJpNJ5PP5O67wNAIr8ct0gEbKn9FoxO7duzEzM4P5+XlBvJ6amoLP5wNQn6FN8N+kTizH322KhGatGio0BoNBuOLknjbAzbRzbpByuYx0Ol032FKphGg0ir6+PrS1td2WMKpWqxgbG8PY2JgghMngxpTrRvj9fhF2y+Vydd2ZE4lEXXYIFx/HkUgksLCwgEgkIpQ8jl+lUgnXe7FYFIX+thpUeJT1Tyh4ZCEkfwa4qXE3O6S1HOitk//PdUnhyfmThaW8WeXNvJ1A4j4VMaWXTsm5Ur7O8Te7jgeLqhUKBdFkMp1O48yZM2hra4PBYMDOnTsxNDSETCaDbDaLhYUFFIvFujRyGlOsrM1DxGQyoa+vDwaDAQaDAWfOnIHf7xeZPTabDSMjI5ifn8dDDz0kUpq3K5Rh6EqlglgshnPnzuHMmTP49a9/jePHj2Pnzp0NFVkqCpR/t2uw0POey+UALMl/eha5n0hKV4ageD/KMCywcmFZOVRbqy1llMpzvp0UHj5jZlLJtAWeFaspaeVyWZTIkFtstLW1IZPJIJlMCt5rM6DkPirJ6QRDW8lkUnB6NBoN/H6/IJo3mj+up0wmg1KptL0UHmZAyIoMQ0OMpZvNZmQyGaRSKQC3lt7m57mwNRoN8vm88JRsFGq1GoODg6hUKohGo0gmk0gkEqIBIesD+Xw+YZn8zd/8jbCaldqrfHjIG5YChVlC9CDwdyic1Wo1nE4nnE4njhw5gqGhodt59OuGUsHhISinjDZyDctZNGTlN5vwKXukDAYDfD6fKLon19eRrU6ZKMneS0xll+PRXq+3rmsz0Jzeb43A2lbLCUxlyE95uDAk1CyEQiF8+9vfRiaTEX2XNJqlNie1Wg2vv/46XnrpJUGIVHpx5MNPGfqRvV0GgwEmkwlerxdmsxkulwt2ux2Dg4Noa2tDsVhET0/PLb3+molGh0YoFMK5c+dEX6WzZ88iHo8jm80il8vBZrPhBz/4AS5evIjPfvazt1QdZi2Yl19+Gfv378exY8c2fH+cC5mjCdwkEcvzIcuTleSOfC2fgazcAPWhWdlDrVKpEIvFttxTvhZwrErDku+tRNoulUqIx+Oizpb83Ojhoget2VDehxxyZIJCsVjEuXPnRLZaIBBAMBgUmbVc88ViEQsLC/jNb36D//3f/4XVaoVer8ff/d3fNfztLS88yAlQZiMRtLLYTC6VStVZ3BRosteA5EyVSiXS37LZ7IbS0lWqpcyVcrmMVCqFeDyOSCSCVCol6nWQLc779vv9woKQv4djljeskhMkpxtysdrtdjidTnH4Mr12aGgI7e3tG3jy68dy3JvV4v/K66jVMzNmrZ+/02CTULrX5XmSFVaGuYB6RYZ9qXizTPoAACAASURBVDifyrDJdhijjLXeT6OMGbluTTNA0nUulxOVlBlGpIeOFi7vlQdcI68b545Clh5UZnQBS+OORCIijM66IQyZbYc6S8DNGi6sYZLNZjE3N4dz586J5sNjY2PIZrOi0J5Wq8XU1JQop6BEPB5HMBjEjRs3RBjhdsC5kw9synJZwVYajLxOieW4N8vVAFMaoZzvTCYj9u52xVq9UaSB0Eiz2+2o1Wq3dElvtlxaLiRL0Fvb1tYm9ikroxcKBSQSCRGum5ubQyqVwsLCAi5duoSRkREMDw+vuGbvOGlZVkpYzyIajSIWiwmBRKHDDcHMCBYjokVH9x7TSdnjSa1WY+fOnUJgzc7OYnR0FPfcc8+671mj0eChhx5CrVbDJz7xCVHSfnFxEcFgEM8//zxmZmZw7do18eB7enoAQJCc5caFyz0XACLLTKPRIJfLiZLhR48exec+97mGcfmtOnh4gMjVhSk0lNaHDFlJ4GEJQBQppOu62WD3+bfeekscGhyr3W4Xa65arYqwpExcJddA5hs0W5isBOXB3yiEpfR+cM3R69osWK1W3H///ZicnMT169dFZehjx47B6/Wit7dXGCRtbW2CG8h9SGVGPii5l1gzSavVCqEaiUQQDofx/e9/H2azGdPT07hx4wYikYgoFLp79+5tMd/pdBqhUAinTp3C9PQ0rly5grm5OVy4cAFut1usZRZiZSuR0dFRBAKBunAusCSTzpw5g4mJCVy5cgX9/f23dX/VahXT09Oi6SvlgeyJoZebXh8auEA9V05+jfcq/7+R0cIQplxzi4bXtWvX0NfXJxJHmg2loSF7rlYjWDNL68KFC5iensZ9992H/v5+dHR0iNILzQ5NLwd53FarFceOHcOFCxeE04Mtl/L5PM6dOyecAX/1V3+FyclJAEslbdLpND772c/ive9977K/dcdJy/JExWIxzM7OCi4KrSTZW0OlhRaLvLhlMikVi0KhALPZjJ07d4rCRiQGbxTy4aBWq8UmtVgsuOeee6DX63Ht2rW6z1ATBbCssiO/JisR8vg7Ojrg8/mazhHgs14pPXk5ga/kidCK2k4EZt6TLFyBekWAhyavNxgMwlMlW8eykN2uaCQwlcoOXyOo3LIOU7NQqVREUVHy/wCIthmxWEx4YZPJ5C3h1kZZhlzPDH1wjWs0GlHd1e12iyytjo4OOJ1OeDyeO5qaLu+Z5cCMnEuXLmF2dhaTk5OYmZlBIpFALBZDqVQS5FeZfM9aZhqNBoODg+ju7hah2WKxiGAwiFAohF/+8pfIZDI4fvw42tvb6xrubgRKT3EjUrUcfuJ7ShlOft1qxmS5XBaKLH9bzjyrVqtIp9M4f/481Gr1tlF4VsJy+5OgYeJwOOB0OhGLxWCz2cT8yp6R7Q6OhV6qfD6PfD6PyclJPP/882Jdz8/PCy+dwWAQ+3OlkjR3XN2TJygYDOLChQvw+/111ZNZbVOlWiqgVKvdLFYnb1hZ4yVJiaGroaEhxGIxLC4uitL4mwFa9oztM+b/4x//WNyTbEkq3afyM5A3PMfI3yCfZ8eOHVsWtloJPAyUVYdXwnJKHuvUUOnZDpAVnkYuXyrUJKmTAEyLkdYoUJ+6v12hvDclNwCo52XJ3kQaH81CqVRCIBAQISu2i+jt7UU+n8e1a9dEtVbZA0koya7ye7JCyyrmDz74IDo7O9HT0wOtVgu32y1KYPT09MDlcm2qd0cpF2jcLef+z+VyWFhYwI9+9CNcvnwZFy5cEERVtgbp6+ura2VTq9Vgs9mQz+ehVqtx+PBh7N+/H3q9XmTwXLt2DVeuXMHzzz8Pp9OJP/3TPwUAzM3Nie/fKGQlhXtOWSaA4FlAg0I2YFcyZKk0yaEynhPKSEM8Hserr74Kh8OB++67b8Pj2io0CvHJz06r1cJqtYpifeFwWMjcUqmEcDgMn8+3LbzrxHLKGwBRe0fOaA4EAnj99dfF9S6XS8hwj8eDvr4++Hy+FTsR3LHTR7ZUOBi73S76XHERy8oMANGckvwV1lHgwqVXh8RRl8sFrVaL1157Df39/bj33nsxPDy8aVo7LQQ5VZMHuLKglXxg8BmsVVHg97DqpPIZbjWUno6NglVutwMahQjNZjMcDofwHLAVAwAxD1T6uO4MBgNSqRRqtdrbxmpqBO4roDHnjOs5FAqJ5IFmoFwuIxgMCh5NJBIRnhcl6Z9cKpm0vFzIDoAwUri/k8kkgCWBWygUkM/nRQYleT30/GwWGnk+5LXKjJUbN25gbm4Ob775JiKRCEKhEMrlMnp7e4X3lFwOhvqp2LNZM7DUnmd8fBxTU1N47bXXxG8xk/Sv//qv0dPTI+q3sOr4RqBSqQT9gGOtVCrC8OG8rRQmX0s4h2FJUh4IKk3k6zEVnhSJ7SKb1grlWVCpVISno7e3F3v37oVGo8HFixdRqVQwMjIiatrYbLYt76m1ElY6Gx966CH09PTgW9/6Fs6fP494PI7e3l586lOfQqlUQj6fx4kTJxCLxVAoFOB2u3H48OFV2y7dcYVHDg9xM8osbdmiUavVyOVyqNVqYpPIm4HauzLMAADT09Po6OhAR0eH6FG1mWMhKBSoWSpDAvJnVlN2GmntSi+IksS31bidUM3bIdRDQSkfNCTWy14Oxsg5P1SEeaDK2T7bHcqDpZH3R+ZApFKppnYJpyeXxNdCoSD2vmxUyQRrueaKkrtD0FsHoI7Dozw8+dsqlaqupc2d2pMqlQr5fF7U3wqFQhgbG8O1a9dw48YNvPrqq8hkMoKXY7PZxLgymYzwNjMzj/eZTCZFxWj2MIzH49Dr9XC73ejo6EB7ezuOHDmCjo4OJJNJEbrl96239EKtVhOKo3KMjVKtb8e44vcpU/IBiCxg2XhlI8u3K9j3jcVqnU4n3G43LBaLKKhK7yyTY26nRt1Wwul0YnBwUNyvy+VCf38/7r//ftHS6Y033kA8Hke5XIbFYkFvb++q41uXwiNb+zIjXlmuHmicknvixAl85StfQbFYhMlkEi3s5YwnpbVcKBTquB9MQZcPp0KhgMXFRezbt0+4uzcLsjBlW4L29naRmioXapNd56spKrJLl5kny3l4mgF5rpUeLCVWqrOjJBw2U3mTlVOSstPpNFKplODnMF1VpVLBbrdDq9WKukqssCynjVKAsnw7cCtBuNlQ8leUr/H/SgWcz8nv94uiZs0ClUt6c/jMOS+yR0pWtJVcOWUWHr+bCRG1Wk0oEUwmSCQSIuTMhIvu7u5NW8eNZOWrr76K0dFRvPTSS6IDtsViESUQfD6faKsTjUbrvktWxDluhuWZ0VUqlaDVajE0NIR7770Xn/vc5wR94OWXX8b169fx0ksvwWg0wul04tOf/jQOHTq0bhJzuVzGj3/8Y1HdmB4devez2axQRjYaFlZ6yBi2AiDGaTQaRXSAz+3VV1/FXXfdte7fawYalbr4n//5H0xOTor1+NBDD2F8fFwUzbTb7RgeHkYmk0EsFoPL5dpW3mg5sqHcA8899xy+/OUvo1QqwePx4Kc//Sm6urpgsVhEWPe5556D3++HTqdDf38/HnnkkVU9WOvSDCgkZOG4XKw5k8kgl8shHA4jl8shmUziypUriMfjoiWDrJHzu+TFLytV8qKW02RloZbJZESK+J0A3bMMt3GDKol0jaznRjwe+TuIRlZPMxQEOfbfCHKoYDUhJSunsoW11ZDvWT4EmMXCOkh8n2uM90uPo8wpYCVQeh+2Exj/lhu3yvsNuHWtyiEfPi+GkpoJZbyfHh45VX09Xlal3JJDKqwD1iitnf2IbgdUtpnuzgxNGn/FYhEXL17E9PQ0otEo0ul0HafMYDAIZZRKmnxoyDWHgFtr1tBLzbkNBAJ48803Uast8QovX76Mubk5Me+VSgUzMzMwm83rVnhqtZpQEmV5zzXIpr23C3lNUx7LtdHk+5HDuM0yvtYLPrdcLodoNIpAIICpqSksLi6KKssLCwuigzrDdzyf5LIZdwIr8XFWg3x9KpXC+fPnMT4+jlKphLvvvht79uxBb28vHA6H+C0q8qQl2Gw2OByOVZ0da1J4lDyVtcDv92NmZgYnTpzA3NwcLl++jGQyCYPBIJqFycUDuRG5mblgmbUlW2Vk4AM3y+QbDAaEQiGcPn0aR48e3bSeU0pPDbkefLByFUxlWEPJF1B+FzemklS3HTKZaPXK8wPUZ7jw/8tlAAH1Xi/OaTM9ILwfrVYLh8OB9vZ25HI5UfeBXhx2yKaHkU1rqWyzu7HX6xWVTBvVNGkm8vm8qEos15cBGtcqkQ9IHvasR9WsCq3KewRuHlQmk0kcyrJXQ76GkI01/l/2FMvkVovFArvdXpdZyO9SVm/eCDKZDG7cuIHR0VFMT09jZmYGuVxOuOoTiQTy+bzgurCyM7PleK8yB4Vys5Es4SHPcdBoK5fLmJqawvXr1/HCCy+I72bDyba2NuH5OnfuHGZnZ/Hoo4+ue7yRSATRaFTse5L+a7Wlvk5UtIBb5ct6wXnl98vjlpVatVqNtra2W4oubldwbkOhEE6ePImf/exnIjLidrsRDodx+vRp0WOSYc1YLCbmfLPPFfl58v/ArbW8VhqT8rq5uTn85V/+JQDgrrvuwpe+9KVbCl/KuoFavVSh2u12r2ku16TwKG+KYSQW+eMGTafTiMfjmJqawtzcHAKBAAKBANLpNJLJpKg7w+JIZrMZtVpNcCLopaFWSgvG6/XCZDKJdPVAICAWNnk0Go0G6XQaU1NTgvh8JyBPUqMQlnwdsHwRLF6j3IzbyeJopKgplZ31fl+zQnRK8NCpVCoiREBiJjlajH9TCaKHR61WIx6Po1ZbIi0zHZrN+8xm8y2e0GaAhTJrtZpIK5YPfaCepCzPK/dUrbbUYDMYDGJychLt7e0bKui5WaByzT5hLCWvJJ/K61UeI9/j33I4VjZElO04ZD7Q7XjyXnnlFYRCIVy8eBHhcBjxeBypVErcP4vHySHvWu1m4U4eXnwWclhVrVYLErb8Pg9++XoeGnIWptFoFKRehtrp0drIvs3n84jH40J5I9+SPE7KcGXySiNvntLbrFSMOI8sI8JilPTk6nQ6sU44j6w+vZm4nT2vlLfyvEYiEQQCAfz85z9HNBoVNZW0Wi2efPJJ2O12GI1GmEwm6HQ6BINBlMtlRKNR2O12eDyeTc+Qlc+/5d5bDbLj4sUXX8TExAS8Xi/279+P48ePC4+i0vtMhV+lUqG7u3vFVHQZa3oCtVpNKCn8NwlRyWRSNNaLRCKYn5/HmTNnEAwGBSGO1j0zBliin0KEC5CaPzci/81O4iSZZTIZsdjpelar1chkMlhcXNxUguVatNSNHPyNQoJKwbuW37+TWIuCstb3ZWWv2YpArVYTzVgtFgtsNptoskllh3UgisUiUqmUCKHwcJCtUrqauSdY9bvZKBaLCIfDqFQqopAirV9iOQ8dDz5g6eBiDS2Hw9E0hUe+V+59pt0utw4bhbkaeYvk6yiTVjJaNoqrV69icXERZ86cEYU45fAo5SC5ihyz7D2WQ21KZVWuOUR5LXMkaRnzD5UnefxUeHgNib3r9cwynZgeOFnOK5vxyvMhKzNKg7ARlN5n/g6VG5ljKiuDLKy6mWgkw9fzWfn+qPgCQDQaxczMDF599VVYrVZ4PB7k83lYLBYcPXq0LkmHc55IJBCJRGA2m8U5udlQGknktK5n/Ay9v/LKK5ibm0N7ezvuuecefOhDH6q7jvMr7wmVSgWfz7fm2lirKjxMiXzttdcwMjKCZDIpCl9xQ5EjwKZwrPXAVEDeJDVtn88HjUaDN954AwaDAT09PWJTUdEJBALiIc7OzgouULVaFeEHZlJww5O1fqfd72zkKWfsKIUisPqi5wTKG/92LcjNgiwwl8Nynh55YcoWpbLa7VaC91StVhGJRLCwsIDJyUl4PB7YbDZh6TH7xe12I5/Po1gsinYiZrNZWLyyxVkul2EymXD58mX4/X584AMf2BZKTywWw9mzZxEKhcR8yHMgezdkd7e8fvV6Pbq6uhAKhfDf//3fcLvdcLvdWz6WRiEqq9UqyKhKBQbALUqD7PGRvTnys5FrLJnNZkGIJuT6SxvBb/3WbyGRSODuu+8WXnCmFYfDYcG5kjk9Mmq1mgifptPputflMLQcKqcnxWg0Cu8Vxy3zfRiuJdGXz0+v12NoaAj79+9f11jn5uZEWwugnlvD36dBzIbQMrdTVjobGZaN9hc9PLJxwuxfen3kFP7N9jjfTthIzgjlPDGp4qWXXkI4HMZdd92FeDyOcDiMj3/84xgeHl4xFZvfZTQa7xhVolKpYGFhAXNzczhz5gyOHj2K/fv3C+I/Ie87eme0Wi1OnjyJ8+fPY3JyEk6nE1/4whdu6eiu3PuUXVqtFl1dXYLfsxpWVXhSqRRGR0dx+fJlXL58Gel0ui7NkMRbpbeFi5eLjhuYrGuLxYKxsTFUKhWRSim7qzlB8uKk5cNKi+RcyDyarairUKvVRMVSps3Lvysf+I0gKzdKQb6dwj4bvQ9l+JNjWq53z1aiWq2KlE26ypl2W6lUBLnSbrcLBTqdTou5Jp+F3g/yLHQ6HZLJpPAe8bVmolAowO/3i5RqYPnikABuWYdyZlQul8PU1JToeL0doExDBpY3MpT7TPm6rAxRVjGTlNVcgfr1vBGwhg/lmMvlgs1mQzqdhsfjQTabFWRmhlNlw4FKDNetPAauZ3lsshedHjEqA8rK5/w/w5l2u138Zn9/P7q7u9c1VnrjlUq1rEDKobZGSi0hhxZXghwqkw9GmaPVbENkLeA4YrEY5ufnEYlEkE6n4fV64XA4YDKZVmyLISsWVHrWM+5G65thznA4LFo50KBYWFjAzMwMLly4AJPJhFqthrvuumvFRrvUB+bn5zE6OgqHw4EdO3agr69PkOob3bO8DzQaDdxud53XeaU5XlHhqVaruH79Or7xjW+IBy4/OJLfCG5ApRuxWq2KwyOTyeBDH/oQ9u3bB51Oh/Hxcfzyl7+Ew+EQHiGr1Ypdu3aJugKszMzUWD5kebOwEi41/DsJeUNx3HTZNvJ28DPKcI4yhi5/Fz/TbCgtLTlEJV+jRKMNw4JoNptty8fG3ysWizh//jxSqRT6+vrqDjcWtNJqtfB6veJ+VaolfoPP50OpVEI2mxVN+ZLJJPR6PaxWK2KxGIrFIubn5+Hz+dDR0bGlY1QimUzi6tWrdYIJuDUFnc9AznykUk8rLB6PC0VxO4HkTMqARiFUQvbuyKF0htq59+jB6+joQK1WE2NeSWlcK2q1mgijskgclZlisSioAX6/X9AFmFas9CzKnjp6u6ikWa3Wun9zjTKrUM6OojEij4ukfoZ4N9IWRuZCydwaOfyRzWbF+pO9bfzMWkLhctiO88oDk1EAubGmXL1/O0J+zqdPn8aJEyfgdDphNBpx48YNPPzww3j66adXPOd47iaTSVGHZz1o5DXVaDQIh8P4zne+g5GREZw5cwZOp1PUw0qn01hcXMSJEyfg8/nw7LPPYt++fXX3RDCzNRqN4uLFizh16hS+8pWvYGhoSDhR2HdRybNjqLRarcJqteLgwYNCGV9tb674FK5evSqa5tVqNdHMixuE2iNQ3y+lVquJhoNGo1E8MJ1OB6PRiEuXLiEWi6Gvr68hJ4ALuFgsYnFxUWTL0Hrh78uWnbw5aJVvVs8b5SbjIcB74e83ul65aGTLme/L1xeLxW1RDItzsJxHYK0Ki6wsKbkIzUC1WkU4HEahUKhbH7IFzfmRQwqyMAVuejZ5GLLytkajQSqVWtGy2SqUSiXE4/EVSwHIh2Yj8LkwSYGZlVuZScj9pmwTANy0EomVBLusDCkPPHl/cn3b7Xbx/OS5v531S+Iu94KyH5tOp4PdbhfZJy6XC8ViUYTzZU+4LFvoTacBRSVF9uRQwaOMUSoTsidWo9EI5Z8ZiswYWyucTie6u7vreEJyiEoZUpXlKiHLIaXcVELOyOI1sndH/p2tAOeXUYh8Pg+n07mmcykWi2FsbAwLCwuoVCqIx+Ow2Wx4+OGHMTQ0VFf5upHxKSuUG+EpyWtdeT75/X4EAgFEo1FUq1WxplkygfXMpqenYbPZ0NXV1fCZBwIB/OIXv0AikRCVvdm2RelUkCF3ujeZTOju7q4LaW3Yw3PmzBlcuXIF0WgUHo8HDoejLhZKvg5/hFpZNpuF0+kUnXoBCI+PXq/H66+/DovFgq9+9atwOBx4z3veg0gkIhj96XQaY2Nj8Pv9mJubE4uVk1yr1W4hKtKdW60uVYaNxWJ3rMkfJ4MbVhaUfJ+gsFS6zeVJkS0b8kaaDaUHTbmI1qr0yAJGdsM3C5VKBX6/HyqVCg6HQ8wZ167sYVN6CeTQAa3ERCIBlUol6vjodDokEok72mByrSiVSohGo6JT8kqhq0bg3HNPM5xXLpe3tCcPCeOywiMrL8oO2/KhKv/N6/m3rOzJn+c19DrLhtXthLOAJcWYBfDI15GhUqlgsVjgcrlEOxPKG0KeL94PvT9U0imfqeDQgCQHiZ4VuVcclalcLidCYCRIe71emM3mdSk8Xq9XhEQ5NpkQLntzZGVN6eGWlR7OkezRkQ1h5fzKmWn8Tn7uToNGQiQSQTabRSwWw+Dg4JpkQyAQwMsvv4xUKgWVSiUKTz799NO37D05JCivYdZ72gjnbLnnwxBUMBhEKpWCRqOp87AyPJtKpTA+Pg6TyYTOzs6GZ8XMzAyeffZZ7NixA/v27UNHR4dQXGQDQwnqCOQJ9/X1rbnVy4oKj9PphMvlEilt5AIYjUZYrVZRil3eKHq9HjabTcSmAYhsAHmxV6tV/NM//ZPo7xGJRJBMJutSgeVQkbxJqGhwImXFo1KpIBqNYmFh4Y51wVUKIKKREsDN1mjyZMFJpYfKZLOxksBYC+TxkovVbDAsY7fbxQFAvph8ODKcw/mUw7NUjHgtq4WzO7FarRYcjGaBXk6my2u1WpjN5obeEaBxeQQeMmzMOzs7i4mJCQSDQUxNTWHnzp1bMqc8lGVBLnuOKdiV/A9eK2e6KN/jASpnpAE3n4vX6xX9qkimlb0FGwUVOJ1Od4snUGlk8P6Y2i3zFzgOeZ/Kc6j0gLPIppwK3sh7Qvkj9zNkLaD1YGZmBuPj4yKrlkoNzxImt9CzI3vA5bHwfaWi2cgLLStwDBfyOuCmnN0KL084HEYsFkMwGBRnAL2kTOZRgkbUwsICRkdHhUPhj/7oj7B79+6GYSzZmJaVyEplqYmoTG5fKy5fvoxSqYREIoGdO3eis7NThAfZlgRYKsZqNpthsVgEOdpsNqNQKODFF1/E5OQkjh07VqekpVIp/Mu//AtGRkYwMzODRx55BO9///vXnFpeLBaRzWaFgbmeuVxRYjGdzWQyiU3FhcS+KnIKIF2KFLBmsxnZbFZsXL7HxnwjIyOwWCzo6upCIpEQdUyY8SU3BmRcVsncl4UesLQJGPPeCqxm7cnhBKWFrfQc8EBVEg+bBSVBcz2hLPk6WVtvZuycHDKmMvNeeADQSlEKUnmelAKZGzmfz4v6O80K/RC1Wk1U8C0UCmJcslu/UYYKDRH5e9RqNZxOpyhglkwmEYvFtmweleuvVquJvkDK0IVyLHxPeRAov1t5YHBsVqtV1Apr5MndKOQDWQm5oKC8HmUlRrk25T+yla/kFPKAIOdCXuf0mMjjp8xlmI2emrUiEolgZmZGGAgcs+yVocJHw1mW5UoZojS8lPuT/5ZDW7L3XJ5j+fo7tU9Zx6hYLAolV7nHlKhWqwgGg/D7/aJFhMViwYEDB7Bz5841/zZ/Z6N98EKhEJLJJGZnZ8Vrer0e4XC47jt5npMTxvlVq9UYHx+HRqMR4TidTodsNotgMIhTp05henoaxWIRXq8Xe/fuXbNCzSSTRo6H1fbmigpPW1sbFhYWBCmO8blisYhEIiGsY2YduFwusWjVarXwVtATxMVFHgezAKampoQ70+12C8uH1/NhUeGRLW7eTywWE4MNBoMYHx/HE088saYHuF7IwkfGcpYkM3kavcfPcVyszdFskBshj1H2rq1V6PMzLDfA5rDNANM5+fuy50AWTABEyLRardZlaQEQSqlWq0VnZ6cgmtJzxMrLyWRSZMZsJcrlMiYmJjA/Pw+g3iMpk+t5uFCAyOUR+JlSqYSZmRlEIhFUq1UEAgHMzMzg0KFDWzYe+ZAql8twOBzweDwND0GlW7+RUrTc4cY9yGrK3d3dCAQCdTyuRvt+MyF7JmSstOfWs5+WU/5W+8xGFIKpqSn85je/ESnn9O6wl5XFYsGf/MmfYHp6Gt/+9rfFucDzhJ+RQ5VKoxG4qcDLVABGIw4dOoRkMgm/3y8Kh8pgLzJW1t5MDAwMoLe3t2598pxbDqlUCl//+tcxOTmJ6elpfPKTn8QTTzyB9vb2ZT/TiFvGaMHMzExdr7W1gt0Rvvvd78LlcsFut8NkMqFYLGJubg7FYhEWi0XIjFAoJJrRyhSGSqWCb37zmzhw4ADuuece/OAHP8DIyAhee+01mM1m3Hvvvejv74fD4VizMc0SDmwBI2O1tb2iwuNyueD1etHV1VXXMVijWWrcR4WHg6ayI3sp5KJTSrciF6pKpRLuU5Lv9Hr9La5WDkjpYWIsjxq1HI+/U1DG9VeDcjEq3c6yIGqWQiBDeW8rxVSX+7x8eKrVapGp16zxMTwih6pk76QcPpXHKrcu4Ou0cKxWq+gbR44EladCobBuq3izxrmwsCBqRS1X20T2Figta5mjxhRplUqFVColiLxbCXlvsOuz0lPKfys5WPLn+bq83yiH+DkqgTabTYQstxpK4d9MT+9GQQOH60nm4VDm79q1CwaDAQMDAwiFQiL8wvXI+eZrK0H2fDFjbdeuXQgEAhgdHa27ls8znU5jfn5etAzaTHDM8vwUXwAAEENJREFUly9fFtzMtrY2uFwutLW13RKemp+fx8zMDMbGxpBKpdDe3o7u7m709PSs+954DlMurRfd3d1YXFyESqVCIpFAOp0WmW+soSO3M5G5p/LZlslkcP78ecTjcQQCAVy8eBFTU1MoFArw+Xw4dOgQfD7futY3ZWyj0h+rfc+KCs/AwABKpRLe/e534/Tp05ienobH44HJZILT6RQ/yAZ+srXIA0Qu7y0T0tRqtdDk5biqkpRGK5MN9rgwmfFFIjTJorlcDmazec3xwNuBUtACy3t5lvMGKYWyLHibCVngyArPWpU8pfcKQF22XTNq1CizsNjPjSmrlUpFWNiyckCvG708rLljsViwY8cOxGIxxGIxuN1u6PV64RVi1WWTybSl42TzyZmZGVFyXp4//q1MEaYwY4E6ed7y+TxUKhWCwWBdIsFWo1aricrrHId8kCqzRWXlhp8n+LocxgRukk27u7vrSMvbxRh5u4D7jUqj3B4DWAqRHDlyBP39/ZiamsLZs2cF54dthNghnvuToJdE9l5SuVKr1ejp6UFXVxeefPJJXLlyBb/61a/q5ptYXFzE2bNnhRdjs5FMJvG1r30Ni4uLiEajeOSRR3Ds2DE89thjddWRAeC1117DG2+8gTfeeAPt7e34wAc+gP3792+Ii8r1nEqlNqTwPPzww9Bqtdi5cycWFxfrKCIq1U0eL5Udchb5Go2nbDaLF154QXzW4XCIYpMDAwP46Ec/ektK+WpKCzk8er2+YYHXlT6/Kuuwra0Njz/+ONra2rBv3z5R1yMSiYhwluxxkUMh8sAb8QbkzABlLFqOw5KIJm8cVhrlRkgmk9DpdOjo6MCBAwdw8ODB1Ya2YdA6IU+j0fuysAVWV4gahYvW63reTJCwptTaieXurdE4KWS4SZplrbKuDrOMrFYrotEo8vm8CGGxMa0cxmDYVCZERiIRFItFWK1WWCwWodRQgcpmsw15MluBanWpuCLHlslkxL0BEAdIIpHAY489hl27dmH//v24du0avvOd76BUKgmDgnNFpS8ajQoS5lZCXjNMipA9BSaTCRaLRSRHLBf+ABor61SC6e0rlUqw2Wx1pOLt5IF9O0FJaVCiq6sLH/rQh3Ds2DHEYrG6ekg0linzCVmJpfEkc3GYIbx//36x/oF6ojMbBF++fPmWBpUr4a233qoL1TPSodFohDdVNuTZPzIYDOLSpUvI5/M4dOiQaKQdi8Xg9/tx6tQpXLhwAV1dXTh06BCeeuqpdfN25HHy2Ww02qHX6+H1ekUZDyZlcB5ZaLVWqwlSuZJwzmgQDShGcMxmM1wuF7q6utZUwkPuL8d6aeQXAzfL4qxWYHFVhcfpdOLee++F1WpFR0cHgsEgCoUCEomEGCy7SHPQchEvZdYBcHNiuIBlb5DyWuAmeZITx5oGHCDrDXg8HnR0dGDXrl3Yu3fvqg9xo2hEflO+r+QLyMqPfJ3y38pDslnKAbkoVASWs5YbjUup6PEzBoOh7hDdajAVuFwui2yCeDxex9+xWCxC4ZHXqRzWYq0dlUolNh3HxWxFOatmq1GrLRUBo+eVGRXMRtJolhrz5nI53HXXXXjwwQdx//334xe/+AX++Z//GSqVqi4DEoAI1yWTSSQSiaZ5IWu1msg2kr3CtPYYapeteDnDTv4e/k3PluxJLpfLsFgsdSnVwPZp/fJ2AMnl9PwvZwC43W7cd999Db+Dc5HL5eo8zsDNOeHZs1yPt2vXrjWUURqNBplMBtPT0+uqID47OyvkgMFgEJnJNBKoBDE5gmuRv1UsFkWpF61Wi0gkgomJCVy9ehXj4+PYt28f9u/fj/vuu29DmZCyXOaz2Qj0ej3cbrfgtVGha9RUluR+OYOVf5N/y+eg0WjgdDrh8Xjg9XrF7610LtABAtwkLZPDw/NW7riwHFZ8ErKC0t/fj87OThw4cEBoxefPn8fJkydFhVkWx5K1aNlS5neSAM2Hytx7FitUCicOgkxw9uxSq9UYHh4WMdHOzk709/evSPDaTFCjVWrQstImh/CUm10WosrPNhs8RADUzaF8f414R0rrWq1WC4XB5XLVZddsNTKZjDisWTp/bm5OKHZUyBhGZe0ak8kkQlXy3DDV12KxiCq2DNXWajWEw+E193jZTOj1ehw9ehQqlQqXL1+uO/xrtVqdtcwMC86lXKSO88RQnsFgQHd3N/r6+ra8zIC8xmQODwUdFRatVisKk/HelRV8lcoOM2h4sMbjccRiMTFuHtg8fLdD2Yi3Ax577DHs378fIyMjeOutt+pKmCj5ZMuBnoFGaelKxWc5GI1GwUPlGuGh6XK5cPDgwXWFs9igOhaLie9keJhZcCxEytIOg4ODaGtrQ3d3N9ra2jA3N4f5+XlMTk6KenMajQZHjx7FH/7hH2JgYEA0k+U9r1VuqlRL2aNOpxNHjx5Fb2/vmscmY8eOHfjEJz6BRCKB0dFRRKNREamht0aO8HAvyZxIzhkNSHqK/uAP/qAuCtPIcJbfk0EPrN1uF/O2Vs/rmqUWw1cOh0O461lUiXHCeDx+i0CgG52KDHCznwywJLzI0KZSw4cqQ3Yh8vBRq9XYu3cvXC4XOjs74fP50NXVtdYhbRiygGxU50NW7uQMA+UGbeQl2i4Kj1arhdVqFQqCXNyLaOShUipB5HYxTNLMppo8FJUeOrmDNA9HWWllmIReSI6HQpMuXX6fTGxuBtdFq9Wiv78fc3Nz4p7lg1/2UtD1Ln8WuLVUAj/j8/nQ3d29YqbJZkLpWaQc4DPWaDSiWjsFMJUgXqssCyH/m3vUZDKhUCgI5ZZzrvzcdtmfbwd4PB7ReBSAqNgPLKWjO53ONckCzmOj19cCg8EAn88Hv98vPK80eNrb27F79+51VUZnnz2G3FgcknuKGWM0+u12uxiv2+2GzWYTzbAnJiZEQc/Ozk60t7djz5496ybyKsG9Ti/vRmA2mzEwMIDh4WHMzMyIqs3hcFjIA+4zenvkxBDKPp7VNDQtFgv27NmDnp4eAOuvXJ7P55FMJjfkaV1R4VnugZvNZtx99904dOgQPvGJTwBYOkzYXDCZTArBwU7N7e3twtXc0dFxS2x8PeCBKgvBlUJMmw0W4qLSR/em8pCQ3XC8b14LQBxEMpaLc281bDYbhoaGRN2EbDZb5/UBbm4qmUAogyEQvufxeLb0sFSCXAIeZlTUqZjRouf8JZNJ5HI5UY+FLUvi8bjIVFpYWEAymazjX1H4kS+01TAajXjssceQzWaRyWSEQudwOOoyG4H6Gkm0Clm/R3bFFwoF5HI5HD9+HI8//vi6i9BtFBSsDBMolRCfz4cHH3xQcIyYTCFX4iUaVeSV67UkEgl4PB7s378f/f39Yi0on1lL4VkbWFzxH//xHzEyMoLPfe5zovn04cOHsW/fvhWzGOVQupzaLUNpdDU6B1wuF44ePYpTp04JRcPhcODo0aN44okn8LGPfWxdMunJJ5+suz+iXC6L4rnMbMrn8wgGg0L5icfjSCQSmJ6eFvuzo6MDBw8exAMPPIDe3l50dXUJw2O5ek2rQafTIZ1O44c//CFsNhs+8pGPrPs7LBYLdu7ciS996Uv4i7/4C5w5cwZjY2P47ne/i4WFhboaPWvFvn37sHfvXhw+fLgunLUSlPO6uLiIq1evwuPxwO12i2vWgg37pSkAODEkDBeLRaHZ0/oCIPrDqNVqEe98u4IaPOOYKpVKWPvyNctxOPjMyCVh48pqdanKZjOr9BJUbtgHJRQK1WXf0a2oVHiUvB6bzSZIpsxiapaHh4KImTfKvl6yB0er1Ypq4STpVqtVOBwOqNVqJBKJOm9CoVAQcWqr1QqPx4MdO3Y0rcWERqNBX18fPvnJT+LcuXOYmJhAPp8XcXgqYlRkaJFls1lhvWq1WqHU9/X1Yf/+/di9e7fIkNoKKDmAVE6pqLrdbhw/flx43NhAuBFhuRE3UPYa5XI5ZLNZDAwMCO6C8gAlKb2l9KwNavVSfTZmzVJOkufGOaFxqFxXSq/pSrJDnitluJJGNkMuOp0Ohw8fRm9v77rDs8spIGytQD4ZlXCv1yvOAZZ3GRwcFGeIw+GA1+vFjh074HQ6b9sgVKlUsFqt8Hq9GBoaui2Kh8wDGhgYEFWUE4kEYrGYUGCj0ahoPyOX5eA+Yej8sccew4EDB0TzZf7GavtJJiP39PTg8OHD6OzsRHd3d53RstrZsmmBeLVavWaN7e0OkqaZvgosCWaGf2QB3ahZJiemUqmIA5EHZjQaFV6RRkJ7q8CFPjAwgO7ubpw7d054NtiWQa6rRHDcjOt6vV709vZicHBQKMHNQiKRQDQahdForMsmVIbgyBdoa2uDRqPB1atX60I6XV1dYo6Am+uBvYdcLhc6Ojqwe/fuptThIQYHB/HFL34Rf/u3f4srV67UcVLo5mZfGq7VVCollAceMrlcDvv378dnPvMZHDhwYEt5ScoQOTlILLnv8/nwyCOP3NF7kA9RZuy1sHbQ6KFBUSgUkMlkxLqjASmHEYG1c3QaQVZu2WCSRjf5evfdd9+6sqBWg1qtFh0GlCnnWw273Y7u7m685z3vwa5du277+8jj7e/vx7vf/W7x+sLCAqLRKMbGxhAOhzE9PY14PI50Oo1kMik8rSwl8fu///sNM6hXVVQkpXTv3r2wWCzo7e0VmWO8ZssUnv9PsFqt2LlzJzo6OlAqlUR2AIUwUJ8WKPNagHoFwWAwwGAwIJvNQqVS4eDBg9izZw+A7VFsjJkIQ0NDggNDr4Dc94xgxhKLtjGLSSnMmoH3v//9uP/++wXfQ6/XC0uTodeuri5hnXR2dsJoNIp4OtMrVSoVhoeHAUCEgB577DGh0NFrR+Jis2A2m9HX14fPf/7zeOaZZzA7O4tUKoVQKCSIkocOHcKePXug1WpF7Q+SuO+55x60tbVh9+7dgqy81SRss9kMh8OBrq4uOBwO5PN57N27F/39/VuiPHu9Xjz44IMiPNnf34+enp5tsTffTvB4PPjkJz8peBxdXV3w+Xxij8jhys0C54itMcxms+CI8nffyfPodrvx0Y9+9I7WpCMnyePxiPo4NFJkY5/1y/r6+m77N7u7u+F2u0VG5noU45bCswFotVqxgarVqmicprREAYi0OaXCI5MvmR6pUqnEZG4H0OXPlh8ELbR4PC7i0BwbCXoMa251W4WV0N/fL7xPQH2bBbZhaG9vF2NzOp0wm80N52MryPG3C61WC5vNhnvuuQd33303Ll68KKwwtm7x+XwiRGU2m9Hf3y/c0gcPHkR/fz/uu+++piluzA5j00uj0Qin0ylCi8Cd9YSaTCZ0dXWJjtc2m62ptaTerjAajdi3bx90Op3wgNjtdmG530nlg+F5uSfjdmhmfKdhNBqF8Xwnf4MG7lbBYrGsi2QuQ9WKRbfQQgsttNBCC+90NDfG0EILLbTQQgsttLAFaCk8LbTQQgsttNDCOx4thaeFFlpooYUWWnjHo6XwtNBCCy200EIL73i0FJ4WWmihhRZaaOEdj5bC00ILLbTQQgstvOPxfyQe5BtZ2cv9AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 720x720 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "fig, axs = plt.subplots(1, 10, figsize=(10,10))\n",
        "for i in range(10):\n",
        "  axs[i].imshow(train_input[i], cmap='gray_r')\n",
        "  axs[i].axis('off')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SZjVIoyqmVy-",
        "outputId": "7a29e971-32f7-4c16-8d45-3ac8bd6783d4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[9, 0, 0, 3, 0, 2, 7, 2, 5, 5]\n"
          ]
        }
      ],
      "source": [
        "print([train_target[i] for i in range(10)])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Rx32OX7NqNnD",
        "outputId": "2199611b-2312-4116-9b51-23f23770303b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9], dtype=uint8), array([6000, 6000, 6000, 6000, 6000, 6000, 6000, 6000, 6000, 6000]))\n"
          ]
        }
      ],
      "source": [
        "import numpy as np\n",
        "print(np.unique(train_target, return_counts=True))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lrhnk0zXYj5e"
      },
      "source": [
        "## 로지스틱 회귀로 패션 아이템 분류하기"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "2HSAg_UFNH52"
      },
      "outputs": [],
      "source": [
        "train_scaled = train_input / 255.0\n",
        "train_scaled = train_scaled.reshape(-1, 28*28)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kwOFVfpxxuLU",
        "outputId": "8e7fb44c-3c9d-448c-b94a-5c4f6cff0fd4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(60000, 784)\n"
          ]
        }
      ],
      "source": [
        "print(train_scaled.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ui3iJlfLNLzr",
        "outputId": "07b65376-1ba5-4636-9c1b-fe43ffb6fc50"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0.8195666666666668\n"
          ]
        }
      ],
      "source": [
        "from sklearn.model_selection import cross_validate\n",
        "from sklearn.linear_model import SGDClassifier\n",
        "sc = SGDClassifier(loss='log', max_iter=5, random_state=42)\n",
        "scores = cross_validate(sc, train_scaled, train_target, n_jobs=-1)\n",
        "print(np.mean(scores['test_score']))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Y0W7D4ED3r0B"
      },
      "source": [
        "## 인공신경망"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zaFtnKmzcQeJ"
      },
      "source": [
        "### 텐서플로와 케라스"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "_WOZURhzcTAi"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "rbKFPAUucVkV"
      },
      "outputs": [],
      "source": [
        "from tensorflow import keras"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YdNdad0mcoGD"
      },
      "source": [
        "## 인공신경망으로 모델 만들기\n",
        "\n",
        "인공 신경망에서는 교차 검증을 잘 사용하지 않고 검증 세트를 별도로 덜어내어 사용합니다.(cross validation - lecture 2)\n",
        "\n",
        "1. 딥러닝 분야의 데이터셋은 충분히 커서 검증 점수가 안정적임\n",
        "2. 교차 검증을 수행하기에는 훈련 시간이 너무 오래 걸림"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "4smUJC1hOWTF"
      },
      "outputs": [],
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "train_scaled, val_scaled, train_target, val_target = train_test_split(train_scaled,\n",
        "                                                                      train_target,\n",
        "                                                                      test_size=0.2, # 훈련세트에서 0.2%를 val set으로 쓰겠다\n",
        "                                                                      random_state=42)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "G9OwZTRorHWQ",
        "outputId": "bb977b82-7323-4526-8abc-c1754faaff70"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(48000, 784) (48000,)\n"
          ]
        }
      ],
      "source": [
        "print(train_scaled.shape, train_target.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Van_eJ7BOo-y",
        "outputId": "7d07f834-53c6-4c97-83e7-ab0d098d06c3"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(12000, 784) (12000,)\n"
          ]
        }
      ],
      "source": [
        "print(val_scaled.shape, val_target.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "DEXlpF_8gmJv"
      },
      "outputs": [],
      "source": [
        "dense = keras.layers.Dense(10, activation='softmax', input_shape=(784,))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "k8mtAypNPIz7"
      },
      "outputs": [],
      "source": [
        "model = keras.Sequential(dense)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xPTmHpMe3r0D"
      },
      "source": [
        "## 인공신경망으로 패션 아이템 분류하기\n",
        "\n",
        "- 이진 분류: loss = 'binary_crossentropy'\n",
        "- 다중 분류: loss = 'categorical_crossentropy' => 타깃값이 원-핫 인코딩이 되어 있는 경우(0,1로 바꾼 값/ 타깃이 1, 타깃이 아닌 값 0)\n",
        "- sparse_categorical_crossentropy => 정수로된 타깃값을 사용할 때"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "VqJ5dpwwPWPr"
      },
      "outputs": [],
      "source": [
        "model.compile(loss='sparse_categorical_crossentropy', metrics='accuracy')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JLLytXw7UcvB",
        "outputId": "25dbee38-da77-4293-c1cc-393630f0dc24"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[7 3 5 8 6 9 3 3 9 9]\n"
          ]
        }
      ],
      "source": [
        "print(train_target[:10])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KiZnrOgQP4ps",
        "outputId": "c30155bf-82b7-4c0c-b18e-e6eca1df3c58"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1500/1500 [==============================] - 5s 2ms/step - loss: 0.6075 - accuracy: 0.7944\n",
            "Epoch 2/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.4778 - accuracy: 0.8394\n",
            "Epoch 3/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.4557 - accuracy: 0.8472\n",
            "Epoch 4/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.4441 - accuracy: 0.8519\n",
            "Epoch 5/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.4365 - accuracy: 0.8559\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fca1866be50>"
            ]
          },
          "metadata": {},
          "execution_count": 19
        }
      ],
      "source": [
        "model.fit(train_scaled, train_target, epochs=5)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OdXPQEXOQIFm",
        "outputId": "9a1ba594-9708-4fee-f6bb-85fc23e0b79e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "375/375 [==============================] - 1s 2ms/step - loss: 0.4444 - accuracy: 0.8490\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.4443625509738922, 0.8489999771118164]"
            ]
          },
          "metadata": {},
          "execution_count": 20
        }
      ],
      "source": [
        "model.evaluate(val_scaled, val_target)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "input layer\n",
        "hidden layer\n",
        "output layer"
      ],
      "metadata": {
        "id": "fUmtu880DmYu"
      }
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "7-1_Artificial_Neural_Network",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3 (ipykernel)",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}