tf_tutorial_02_classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "tf_tutorial_02_classification.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyMV4Hf3/QGtbs6pPTSvufJE",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3.8.10 64-bit"
    },
    "language_info": {
      "name": "python",
      "version": "3.8.10"
    },
    "interpreter": {
      "hash": "0adcc2737ebf6a4a119f135174df96668767fca1ef1112612db5ecadf2b6d608"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/churry75/tf_tutorial/blob/master/tf_tutorial_02_classification.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YKSSCc4Hfl6r"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kPJQ4KWRuUbq"
      },
      "source": [
        "# 패션 MNIST 불러오기"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_E59WVj8uaWW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0e7fe267-4f9c-4513-f44b-d4083416f10c"
      },
      "source": [
        "fashion_mnist = keras.datasets.fashion_mnist\n",
        "(train_full_images, train_full_labels), (test_images, test_labels) = fashion_mnist.load_data()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 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",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "8192/5148 [===============================================] - 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"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fK2zJDIlupAI"
      },
      "source": [
        "1. 데이터 셋에 클래스 이름이 정의되어 있지 않기 때문에 0 ~ 9의 레이블에 맞춰서 클래스 이름 선언"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FCtbcxX0uzne"
      },
      "source": [
        "class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',\n",
        "               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FOLKBpcnvQ5O"
      },
      "source": [
        "2. 데이터 확인"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vGZvsKfcvWA1",
        "outputId": "bd794821-7e9c-45c7-8cd7-3d329d900063"
      },
      "source": [
        "train_full_images.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(60000, 28, 28)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Z4rwAGdnvX2w",
        "outputId": "ec1b0188-160b-4549-da13-2b41e8ac48a3"
      },
      "source": [
        "train_full_labels.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(60000,)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fbodlzfivbpR",
        "outputId": "b790b92e-bd35-45fd-f8f4-3b21db2213ed"
      },
      "source": [
        "test_images.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(10000, 28, 28)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "apyReQwIvo8i",
        "outputId": "efa9dc93-d509-43ec-d5aa-fa7a3167e8e5"
      },
      "source": [
        "plt.figure()\n",
        "plt.imshow(train_full_images[1])\n",
        "plt.colorbar()\n",
        "plt.grid(False)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8pAKS7Nbvg9w"
      },
      "source": [
        "3. 데이터 전처리\n",
        "    - 훈련 전 데이터 전처리 단계\n",
        "    - 이미지의 값을 0 ~ 1 사이로 \n",
        "    - 전체 train 데이터에서 vaild 데이터 분리 및 정규화"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gBN2sjjp2Orb"
      },
      "source": [
        "valid_images, train_images = train_full_images[:5000] / 255.0, train_full_images[5000:] / 255.0\n",
        "valid_labels, train_labels = train_full_labels[:5000], train_full_labels[5000:]\n",
        "test_images = test_images / 255.0"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iH_rpeCt8IED",
        "outputId": "66127668-e148-4feb-9adf-72546a93a27f"
      },
      "source": [
        "valid_images.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(5000, 28, 28)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ksCMglFA8RqP",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bc8f869c-2084-4be6-b382-ca1ec6ffade4"
      },
      "source": [
        "train_images.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(55000, 28, 28)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 589
        },
        "id": "NFhkAYYDv9S7",
        "outputId": "cbc4637f-61e4-45e1-efd5-e9b724a2c172"
      },
      "source": [
        "plt.figure(figsize=(10,10))\n",
        "for i in range(25):\n",
        "    plt.subplot(5,5,i+1)\n",
        "    plt.xticks([])\n",
        "    plt.yticks([])\n",
        "    plt.grid(False)\n",
        "    plt.imshow(train_images[i], cmap=plt.cm.binary)\n",
        "    plt.xlabel(class_names[train_labels[i]])\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAj0AAAI8CAYAAAAazRqkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOydd7hU1fX+362xYqQIKNKbgoKCgAURkaixoGg0ahCNNXZN+Vpi/KkYH8w31iR+IzH2WFAREYxgBAWlCNKLqHSQIoIIFjSW8/tjZjbvXsw5zr3cuWXO+3keHtaZs+fMmbPP3nPuetda20VRBCGEEEKIUme7qj4BIYQQQojKQA89QgghhEgFeugRQgghRCrQQ48QQgghUoEeeoQQQgiRCvTQI4QQQohU8KOyNK5fv37UokWLIp2KyMfSpUuxbt06V9HHLWZfbt68Odj+/vvvvf3VV18F+9auXevtVq1aeXunnXba5vP4/PPPvb169epg39577+1t57ZcXvu522+//TafBzNt2rR1URQ1qNCDQmOzKqiJY1PEU4yxqb6sGpL6skwPPS1atMDUqVMr5qxqIFzTiH8oi0nXrl2Lctxi9uXcuXOD7S+++MLb8+fPD/Y98MAD3n766ae93bp1620+j/Hjx3v79ttvD/b98Y9/9DY/2LRs2TJoV7du3W0+D8Y5t6xCD5gl7WOzKqiJY1PEU4yxqb6sGpL6UvKWEEIIIVJBmTw91RH2KrzwwgvBvsmTJ3v7u+++8/Zee+0VtGvfvr23jzrqKG8fcsghQbvK8u7URJ588klvs6wEAA0abPEy2r+OGzZs6O1OnTp5u0mTJkG77t27e3uXXXbx9tixY4N2Cxcu9PbXX3/t7Z/+9KdBux//+Mfenj17trcnTJgQtGvevLm3Tz75ZAghhKi5yNMjhBBCiFSghx4hhBBCpAI99AghhBAiFdSImJ5333032L7wwgu9zZHx3377bdDuRz/a8vW22267vDYQplHzvn322Sdo97vf/c7bF110UUHnXsq8/PLL3p4+fbq3+/fvH7RbtWqVtz/99NNg3yOPPOLtESNGePuee+4J2r3++uvebtOmjbfXrVsXtON91113nbf79esXtHvnnXe8zTFHHMMDAIMHD/a2TWe3cUJCCCGqN/L0CCGEECIV6KFHCCGEEKmgSuUtrtRrJSeG08iBUNKoXbu2t7l4IADssMMO3mbpy1bZ5RRo5r333gu2BwwY4O3yyltVUeCwWCxdutTbLP3MmzcvaNesWTNvc+kAAFi8eLG3WS465phjgnavvfaatzdt2uRt7hMglKrq1avn7RkzZgTtuB/4PlmyZEns8WxhRclbQghRs5CnRwghhBCpQA89QgghhEgFlS5vFSppcZbPnnvuGezbeeedvb3rrrt6u127dkE7zvpiKckejyWX5cuXe7tOnTpBO67iy9lKBx10UMy3CL8vkPydaxqzZs3y9r777uvtzz77LGjH2XFWStxxxx29zdWUd99996Bdr169vM3ypM3Y27hxo7c5a4zvGcs333zj7Y8//jjYx+/74IMPYo8hRFqwYQQ8t/L4s+3++9//envFihV5XwfC8ciyuZ2PuZq7lc3r16+f99zt+F65cmXedqJ0KZ1fYCGEEEKIBPTQI4QQQohUoIceIYQQQqSCosf0FBrTcthhhwXby5Ytiz0Ga8is0XJ8j33fokWLvM1xO0AYj9KiRQtvs+4MAGvXrvU2p1Tb78TnZPex9mxT52saHO/C14avJwCsXr3a2y1btgz2cfwP999uu+0W2477lXV9IP6a8orrdpvjEOz7OS7I3oelVH6gKinPdUyKK+GYENtnHEOmPisfSdft/PPP97Yt/8B8/vnn3ub+AoA1a9Z4m+N47OdyOQlb6Z0rq3ft2tXbffr0Cdo99dRTsedYk0kaH4XG1RbznKpy7MnTI4QQQohUoIceIYQQQqSCostbSW6s66+/3tucrgyEVXxtWjLLKpwObeWoDh06eJslDJv6yMfgKsOWVq1aeZsrQXNVYQD41a9+5e0HH3ww2FeTJS1e3BUI3ZUsC37yySdBO75u3A9A2EcMy0pA2Oec4mqrP3NZAS5NYN3fX375pbdZVuPXAaBp06beZpc8AMyePdvbBx54YJ5vIQohaY7gMgG8APDdd98dtOMSBzz+RMVj5SiuaH7IIYd42y4UzeUq2LZzYseOHb3N4QC2FMYXX3zhbV5cGghDGDjcwI7va6+91tuPPvooSoWkMZUkb7355pve5gWb27ZtG7TjfuG53y42vd9++xV0Tjy32vuGf5P32msvb/fs2TP2eEnI0yOEEEKIVKCHHiGEEEKkgiqVtyZNmuRtlkDs+6y8xbIKS1XWTcrv69Kli7c5kwsIpY/27dt7u1GjRkG7zZs3e5tdq7ywJQDMmTMHpciIESOCbXY1sovbVi5mSYtlSyC+UjLLh0DohuWsDbaBMFNs8uTJ3rbyE0twLE9y1iAQ3pf2fhg5cmTs8UXFMHz4cG/zQsPWLf/+++97++9//7u37f3BbvoTTjgh2McZpHvssUc5z7j0sZlBTOvWrb1tx3ZcWIL9jWDphGUPm4HJkpat6szz0UcffeRt268835cSSdlbSfC14vndVtLn8IOxY8d6mxeGBsL5c/Dgwd7+61//GrQbNGiQt+2YZTm1TZs23rYZ39znScjTI4QQQohUoIceIYQQQqQCPfQIIYQQIhVU+irrrNdyXIXVDDkF1VZaZv2WbV6RFwg1YE7Ts/omV+zkSsA2zZljP1gbtumSHHtgqz/bmJaaRLdu3YLtd955x9ucZvj8888H7Y499lhv2/in9evXe7tz587etvFZHAPAOr/Vrps0aeJtvqfefvvtoB3fUxyDNHTo0KDdBRdc4O1atWoF+w4++GCIioXvByC8rzil+Oqrrw7a8djnGDIbH8Bjk+OFgLC8BMdo/exnPwva8X0qQjjmws6zXA2bU8ftCuk8Nvk9doV0jtm0MV4cf8lzwsCBA5O/QImQFMOTVDblww8/zHsMezyOs9mwYYO3jzzyyKDdSy+95O2DDjrI2/a3lUuD2N8I7luO2yk0hsciT48QQgghUoEeeoQQQgiRCipd3uKUYK6waReYZPeZlY/Ylc2uUZvazu14YUqb3sjp57xwJrtWAaBu3bp5P9emUrIUw+50oGbLW3axPt5mF+crr7wStOPU8d69ewf72M09ZcoUb3NVViDsI3Z/8+fabb4frHzK1Zr5/Gy17ptuusnb3P8iQ3lTYxkeS3Ye4PuF29lFbf/0pz95m6VxrtANAPXr1/c2p1fbY3JZi1tvvTVoxy77NJKUss7YlHWWKQq9T1gGsRJW0gKWPMdbKSXtJF23uDIktpI+/45zFeaTTjopaMcp7Dwfc3V1IPytTZKtWLYsL/L0CCGEECIV6KFHCCGEEKlADz1CCCGESAWVHtNjVyTPYVe/5RgOq/NzrA7rhFbv4zgOTnO2n8UxQ6wt2tQ+PsamTZu8bVOZOfbArgLOaXulBMe78PIMAHDllVd6m0sHAGFszfz5873NZQTs8bnd3nvvHbTjdMcxY8Z428Z2LFiwwNtciv7222+P/VyRgWMCbH/ymEmKHWA4BuzZZ58N9nFZBC5VP378+KAdj0G2bUwdp9Ta+C0e0zzn2PihtGNjLJmkeB+OfeT4nMaNGwftOCaSl0Ww/cVzv42/5HO070s7SWOR4yWXLFnibV5JHQjHPS8BY5eT4rhdnoN5uSB7PI6/BcLYsIpYNkSeHiGEEEKkAj30CCGEECIVVLq8xXIPuzht6vHKlSu9bdOX2YXOkpZNRWeXGbvWbGo7H4OlKVvhOc7tymmw9vx4JXkAOOecc1AqxMkXVhZkVzNLCEDoumQXta2g3K9fP2+zPGnlUi51sP/++3vblhXg93Gfx636btsByZVNS4mktPSka5DkRufxzpWt7SrNLFVx6riVlGvXru1tmw7LcL9byZMlLT4+V44HtlRZt/NNWkjqVx5nHA4AhCvc85xgxyZXXuYQBZ5/7THsOcWFNojk/uOVBnhs87wKhNd+1KhR3v73v/8d247Hm+0T/k2258fzvf2tLQ/y9AghhBAiFeihRwghhBCpoNLlLV7QjOUn6yZnl7qVmTizi91u9hjsMovLHADiK0BaGYzdq3wO1k3OEfDvvfde3mOXMpyRBYTSg8324eyaLl26eHvmzJlBu7/97W/e5myauXPnBu1YiuBF7Gyfs1TFsprNJGHsMWo65a2mzGPOZkJyX7OUabNreGHXV1991du2oivLmo888oi3J06cGLTjDB0efyxxAuH9ZuVKPl++T5944omg3dlnnw1ga5d/RZLrm/JUuC4LPMclZWXxd02qmPuPf/zD23bhSK72y78DVj7kuZX7KOl627mavwvb9hjlXbSymFRElXM7zyYtHhpHXBV8ILymZ511lrftOH/rrbe8zXMFV00Hwt/uRo0aBft4jud7qLyU1iwuhBBCCBGDHnqEEEIIkQr00COEEEKIVFDpMT1cTbfQiq02Vod1XtZkbUpjHFbvjFvR3Wr+vI/TW23MEeuaNuYkDdi0b463sPs4PZWrd9oyBQyXPTj++OODfZyqynqy1ck5NZPjUJLiGkoNmyrMfZM0Nnm82DHH5R+mTp3qbTsOuEo3n8fs2bODdqNHj/b2m2++6e0jjjgiaMfnyPEHNhaBK33beYBjS3hl7mOOOSZo16xZMwBbxy/URCoijufpp5/29rBhw7zdt2/foB3HN/K1trFyvM33pC1TwPeaPT++H/izcuUGcrRu3RqlAv9e2fmu0HntpZde8jbPkStWrAjacbmGBg0aeNv+FvJv9/r1671txx6fu62CX9HjTJ4eIYQQQqQCPfQIIYQQIhVUui9/zpw53k6q+MjYtFh2fya54dmlV6iUxm4364JjVyu792w7huUbAPjggw+8nVQ5tiZQaOojV9G0bmjeXrNmjbet9MJpjNz/dkFX7hc+hnXv8ueyezaJYqcPVzZJ1af5u1qZl9ODuTI5EMpYp512mrfvv//+oF2TJk28zYsIrlq1Kmg3bdo0b3fq1Mnb1gXOfchjzlZ4Zve7XdiwZcuW3ub7xabJ5r6zTZOuSOLutbgFPZPSnHmfbcfbhZbyePTRR4PtW265xdssN/NClEB4H/HYtBIkX1eeWwudt+25Jy0AXR3lraTvyd/Fhn0kVUfn31de4NdWUOaFQPka2jIkvGLCM888421bvoXHM49RK1XyvG2r9vN9+dxzz3n7/PPPR3mQp0cIIYQQqUAPPUIIIYRIBZUub7H7jCt22iwQzvixrkuO5ma3mHXXs0vPyiUMu3jjXLBAfIaIleaS3N7sXq3p8lahsCvT9jNfb3aNWjc8u0lZ7rQuee6jpKh/7ltbObbUYHcxV8K10hRvs2vbLg542GGHeZtd5QDwpz/9yds9evTw9u9///ugHWdlDRo0KPacOnfu7O0DDjjA21by5r7maq+2HY9NKwcsWLAg7zHs/JNz7VfGgqOFLnKbJImUV5ZlKf7hhx/29vDhw4N2Rx55pLeXLl3q7aSFPpPkU5YneR6wEnXSwrc8l/C+GTNmBO1OPvnk2HOsKuz9xnNckoTF2a8vvvhisG/kyJHe5qw3WxmZJV7uh4ULFwbtWGreuHGjt999992gHWeAcZiDvTdYdra/45yRywsVl3cBaHl6hBBCCJEK9NAjhBBCiFSghx4hhBBCpIJKj+lhfTIp5iZJ42PtLi41Ewj1e45RsBp33ErDViPk9/F7eDVaIDmWhKv/pgXWXm3FzrgK2FbXZo2e75uktGuOZbGfy/eh1YZrOkuWLAm2J0+e7G1OG7WrGbNOz/Eqb7zxRtCOU5G7dOkS7Hv88ce9zWnq9py44jFXN0+Kc+OKvrbMAMcOcCq6nVc4fsveY7wCO+/jWAlgS0xEUqmKbSV3TxYap1AoNv6QK2BzOjAQzmvcLxzDA4TlAzhWw8bKxVW8tvBczXOpnbfjKjcD8ddt8eLFsZ9b2cSVUbFxigzH2QwdOjTYN27cOG/buFK+jrZEAMNlQ/j8mjZtGrTjlRX4N+2oo44K2nGJGj62TW3n802qFM9jlL8vAPTu3RuFIE+PEEIIIVKBHnqEEEIIkQoqXd6KSym27k52X9tURXahcjobp+LZz+K0aVslOe491g3I6XJt2rTxNrvdgdA9aSvHchpooe64mk6StBhXYdZKTiy38L6k1GXGSpC2ImghJFW9rWq+//57Py7sfctpuSzV2HHFY3D8+PF53wOEY46lMyAcm5zmymMHCK8/jxd2XwOha5vLHdj5gqU6TnvnNFl7fLv4JH8v/ixbITY3z8TduxVBTp6xKfx33HGHt6dPnx77/r333tvbfH3tHMn9wIuxAsDRRx/tbe5LXsgXCO83lilYzgDC1GY+J3sfcvVfluPs9eb70u6Lk3DsfViVxM0fXDoBCGUsTgm3ad9c5sVKZDx/ch/Zsc19wWPASk4tWrTwNkuGvMgwEMrffO35XrDnZL8XXyd+Lhg7dmzQTvKWEEIIIQShhx4hhBBCpIKiy1tWfuCoes7aWL9+fdDOuqWZuOj2pKh3dtVZV2jc4nQ244e3u3Xr5m2bmcLR7NZ1a12XNZm47ANbdZld1DargjNgCs2o4vfYLIW4RRJt1gq34+NZ9z9LHtVJzrI45/x3smOOv3vc9QbCa9mhQwdvW8mJ72E7lviYhx9+uLetrMALFrL0aM+J5wG+d+x9xP3GLnCbYcWSi/0sHqssFdiqtY0bNwYQf69VJH/+85+DbZ4zr776am9bqYsr3/P5W9mHM93snMsSJ0sY9ntzdW2WQWzFah4/LGFZiYXnD74nbearnWeYuPvczsdVCd+LDz30kLc5BAIIz5nle/v7xGPAzndxi3tbKYn7lvvShgdwX7LUxfcdEEpQ3bt3z/t+IMzss6EH/J353K20Xijy9AghhBAiFeihRwghhBCpQA89QgghhEgFRRc4rRbI26zXWh2PtWZeYRsIK33a1DcmLgYjqSJzUpVPXqGaYxms5r9s2TJv25gCq3mWIvz9gVCHtSn8rClzHE9ShdWkVOG4PrfH435me968eUE7m8ZbXfnss8985WR7P/Lq5Kyd23HFcSyc8tysWbOg3RFHHOFtG7fBcQUcD2XhPowbf0A4fmzabBx8Dja1/fjjj/e2vU42PixHXDXvpO+3LXz11Vc+NdnOFxw/yJWxbXXtuFXG7fjjeBEbV8mlODh13sZT8XXg+T1pDPN9w/FdwNZxdTmS5k6OdQLCe4Xnn6Tfi2Jjr+9tt93mbb7vmzRpErTj+5KPkVTKxd7LhcYyxa1ibudc7kuOIczFu+XgODFOt+eYWPtZtqQMjz+OObL3Q1KVb0aeHiGEEEKkAj30CCGEECIVFF3esi4ndoUmLTDZqlUrb9uKqHHSRNICgNZFzfB5cEqndc+yy5i/h00DZRevdaFbN2wpYvuLSUr7Tuq/uPRU68blbe4jmzLN7bifufIsUHPkrZ133hnt27cHAAwYMCDYt99++3mbU4Vbt24dtOP7mPvJpr+y1GHla5ae2RVvxx/3NbvHbbu40gI2LT9uPNqFSQudL7idvfdy0kmcHLatbNq0Ca+//jqAMM0XCPuS5ckVK1YE7bjMAPeRrVjO19vKSnyN2bbfO2kRV4ZDGPgYLKsCQLt27bzNMqtdwJSlOhsewducEm/T7eOktIoiiiJ//9xyyy3BPttnOWy5Du6/JGmf29l+iKu0bD+Lt1kitO24/5IWxY1buNeuYrDXXnsV9Fnc53YOsOUd4pCnRwghhBCpQA89QgghhEgFRZe3kiKx2VVnJRF2wVmXpHW3xxHngrMudHaTsVvbus/YRRiXdWSxGR5JrsCaRpxUZRdJZLnBXqu4BUJttgHfA4X2P1/rQrMZrFxTU9hxxx3RvHlzAMAjjzwS7ONsOl4scuDAgUE7ll5ZymUbAA499FBvW/mW5WyWDe29wrICV9q1Ln++P7hvbP/xmOZsIDt3sJSSlA3G96mVQE455ZS8x64oGjZsiMsvvxwAMHHixGAfV+vl79KxY8egHZ8zz1W2Ai/PcZzZB2xdKTmH/d7cz3xOdp7l+2uPPfbwts0oi6sebOcOPj97f3FbPieuLA0UP5tr1apVuP322wFsHerB9zPvs9IU9wtfa/tbwv2SJBPzMezY5mMmLezMv5PcR/b3rtBq0jwHWKmSZU1uZ+9DHgO8MK9Fnh4hhBBCpAI99AghhBAiFeihRwghhBCpoNJXWWddj3VYm8rNmi9XcgTi44JsOl9cvIjVf1kbTErFi4sl4RRLAHj55Ze9bVNmi5XmWp2wsQDcD7biLsMasq30y8fgfkk6HmPLD8StPF5o1d+aRC7Wx9r9+/ePfQ+nlE6YMCHYx/EHy5cvD/ZxPA3r+VbD5ziF3/zmN95OipPh2A+bvsxxG1xV28YscOygjWHhWAKOt7DHyMUVFDM+L3df9+jRI3idt/letdW1uVotl2GwsZN8DL43gHBc8Piz8yxvc0yHjdXZf//9vc3xF4899ljQ7p577vE2/w7Yz+V7xfYFz/FcJXjt2rVBu6RSJhXBzjvvjDZt2gDYulI9l4zge8/+PvH35nnMzpGM/V7cz/w++5vGn8XjNymOi+8TG4/E83NStXY+PzsH873M97mtyGzjteKQp0cIIYQQqUAPPUIIIYRIBUWXt6zUEZfG2qJFi6Ad77MLtXG1Zk5tt5VT2cXHrmyuVgqEabfsJrXyGJ8Tu9bOOeecoB3LW0luwVLFukw5Zd2mGrOrNa7SNhDKguyCta5gvr5J9wb3M6dBWskjrbBka+Xb6g7LZTWVuMrGfL/z/NayZcugHW/bqs5x2HR2HmdxcjAQLipZHk488cRgm2Uxlqas7MHnYecVvk5JJUWsBFfR1K1bF2eccQaAUKoDwirxPAdNnz49aMfzJ//uWImIj8+LYwOhdMnXkT8XCKVrtu3vM0vevCCx7SP+LgcddJC3Z8+eHbTjisz2d5dLoHAYjF0J4eSTT/b2XXfdhTjk6RFCCCFEKtBDjxBCCCFSgR56hBBCCJEKih7TY8v6s4bIS1T07t07aMcan00ZjUsxTkp35X12Vd847TopNoU145/85Cexn5u0Ym5a4Pgcq/nyPtblk1ZrZmwJAH4fx+3YWCo+D+7XuDIHQlQmhZZiqEhs2ZDKomHDhsF2bpmPUmC77bbz88tJJ51U0Hv69u1bzFOqEC677LKqPoVyI0+PEEIIIVKBHnqEEEIIkQqKLm9ZyYklDJaVOnfuHLSbPHmyt2fMmBHs4xRarvJo05dZFkuSreJWWbcVL7kdp3fuueeeQTuuwmxlmjTIW0lVuG3l1LgKoFYWjHP3276MWyXYrgbO58jHsBVrk0hKnRdCCFH9kKdHCCGEEKlADz1CCCGESAVFl7es/GAr4+ZYsGBBsP3oo496mys+AmElS5aL7LG5GjTLD1zRGQglKJatbJYXS3WHH354nm+RgWUxlt8AYP78+bHvKxU6dOgQbHP1TSvvxUlftsIqS0lx0hQQZmlt3Lgx9hy5yjdXMs0tDiiEEKL0kKdHCCGEEKlADz1CCCGESAV66BFCCCFEKih6TE+nTp2CbV5pdd68eVtOxMRwcPzMwIEDi3R2xYFXebap1jY1vxTp0aNH7PaiRYuCfbxqMK9wbtPeeZtLEdjUdl7xmeNz7Iq83K5169Z5vsUPozR1IYSoWcjTI4QQQohUoIceIYQQQqQCx6nAP9jYuY8BLCve6Yg8NI+iqMEPNysb6ssqQ/1ZOqgvS4sK70/1ZZUR25dleugRQgghhKipSN4SQgghRCrQQ48QQgghUkGNf+hxzu3lnBvsnFvknJvmnHvFObdPGY9Rxzl3ebHOMU045/Zwzs3M/lvjnFtJ2zsmvK+Fc25uzL7bnHNHx+w7zzm3t3ntLOfcH5xzvZxz3bftG4l8ZK/vPOfc7GzfHlKBx+7lnHu5oo4nMjjnTnHORc65dgW2X+qcq5/n9c/ztU84TpnaJxxnq7GeZtSf5aNGP/S4TKGUFwGMjaKodRRFXQD8HsCeZTxUHQB66KkAoihaH0VRpyiKOgEYBODe3HYURf/9offHHPPmKIpG29edc9sDOA+AHTjHAxgFoBcAPfRUMM65wwD0AXBQFEUHADgawIqqPasMzrmi1x6rwfwCwPjs/zWR87D1WE8z6s9yUKMfegAcBeCbKIoG5V6IomgWgPHOuTudc3Odc3Occ2cCgHNuN+fcGOfc9OzrfbNv+xOA1tm/WO+s/K+RLpxz+zvnpmSv92znXNvsru2dc//MehD+45zbJdv+Mefc6Vl7qXPuf51z05EZ7F0BPJU91i7ZB+FOAD4BcCmA32T3HZH1Jr2e/cwxzrlmdPxBzrmpzrkPnHN9Kvua1DAaAVgXRdHXABBF0booilZl+2YAja92AOCcq+WceyTb5zNy4y7bH29l20/P55VzznXLvqe1c66Lc25c1qP7qnOuUbbNWOfcfc65qQCuqbzLUHNwzu0GoAeACwGcRa/3yl6/Ic6595xzT2XHEL93F+fcSOfcxXmOe61z7p3smBqQ8Pn3Zsf1GOdcg+xrnZxzb2ff+6Jzrm7c69nxH4z1CrkwNRT15zYQRVGN/QfgamQ8Cfb10wC8BmB7ZLw+y5GZqH8EYPdsm/oAFgJwAFoAmFvV36fU/gG4FcD/5Hn9bwDOzto7Atgl2wffAuiUff05AP2z9mMATs/aSwFcR8caC6ArbR8E4Il8nw9gBIBfZu0LAAyj449C5o+AtgA+BLBzVV+/6voPwG4AZgL4AMDfARxJfXNV1r4cwENZeyD1ZZ3s+2oB2DV3nbPXfWrW7gXgZWS8dNMANAOwA4CJABpk25wJ4BG6B/5e1delOv8DcDaAh7P2RABd6FpvBNAke/9PAtCD+rMFgNEAzqVjfZ79/1gAD2bn0O2yfdYzz2dHNN5vBnB/1p5N985tAO77gdeDsZ7mf+rP8v+r6Z6eOHoAeCaKou+iKPoIwDgA3ZDpzIHOudnIdHxjlF0KE9vOJAA3OueuR6aewubs60uiKJqZtachM0QPI/8AACAASURBVEDz8WzCsY8DMDJm32EAns7a/0LmPsnxXBRF30dRtADAYgAF6eRpJIqizwF0AfArAB8DeNY5d15299Ds/9x/xwK4wTk3E5mJbmdseZD5p3NuDoDnAexHH9MemQn4pCiKlgPYF0AHAK9lj3MTMhN7jqR7QmS8ooOz9mCEksiUKIo+jKLoe2QeZlvQvpcAPBpF0RN5jnls9t8MANORGTNt87T7Hlv650kAPZxztQHUiaJoXPb1xwH0jHu94G+ZHtSf5aSm69/zAJxehvZnA2iAzFPxN865pchMwKKIOOdOBXBLdvOiKIqeds5NBnAigFecc5cg86DxNb3tO2Q8QPn4IuHjjkXG01dWbMEqFbBKIIqi75B5gBmbfWj5ZXZXrg+/w5b5xQE4LYqi9/kYzrlbAXwE4EBk/rL8inavRmZsdgawKnuMeVEUHRZzSkn3RKpxztUD0BtAR+dchIwHPHLOXZttYscd/y5MAHCcc+7pKPvnOR8awB1RFP2jjKeksbUNqD+3jZru6XkdwE7OuV/lXnDOHQDgUwBnOue2z+qNPQFMAVAbwNrsA89RAJpn3/YZgB9DFIUoil6MtgQzT3XOtQKwOIqivyLzl8cB23B433fZvyp+FEXRersvy0Rs0b/PBvAW7fu5c24751xrAK0ABD/QYgvOuX3dljgsIBNDlVR19lUAV+ViC5xzuVV3awNYnf2L9BxkJu8cnyLzUHyHc64XMv3RwGWCqOGc28E5t39FfJ8UcDqAf0VR1DyKohZRFDUFsATAEQW892YAGwD8X559rwK4IBtfAudcY+dcwzzttsOWP077ARgfRdFGABucc7lzOAfAuLjXs7bm6Qzqz22gRj/0ZJ9UTwVwtMukrM8DcAcyEsZsALOQeTC6LoqiNQCeAtA1+5fpuQDeyx5nPYAJLhP4rEDm4nMGgLlZmaIDgHyu1kJ5DMCg7LFORka2zDECwKnZQLkjAFwF4PysvHkOwqDX5cg8GI8EcGkURex1ECG7AXjcOfdu9lruh0z8VBx/REbKmp0do3/Mvv53AL90zs1CxpUeeGuy0nQfZCbozshMtP+bbT8TyswrlF8gk+XKvIDCs36uAbCLc+7P/GIURf9BZq6dlJ1ThyD/j9gXAA52mZIUvZGJ6wAy3sE7s/dQpwJefwzZsZ7yQGb15zagZShEyeCcewiZ4Nm3y/i+xwC8HEXRkKKcmBBCiGpBTY/pEcITRdFFVX0OQgghqi/y9AghhBAiFdTomB4hhBBCiELRQ48QQgghUoEeeoQQQgiRCvTQI4QQQohUUKbsrfr160ctWrQo0qmIfCxduhTr1q1zP9yybFT3vlywYEGwvf32W+rWcfD97rvvHrRr0KBBcU9sG5k2bdq6KIoq/CSrY39+9tln3l6+fHmwj9dA5P60iRVt226pgbjTTjtV9CluE6U+Nrn/Nm/e7O1ddglLqsT1y3fffRdsf/vtt97esGGDt3/847AUTO3atb293XaV93d5McZmdelLZtmy+DqiderU8Tb316ZNm4J233zzjbf32WefCjy7iiGpL8v00NOiRQtMnTq1Ys5KFETXrl2LctzK7MukDEGzALDnhBNOCLZ32203b/Ng/OlPfxq0u+SSS/Ie7/vvvw+2K3MyZZxzSZWLy00x+7M8/QcAb7zxhrevuOKKYB//UH711ZY6kPazhg4d6m1+ALJw//I5JZ3ftlIKYzOJMWPGeHv+/Pnebt++fdCuTZs2ed//6aefxm4PGbKlJFavXr2Cdscdd5y3a9WqVfgJbyPFGJuVOS4Lvdd5jrTzYt++fb39ySefeHvUqFFBu48//tjbr732WkGfy+dbzHGZPX5sX6pOj6hwyjsYr732Wm+PHz8+2McTKw/GuXPnBu06duzo7e7dtxTsraqHnFKgLBPUCy+84O3TT9+yLN6BBx4YtOM+5B+2nXcOl8Lr0KGDt4cPH+5t+7Ab17/lvRdLlS++CJcou+GGG7z93nvvBfs+//xzb7O3wnoKeGzyj+GiRYuCdvzHCnPffffFbvMfOyNHxq0jnE6S7uV169YF27/4xZZizRMmTPD2jjvuGLR79tkt6/by8e3DEXve69at6+1BgwYF7c4888yCzpe9gnzsYqBfAiGEEEKkAj30CCGEECIV6KFHCCGEEKlAMT2iwknSbq+55ppge8qUKd5mHdoGS65YscLbnGFgMz8uvfRSbzdp0sTbV199ddCOgyVZr05z7E+hgYYPPPCAt59//vlg3wcffODtdu3aefvYY48N2nG8FX+uPR4HOV955ZXebtq0adDu1FNP9fZVV13lbfs90t7XHMMDhMHFHD9l4fge7hMgk8WWg8emPd7ChQvzHrtTp06xn8U2x5sAYbxIKVFoHNrEiROD7euuu87bM2fODPZxlitnW61duzb2PJKyJfm+2Wuvvbx94YUXBu3+53/+x9u33HKLty+6KFwmsdhxPEz6Rr0QQgghUokeeoQQQgiRCiRviaLDcginHQOhC5xdnJzSDADNmjXz9pdffultK1Fw+uyaNWu8bd36LG+lUebIR5y89c9//jNo9/vf/97bVmZiuZHTo++8886g3a9//WtvL1682Nv2/mjcuLG3WTpZvXp10O4Pf/iDt1etWuXtO+64I2iXxr7m8g82LZ1lClsvh1PMOXX8Rz/6UWw7lr6sDMbtuC95nNrjczubDs11klq3bo1SIUlafvTRR709YMCAYB9Lt3vuuWfsvq+//trbXHQSAJo3b+5tTiO3pUFY0uJjN2zYMGjH34XnjQcffDBox2EOloqu75O+GUAIIYQQqUQPPUIIIYRIBZK3RNEZN26ct23FXV7DZY899vC2rRTKmQRcAdS2Y5csrxHEkgcATJs2zdtdunRJ/gIpIU76ee6554Jtdm3brAt2nfM+dpsDoWTG7nG7js9///tfb3Pf2n5v1KiRt/l+E0CPHj28bZeMGDt2rLdtFV+WlljeSiIpy4vHPktddrmK+vXre5uXcLAy2K233urtf/3rXwWdX03n//2//+ftXXfdNdjH48OOZZaIuF/sWoXc5xxiYMdvvXr18n6unQ/4c/mzOBsXAN555x1vd+vWDcVEnh4hhBBCpAI99AghhBAiFeihRwghhBCpQDE9VQivVjxs2LBgH1efrels2rTJ21bn37hxo7c5psfGEPDqzaxl24rMHB/Atv1cTpFUTM/WrF+/3tu2ki7HZtjVlzn2ivV9W2WWU1s5bdbGIvB9wLEDbNv3rVy50tsbNmwI2nE8WFrga3XFFVcE+7iUg62gzNXNOc7Gxv4wNp2d4XgRHo8cIwaEFZ65WrON/bnrrrtiP6uU4Li2WrVqedumb3M/c7kIIL7isR1H3LeLFi3ytu0jLhti5wCGYzY5xtKe+4svvuhtG9NTEWnqjDw9QgghhEgFeugRQgghRCpIjbzFqbR2IbX//Oc/3j777LO9/cILLwTtevbsuc3n8cQTT3h7xowZ3uYKtQDw1ltvAQA+++yzbf7MqsamizOc4spuTF4sFAB22GEHb7Or1qYus5TGLljLhAkTvH3ZZZfFtksrvGAhjx0g7AsrG7KEwX1jj8FyFKeps+Rrj8H9ad3ycSm5s2fPDtodeeSRSBs8XpIWFWU5CwjlJJa3khai5FR0K3Xx+3jc87HtvvPOO8/bNt3eVh0uVXhhUZ7feKwB4fho0aJFsI/HH8u/toQIp6lzuIFtx5IbS1j2c/l4SSUnRo0a5e2BAweimMjTI4QQQohUoIceIYQQQqSC1MhbSRHma9eu9Ta7ZC+++OKgHS9qx9Hx7Oqz2KwVXqCRP+u1114L2p1yyikAwoj3moKVlZIkOnZl8zW1C+FZOSPHhx9+GGxbuSUHSzIAsGDBgthzEqH0au/hpEU7uT9ZzrDv4f7kbB12lQOhNMN9a/uTpVG233777aBdGuWtQrEyE0ucnL1jMyvjqjDz60AokbB0ZmUafp/6CxgxYoS3k35reB9XKAfCfuFqyElj21Z8Znj8seTGlZotLKvZ+WDJkiWx76to5OkRQgghRCrQQ48QQgghUoEeeoQQQgiRClIT05MUGzN+/Hhvs75s44AmTZrkbY4bsKmZHJdgK8Duv//+3l6+fLm3bcXLnL6eVOG0umLjZThdMqm6JscK2O/NmnRcdVEgPm7AxoDYbREyffp0b9s+S4qP45Rlfl9Sn3E7G2PAsT/cZzb2h1Ng+d6xMT1px8bGcb/w3ASEMT08lmzKOlfxtfMYw/3Ctq20bNOe046tKh4Hjw8uyQGEsTZcDiSpQj7HYtrK9xw/xP01a9as2OPxKut2nO++++7e5jR3e+4VgTw9QgghhEgFeugRQgghRCqoedpJGWBXbpJ7nWUrdhHatGleJJHb2TRCdhnadExOseYUvhNOOCH2/Goa/L2AsB+s+5vT1NltzgvrAWFaJEuV1l3P7lSu2MrvB7ZOpxUhLL1a2DW9Zs2aYB9f8yQpk2Uwlsus7MifxX1tXf58X/FYt/di2uHrDoTXqn379rFtWcKyx4i79j/02TnsWMyV67AkSXOlDM+RPD7s9eR2NqyCrzHLVlx1GQjLsnz00UfetnIyj0uWpmzldZ4f+B5iqQsIy5zMmzcv2HfEEUegIpGnRwghhBCpQA89QgghhEgFeugRQgghRCoo6ZieuJgCm1K9fv16bzdu3NjbdvkELp3NGqdN4eVYIBvvE7fae1Jp/5oGa7dAqOuef/75wb6RI0d6m68bx+YA4XXjeB+7+u/cuXO9feqpp3qbVyrOd44ihOMlbIoyxxKcdtppwb4333zT2zxG7DF4bO67777efv/994N2nNrMY+nQQw8N2i1atMjbPJaSVgRPI0klMGwMI7flNOekuCCOK7GfxXElvI/HrD0ek/S5pQwv0ZAUJ8dxMfZ3p3nz5t7m2KiPP/449hgcF2RjdTitnPvczqtchoRjXe1vJvftuHHjgn2K6RFCCCGEKAd66BFCCCFEKihpeStOMho2bFiwzWmA7IKzVSN5O25VZyB03VqXMbsJOX3drkyetMJtdSepguh+++0XbPPq8uzytO5U7kt23doVn/kYLFXWrl07aMcre7N71qbKpxV2X1sJ8b333vP2c889F+zjdFOWNW0/Mey+58rKFpbLrrjiimDfGWec4W2WYmy137STJBE1bdo09n282jmnMgPh6uxJkhO/j+dIe3/FVWSuidXpKwL+neBrYKWpq6++2tuPP/54sI/bcnVlK//yGLPHZ7iP+PfPjjcuP8DhIlOnTg3a8W/wu+++G/u5FYE8PUIIIYRIBXroEUIIIUQqqBH+QiszxUWw24jwOHnr6aefDrbZPctSh/2cpMj5OGwl4F69enmbpTTrSuRo+5rG6tWrY/fZhV85c46zCmzfxckj3F+WtWvXetvKajNmzPD2qlWrvN22bdvY45U6fD+yvGjlB3ZF24UIuT/5fVau5GNwvyfJL9yuW7duQbu4sWkXVBTx2Gu/cOFCb7/wwgveZrkliSQ5iivV26rekiRD+DeEVwWwoRP33nuvt8eOHRvsY8mIKygnjTcevzyX2vPg0AFe3QAADjvsMG/zHPDqq68G7biS++TJk1FM5OkRQgghRCrQQ48QQgghUoEeeoQQQgiRCmpETE+hsTRJVY0fffRRb3P1ViDUHTllz8YD8PHtir8M66L2GMccc4y3WcecOXNm0K4mx/RY/TdpH6dGszZs4wH4mnI/2KrZHGPC17Rz585BOz6Grf6cVnhldR4Hti9Yz7dwvEe7du28bePteJttO9Y5hZ1jG3jsAOGqzXGreQNh7Jxd6TkNJMXZ2BW3uf949e2k8gNxVZft+7iP7BzJ8T5pxMYp8m8Njwc7pph99tkn2OZSHCtXrow9Bo8/7hdbSoKPsXjxYm9zvBAAdOrUydv8m3bttdcG7fj4HGNZDOTpEUIIIUQq0EOPEEIIIVJBjZC3kuDqvNYFx/tuueUWb7PLDQhdeiyXWFc7uxk5jd5KXezWtemXs2bNynt8m0bft29f1FQ4bRkA6tWrF9uWXbksN1i3K8stLHXZPuJ9LFvx4nlAKLMpRTYDpw7z/c3XHggrHo8ePTrYx+/j8ZgkB3NV2KQUWh7PgwcPDtqx/MLytb0/WMJLo7yVxJgxY4Jtlje41ESPHj1ij8HvSZIZeY60Mtizzz7r7Q4dOng7LQuMzp8/P9hmOZnnxaRwDiv/vvXWW97mvrTH4LHIY8fOAdyO51lbmoKrrSctHMrHs/JmRUvS8vQIIYQQIhXooUcIIYQQqaBGylvsNk1aoLB3797eZpe8XXyS3XNxlWLtdtzio8DWi7gxHPXO2RG2InOSHFDdsd+FI/rtgp7cl+zmTspM4Awt2/8sW7EckpRxJHkrAy8ImeRG5+v/yiuvBPtYRuQxYquqFwqfB7vlJ06cGLRr06aNt212JpOUWZh2hgwZEmzzeOSsrLlz5wbteG611ZXj4EyupAVt04j9/lZaypEk9YwfPz7Y5jACno/t7wwvfM1ztc0o22uvvbzN8yf/fgLAlClTvH3uuefGni+Pc3tOnIXL2c/lRZ4eIYQQQqQCPfQIIYQQIhXooUcIIYQQqaDSY3pY22cdz2r+vG1TFeOqip500knBNsdx7L///t6eOnVq0I4rvXKMiE2/4/gTjuNhHRQI4wZsZUxe7Zt1crYB4J133gGQvIp4dcWu/svpk8uWLQv2cbwIX1Or63I8B+vG9tpzfABXe7ZxO3xOHPuTZvheW7Fihbc59gwIx4u9/hzTw31oY7TiYuLsWOcxyKUPZsyYEbRr1aqVt/mesp9ryymkjaS0b1sVnuOkeH6yFZl5zPEcadvZOa4QuMTHgQceGOzj+6uU0tlthWr+njz32fgZhmNHAWDXXXfN285eNx4vPI/buFXuZ36PPafp06fn/VwbJ5g0B69bty52X3mQp0cIIYQQqUAPPUIIIYRIBWWWt3Ju6SQ5im0rRSW5sguFUyYvvvhib9sF84YOHeptrvI5bNiwoB2nsM+ZM8fbNvWcXfcsl9hqvy1atPC2dfGye51dtwsXLgza5apyWqmoJmAlD05jtemuLDOxG9dWc+W+YNeodYuyq52lygkTJgTtWDbhPk8zcfKDHcO8IKAtGRBXhTlp0eAkmYJd55xCa2UqXuiU7wGbvm4XqE0DfK/bOe2ll17yNqchA+H9wNfUyg08x/G8mCRv2YrBDFfkHTFihLetvFVKkhZj53y+Z7n/GjVqFHsMK+fzbxJjf8d5rCeVTYlbMNZWU549e3bez7JhH3y+dgFoltMrAnl6hBBCCJEK9NAjhBBCiFRQ7uytpMXOkmAXF7vtbKXUSZMmedvKUdOmTfN2//79vX377bcH7e677z5v/+Uvf/G2dcH169fP2x07dvT2Aw88ELRbsGCBt9nde8011wTtWM5hCQsI3fCcHWHdveWtYFsdsPfGoYce6m2uBgrEZ/hY1yrLI3EZXxZ2wfI5AMDrr7/uba4wnGa4kjZnYSQt5GsrbMdh72fuN87Ms3Il9zXfE/acNmzYkHeflea4XVpIqhD/5JNPepvnIyCU3HmfrRgcV0G5fv36scfj6t+2IjN/Fs+ldk4oVXkrKVuJ79/u3bsH+55//nlv26xFvlYsd9pxGVch37aLWyTYzgc8fjnEoHnz5kE7/s7296Oix6w8PUIIIYRIBXroEUIIIUQq0EOPEEIIIVJBmWN64mIoxo0b523WeJcuXRq040qRHNNjdTyOcTn11FODfYMHD/Y2xx7cfPPNQbu//e1v3j744IPzfi4ADBo0yNusQ9uqr1yxluMQOHYICFM1eQViIFwZt1mzZt7+z3/+E7TLaZ5Jenx1JZdun4/f/e53wTZrz3xvJcU0sYZsK4Cy7s+pjueff37Q7tprr409flrhtFEefzaFluMDbMwMxwtwFVhb3TxuHrErSvN9kFRpme8Dfo+NlVu9enXez00Lth84zsaWieB5fPTo0d626c8837FtSyDw8XmetZ/Lvxm8WvgzzzwTtON4zlLCpmzzfc/3NpdaAcL0fq5eDoTzIo9ZO47i5mC7ovv777/vbY7JsuOay7k89NBD3rYVouN+B4Ct0++3FXl6hBBCCJEK9NAjhBBCiFRQJnlr8+bNvqLub3/722Afu+DYRW3TFjkl/KCDDsr7HgBo2rSpt9mVBgA33nijtx9//HFv23Q5dq1xZUgLu+fYpceSGBC6+Njlbxd349T2xYsXB/sOOeSQvJ9r3c5t27YFUDPlrSSsvBAnc1i3a9w+2+dx6ZjLly8P2tn7UoRpxCwXWQkrLl0ViO8328/cN0lVxwstjcHnyJ9lU9vtWE0bnAJut5MqLb/99tve5rIb9hhc1dkunMnb3F+2Sjsfn6U0W/6jVOUtW76FxyKHVdixxot72mPwOE2qgB5X+d6OX5bWuF/s7xWXweD5xZYu4XFq5wP727ityNMjhBBCiFSghx4hhBBCpIIyyVtff/31Vgv45eAIa3ZjWllh5syZ3n7uuediP4vlLuveSqoAybDrjt2udiE8duN98skn3p4yZUrQjl2yOfkp3/k1btw477GB0MXHlSbnzZsXe06lhP1eSRWVGb7G7O611Vz5eIUeO8ndmyZ4DLPNWYtAmHlhrzG7qbnPrMwU159WLuP+sJl6cfC5W2kurnpwWrCZMB06dPC2vfY8tvh9NisrTiLr2rVr0G7UqFHe5gytU045JfYc+bfkhhtuQBqwMiOPnaTfPpYJn3766WAfy7r8m2wX7uXFl/k8bKYY/8Z16tTJ27bPeXu//fbz9qWXXhq04znYflZFr05Qmr+sQgghhBAGPfQIIYQQIhXooUcIIYQQqaBMMT116tRB3759AcD/n4Mrdo4ZM8bbNs2QNXaOabGrK3OsgE19jMPGd3Dano1LYHhVX04pP+CAA4J2XL10wIABed8PhFqo1dAbNmzoba6aybFEwJZUd3tdajo2HZH1WtZ1bewPv49jsmzMRtyxk/q/0NifUufhhx/29rJly7xty0lcccUV3v7yyy+DfTwGuT+TKr8mlSdgktpxFdthw4Z5m+MIgK3jBdIGr6oObD1nxlGnTh1v25R1jungmBMbm9KrVy9v87i1MZa8j48xadKkoF2fPn0KOfUah53zOYaO46mS4j779etX8SdWgXDqPRDOMTb+z8aabSvy9AghhBAiFeihRwghhBCpoMwLjsZx9NFH57WTYFedrSDJspB1oRcKp6lbCWpbWbJkibeti5iriFrX7d577+1trgrMrwNbUuLt+0sZlqOs5LStMl+hC5ha12qa4DRUti0sW9nrGrdwIL8nCVsyIK6cgD0ey589e/Ys6LPSCJcMAcJ50coILC3xHMdp7kAoW3H6ui0P0K5dO2+zXGY/l+d+nkv/7//+L2hXqvKWla3iQgBsVWPGStI8FnkcVYS0H3d+QDgf8NyaW1A7H/b729CPbUWeHiGEEEKkAj30CCGEECIV6KFHCCGEEKmgwmJ6ygPHq9S02BVe3V0UTqGrm1tdl7XiTZs2edtq0nExJRVdyrwUKXQ5jtWrV3t7jz32CPbxdeY+TIrVSYrlYpL2tWzZsqDjcWxYWuK3eMkPG3/I6eEcPwMATZo08TanqduV2jl+hPuZU9ktHNNjz4k/l+cLmwLPZRWSYkRqOjyOuL/q1q0b+x573yeV9mAKnSfjlvuxn2Nji3LYGFsei4Ueo7zI0yOEEEKIVKCHHiGEEEKkgiqVt0T6sOmp7BrlVbRt9V128bJEYV2h3I4llIp2kZYihaavspRkK2zHpcYmyWVJlZa5f5Mqs3K/J32PpPMoVS666CJv80rcQFjt3spHvBI6y0x2lfVFixbl3ccSFhDKYlzB/+233048/xw2BOLXv/61t1988cWCjlETsPcv9xHbSRWZbYmPQuWtiq5OHze2ea4HwvO11ZptOZttRZ4eIYQQQqQCPfQIIYQQIhVI3hIVjpWS2JV5+umnB/uGDh3qbXavWxmCt61rlOFMldatW3s7qUp4oa7fUicuc8NWWU1aKHinnXbyNmfZWWmCP4td4NZlHyeXWVhm4/OzGS5pzOLjCsc2o6pLly7eHjduXLAvbvFQKzMOGTLE2yxh2XYsR/F7evToEbRjiey4447z9k033RS0e+yxx1CK5Krx54irRs8rDliqi4wbJ8HVqlUr2ObsOyuxJVWeLtc5VejRhBBCCCGqKXroEUIIIUQq0EOPEEIIIVKBghlEhZMUI3P88ccH22+99Za3x48f7+1Vq1YF7aZOnertVq1aefvQQw8N2nG8T9++fWPPg+MNFNOTIS5d1cYH9O/f39uLFy8O9nH8z2effeZtq+1zHA/3RXlLGnAV37JUqk0DI0aMiN3HVY0vu+yyYB9X23744Ye9bePj4qohX3XVVcH2eeed520et2eeeWbs+a1cuTJ234UXXhi7ryYzaNCgbT5Gdak2Hhd/aeO4Zs6cWRmnA0CeHiGEEEKkBD30CCGEECIVuLKkcDrnPgaw7AcbioqkeRRFDSr6oOrLKkP9WTqoL0uLCu9P9WWVEduXZXroEUIIIYSoqUjeEkIIIUQq0EOPEEIIIVJBtXjocc6d4pyLnHPtCmy/1DlXP8/rn+drn3CcMrVPOM55zrm9K+JYacY5t5dzbrBzbpFzbppz7hXn3D5lPEYd59zlxTpHUTjqz5qNc+4759xM59w859ws59zvnHPV4jdDlJ+0j8vqcgP/AsD47P81kfMA6KFnG3CZAiovAhgbRVHrKIq6APg9gD3LeKg6AGrkYCwl1J8lweYoijpFUbQ/gGMAHA/gFtvIOadCVzUEjctq8NDjnNsNQA8AFwI4i17v5Zwb65wb4px7zzn3lDOVxZxzuzjnRjrnLs5z3Gudc+8452Y75wYkfP692b9kxjjnGmRf6+Scezv73hedaCzMuQAAIABJREFUc3XjXnfOnQ6gK4Cnsn8V7VIhFyZ9HAXgmyiKfGWuKIpmARjvnLvTOTfXOTfHOXcmkLlvsn02Pft6rhLhnwC0zvbFnZX/NUQW9WcJEUXRWgC/AnCly3Cec264c+51AGOcc7Wcc48456Y452bk+s85t3/2tZnZebNttu2/s96jubl7QFQKGpdRFFXpPwBnA3g4a08E0CVr9wKwEUATZB7OJgHokd23FEALAKMBnEvH+jz7/7EAHgTgsu99GUDPPJ8dATg7a98M4P6sPRvAkVn7NgD3/cDrYwF0reprWZP/AbgawL15Xj8NwGsAtkfmr5HlABohU01892yb+gAWZvu7BYC5Vf190v5P/Vnz/+XmU/Pap9l+Ow/AhwDqZV8fCKB/1q4D4AMAtQD8jebYHQHskr0H/knHrF3V3zUt/zQuo6r39CAjaQ3O2oMRSlxToij6MIqi7wHMROZC53gJwKNRFD2R55jHZv/NADAdQDsAbfO0+x7As1n7SQA9nHO1AdSJomhc9vXHAfSMe73gbynKSw8Az0RR9F0URR8BGAegGzIDb6BzbjYyD7+NUXYXrah81J+lw2tRFH2StY8FcINzbiYyfwTuDKAZMn+s3uicux6Z2imbAcwBcIxz7n+dc0dEUbSxCs5dhKRmXFapFuucqwegN4COzrkImafMyDl3bbbJ19T8O4TnOwHAcc65p6PsYygfGsAdURT9o4ynpKJFVcc8AKeXof3ZABog4xn8xjm3FJmJVlQP1J8lhnOuFTLz8NrsS1/wbgCnRVH0vnnbfOfcZAAnAnjFOXdJFEWvO+cOAnACgNudc2OiKLqt2OcvAGhcVrmn53QA/4qiqHkURS2iKGoKYAmAIwp4780ANgD4vzz7XgVwQTZeCM65xs65hnnabYctN0A/AOOzf3VscM7lzuEcAOPiXs/anwH4cQHnLOJ5HcBOzrlf5V5wzh2AjDv9TOfc9tmYq54ApgCoDWBtdiAeBSC36qH6onqg/iwhsn01CJkQgHx/HL4K4Kpc3KVzrnP2/1YAFkdR9FdkvPMHuEym65dRFD0J4E4AB1XGdxAANC6rfJX1XwD4X/PaC9nXn926+VZcA+AR59yfoyi6LvdiFEX/cc61BzApOwY/B9AfW/5CyfEFgIOdczdl9+UC6n4JYJBzblcAiwGc/wOvP5Z9fTOAw7IuXFEGoiiKnHOnArgv6wr/CpnYrV8D2A3ALGQ8cddFUbTGOfcUgBHOuTkApgJ4L3uc9c65Cc65uQBGRlF0bZ6PE0VG/VkS7JKVq3YA8C2AfwG4J6btHwHcB2C2y6S1LwHQB8AZAM5xzn0DYA0ysT/dANzpnPsewDcALst/SFHRaFxqGQohhBBCpISqlreEEEIIISoFPfQIIYQQIhXooUcIIYQQqUAPPUIIIYRIBXroEUIIIUQq0EOPEEIIIVJBmer01K9fP2rRokWRTqVwNm7cUrW8du3a23y87777ztubNm3ydt26dbf52NvK0qVLsW7dOvfDLctGdenLzz//3NubN28pb+TCtWWD7Z133lIQ9Pvvvw/affvtt97ebbfdvM19DAA77LCDt7fffvuynna5mTZt2rooihpU9HErsz+5zMV///vf2H18ze31jyuVYV+390EO22c77rhj3vdwP/8Q/Nlxn8uU+thMG8UYm9WlLz/66CNv85zJrwNArVq1vF2nTh1vf/3110G77bbb4i/hOfdHPwofKRo2bBi7r5gk9WWZzqJFixaYOnVqxZzVNvDKK694+4QTTtjm4/FD1OjRo7192mmnFfT+Qifq8tC1a9cKOxZTXfpy/Pjx3p47d663d9ppp6Ad/8jts88+3v7yyy+Ddhs2bPB2jx498r4OAHvttZe3eXAXG+fcsmIctyL6s9AffX7QWbYs/Do8oX7yySfe5j8mAOCbb77Je2z7EMsTJZ8TT84A0LJlS2/zgw738w+RNHnno9THZtooxtisLn157733evuzzz7z9j33hLUmDz30UG//7Gc/8/aiRYuCdvxHBs+t9evXD9pdfvnl3uYHoGKT1JeSt4QQQgiRCqp6GQqPfZK8++67vT1t2rRg35IlS7zNT5bW5X3ggQd6m/+CnD9/ftBu3bp13ua/du+///6gXfv27b19xx13eNtKbPxZ7AYUW/PEE094e8yYMd7maw2E90eHDh28/eMfh8u/8F/oe+65ZTFgK8OwF6J79+5lPe2SoFAP5ahRo4Lt5cuX57WB0PPD0qX14PBYZc+M9QDFSVV2XPF9cNBBW5Zyatu2bdCuVatW3rayA987SZXqK9KTW9l88cUXwfa///1vby9evDjYN2HCBG937tzZ23ZsLl261Nvcf926dQvarVq1ytt8TzVoEKoQ3H88hk888cSgnebWeKx36cYbb/R2v3798toA8MADD3j7pptu8vYll1wStOPf1mOOOcbbf/nLX4J2PO8OHDiwoHMvNrprhBBCCJEK9NAjhBBCiFSghx4hhBBCpIIqjemZNGmSty+44IJgH+vEnKIMALvvvru3OZWuXr16QTuO1fnqq6+83bx586Bdo0aNvM26tk3T41iif/zjH96+4oorgnacWWJjGaRDh3B8RJ8+fbxts3M444D7+eCDDw7azZw509vvv/++t23cTlr7odAMrWeeecbbNt5uxYoV3p49e3awb5dddvH26tWrvW0z5Dgm4J133vG2TW1nOA6Lx6z9XM7u5Aw+IMz2tHE7nAHG1OQYHiDsh7vuuivYx2U57Lx47LHHepsz22wc16xZs7zNWXpJ5QI6duzobY7bAcJ7gOM5OdMTAH7zm994uyxZemlg1113DbZ5/PF1PPvss4N2N998s7efeuqp2HZHHHGEtx999FFv2/HLsWDVhXTO/EIIIYRIHXroEUIIIUQqKIq8leRC5zTWlStXenvGjBlBu48//jh233vvveftX/ziF97eY489gnb82Zxizil2ALDvvvt6m2WwvffeO2jHKa3PPfect88///ygHe9Lq4wSh3VRc7oqF8ZimQoIXeBJ15SrMHM7W8Qwre7wpLHJshVLw7ZQH7usbXHCM844I+8xrER99dVXe5tlFVt2gsc0y81WOuFtLmPB0gsQ3m+21ATLW3xtill8tDLgtHSbws8yMvcXEF4Plq1sWYGzzjrL2xx6YFPg16xZ422WX5o1axa04yrBPCfYe2348OHe/tWvfgWxBRtWEVeCwRYnPOCAA7zdpEkTb9s+5/RzHtt2/HJYQnVBv8hCCCGESAV66BFCCCFEKqh0eYszoCZOnOjtN998M2j385//3Nsnn3xysI8rc7I7jaUpIHS79e/f39vsxgdClzefr61e+vDDD3ubs0fefffdoB27ie1aJGVd1LDU4D4HwowcXnvLulPZHZ7UR9yO7wfOBgTC+6Zx48aFnHqNhe+5JGmQZQWWJmxWU1z2JBC6y5s2beptzq4CQnmL+8a64fl8WTa+8sorg3atW7fOe042I4nldZZbkigleeuoo44K9vF6SHYsffrpp97mtZastMgSIo8rW2mZZSzOwOTPAcJMIx7D9rpz5iBnlwGVu7hlTcDKXTnselgffPCBt7mqs+1zzvrjPrd9ZCXT6oA8PUIIIYRIBXroEUIIIUQq0EOPEEIIIVJBUYTPJM2b00QPP/xwb9uUYq7YajV1jj1g/dfGbXA6Jqe+ckyCPd/169fnfQ8AnHTSSd5+7bXXvG2r/XKcio3pSTvz5s0Ltu01zmHjs3gVbY492LhxY9CO4wg4XdLeQ6Uex8NwldSkWAdeBZvHmF1Vm8eL3dezZ09v8yrbt99+e9Du1ltv9fb111/v7aeffjpot3nzZm/ff//93rYVu/me4PdYuFQBlzcAwgrxbdq08baNTUlagb26wCU/eDX5DRs2BO04ZmqfffYJ9nGsDX9ne+33228/b8+fP9/bSWOM70NbrZtX5uZUeQvHlXDVaSCMJ0sjNoaH5wCOl7T3MsfecRyXTUVnOJ4q6be6uiBPjxBCCCFSgR56hBBCCJEKKl3e4vRurrT84YcfBu1YtmD3KRBWZOaqyeyOA8J0vKFDh3rbpqqecsop3n7++ee9bVPbeWFEdq3ayq68GOndd98d7Ktp6a4VjXWTsvzHcoNNWWd5a8GCBd5mKREIZVGWO61EwS70UiOpdAPDUi4Qygw8dmxVVXZhW7mM5ZIhQ4Z42/YTy1uMXcCUU9O5z3hhRCB0y7N8vXbt2qAdu9t58VEA6N27t7cff/xxb3OVaWCLO786y1x2Ps1hSwzwHJeUlsy2HTt83yTJSjw/c+kAW0aBpRguncDnYD/Lyihpl7fsgqM8Tvm+tQuE8rybVJWc+4yPYdtZObU6IE+PEEIIIVKBHnqEEEIIkQqKXrbSVsrkTA3OkLDR97xAqJUmOCuAFwu11R/Zfc8L0tnsA+awww7ztl0ck13enIXUp0+foJ1d5FBswUolvKgdS5W2L9l93bVrV2///e9/D9px3/IipVxBu9SxchZnM3F21EUXXRS0e+ihh7zN1XOt1MhYSeSvf/2rt7mfWF4GgJdfftnbXJ35nHPOCdr97Gc/8/acOXO8/Yc//CFox+52llCtFDN9+nRvczYmEEppSfJI7jtXZ6maQwD42vAiz0DyAqE8fjg7zs6fLF3utNNO3rYZuXwMHs/2OrJUmZT9w5XC7WKkfO+lEZuZyPcAZ3ZZaZHHjpW+SgV5eoQQQgiRCvTQI4QQQohUoIceIYQQQqSCosf02JRWjhvgCrCjRo0K2nEcgtUdOY6HsZU9OR2TtWxbeZS1y/fff9/bthoo68YdOnTw9sUXXxy0e+utt7yt1X/DmBAbH9KqVStvc3Vmq9FzHNeRRx7p7QcffDBoFxf7Y1Mp49K4SxEuycDXn6uKA2E824033uhtrugLhDEddvV0TkXn6rwcwwMAN9xwg7c5JsSu+nz22Wd7Oym+Iy69lksdAOGcYO+xO++809v//Oc/vc2rTQPAVVddFXse1QVeLZvhuCgAOPTQQ71trxVXZOZYD1vtl/fxnG7jvfi+4f6ypUa4ajaXX+BSFUBYysTGI6UdLgkAhKUKeO6zcTuFxqnFlWuw76+OcW/y9AghhBAiFeihRwghhBCpoOhaS1KFRpacbJXk4cOHe7t///7BPq7iy2mQVrLo1KmTt9n9adOmuXolp1keffTRQTuWUq677jpvz5w5M/bc2X1szz0tsJvcSpUs/7Gr1aasn3jiid5mGcwuOMpp7+xatanLNqWzlLALbrLkwN/717/+ddBu7ty53mYZiMsKAGGVY1tpmT+Lue2224JtTm3+zW9+k/c9QFg2gseSTSnn+4ptOyewm5/vSwuPdSvn1AR5i+dZvtZW9uD508KyFdtJFZR5vrfSJ8+7XF3ZnhPLXRwOYEuBnHXWWd62labTjq1EHpd+bn+fuZ297+NIqkxeHdPe5ekRQgghRCrQQ48QQgghUoEeeoQQQgiRCio9f5rjAzjt26a0sib54osvBvt4NWROgbca/b333uttTsGtV69e0K5nz57e5qUmbKoux/tw7IjVLTkN265kncaYHr4GHF8AhHow77PXlNNYGY49scfguAGrT9tSAqWE1dg51ZtjPWyKMq9wzjEcNt6Oy//bfRx71aVLF28/88wzQbthw4Z5+9JLL/W2jQN59913vc3py0mpsDYFmuH7KukYZ555prdtan9NgGN1OGbRxjNyP9u5idsmLUPB15tja2bMmBG0a9++vbd5LrUxQjyGk/qIV5JPKmeQRmyMU3lia3gesXNKUhwPY+eH6oA8PUIIIYRIBXroEUIIIUQqKLq81a9fv9h9Bx54oLeXLFkS7OO05O7duwf7mjdv7u0BAwZ4e9KkSUE7Tle97777vG1X/23durW3uSKsdQWff/753r7rrru8bV2w7NYdOXJksG/cuHFIGwsXLvS2lR7YDcsrQyetcs0cddRRwTanv7KkZatr8/1VarCcAYTVrFnys1VsR48e7W1efXv16tVBO5ZELCz7sjTFVbSBsAL7O++84207hnkF9j59+nj78MMPD9rx+SZV22bphN8DACeccIK3r732Wm+/9NJLscerLtj7m6VL/s42RZ1LOdgqzixNHHDAAd62VdV5DPOY4+r2QCh3ckkKm17N/cL3clL5EyuR8fWw/ZwG7G9XofKWvY45ktLXuV/sigN27qgOyNMjhBBCiFSghx4hhBBCpIKiyFvsuuzWrVuw77LLLvP24MGDvX3//fcH7Q4++GBvc+VmIKwCyxV5rQtu6dKl3ma3foMGDWLPvXPnzt62VXv/9a9/eZtlMHazAmHmx3HHHRf7WWmBr491f/I1Zvc6Z8AlwYuKAmGGDx97+fLlQbs0LTjK37VZs2benj59etDu448/9ja7pW1GFe+zWXBDhgzx9rRp07x90003Be14DHP2pOWRRx7xNsvGtvIz31c81llus++zlabjspV4wdbqip2DWDJk+d5KRDxn7rnnnsE+lqJ5n61uzsfga2+zabkdX3srefM5cgYtVwwHwu/Mn2v3pVHeSqq0zdj7gUM1WBKz86WVOHNYeUzZW0IIIYQQVYQeeoQQQgiRCvTQI4QQQohUUJSYHtb5bfVOjg+45ZZbvH3qqacG7X7yk59426bgcjzNk08+6W27ojmn6nLsj40rYe2SP2vy5Mmxx5swYYK3bervPffc4+127doF+1599VVv//SnP0UaYF2Xyw0Aod7OMRVcvTUJq9cvW7bM29zPVmu21YhLGS4NwbEOttQCV7XluB0b68Ep0HXq1An2/fKXv/Q2399cWgIATj75ZG9zWYspU6YE7YYOHert4cOHe9umrHM1Ya4s3aZNm6Adx6Nw/BcQ3i9cPmHevHmo7nAcJRBWMOfxZ6suc+kG25dxla3tfcMlDDj9fPPmzUE7nvv5N8LGTvI5cjq87S8+vp3T+XrYOScN2LgrJin9nPu20FXW+T22nIUtpVAdkKdHCCGEEKlADz1CCCGESAVFkbc4tXD+/PnBvrhFzKwMxilxXNkVCNNu2X1m01h32GGHvOfHFYKB0AXOrjqbNv3mm2/mPR4vbAqE1YQLrSxcyrCkst9++wX7uDo23w8sYSZhKytzKjTLltatn1RVuNS46KKLvP3EE094214TlgZ5XFkX9WGHHRb7WSwtcZkAK2GwdPLwww97OymFltOX+/btG7Tj6r9c7sKm1rIMUrt27WAfyyd8fja1P7fIbZKEUNlwJXIgTNUfM2aMt60czO3smODxwynQVvaIWxTUHo9lUU5nt+34uvI+245LLHDoAbC1tJY2yisrcV/yWCy0IrPtI7v6QXVAnh4hhBBCpAI99AghhBAiFRRF3jr66KPz2pacmzgfl1xySey+mTNnepulKutmZfc9Z4XYxdjatm2bt92cOXOCduy6Z8nmggsuCNrZbLO0w9dj3333DfbxQrOc0WH7kl2ovK9evXpBO67ezdIGu9aBreXUUoarVr/xxhvetpWWeTvJLc3yg5VVONuG9zVq1Cj2eEnSCWfZPfXUU962Y47vA5ZQbVVgPncrpfG58+faa5FbNNh+96rEniPLWCx1WLmPr1XSfMz9YjMheT7lfXaM8bjlsW77gc+XKzfbOYElLCvb2Tk+baxfvz52X5JsxfKU7ZdCsPJWeY5RbOTpEUIIIUQq0EOPEEIIIVKBHnqEEEIIkQqKEtOTBKdBsv5nteaOHTt621Zrfu6557z90EMPeduuBMtxN1zB1aayc/XOpBWV7YrxOZJieJJScNMC6+02joTTmnv06BF7jLiYHl5BGgjTjm08R1qwK59zmvbVV1/t7euvvz5ox/ET3E82/ZfjJWzV3u7du3ub04gHDx4ctOO4vA8//NDbtrLwueeem/ezbLo9VyDm72+r+HK5C7syN8fF8D3LcxawZbX36pSOa2NYOIaNY3Xs+OO5y8Z38DGSSjzwMThGz54Tx/twX9rry9tsc4o6EJYD4VIVwNZ9mzaSVlkvNM4mqR3fK2wn/b5xzJyNsaxM5OkRQgghRCrQQ48QQgghUkFR5K04KQLYOt0xrh1XX+XFQgGgc+fO3ubUPFshlavAxlVntvs4zdK6dE877TRvJ31HdsnafWmUt1hesW5olh/YDW8Xj41LmbX9ynIXLxpp3amHHHKIt0utxIBdfJHvVb4OVt564IEHvM3XxMo4H3zwgbdtqjAvtstyGafKA8Do0aO9zeOMXeAAcOONN3p76dKl3r755puDdnx/8PdPqvTO8rfdtrIpk6syX52q/to+58VD2bZyZNxiy0A4NnnM2T7ia8z3l5XSGL6n7OeyPGmlWob70p5T3GKpaaFQaT/u9xhITm2Pk76Sft9YurYVtCsTeXqEEEIIkQr00COEEEKIVFAUeSvJxRW3z7rL2D1pXbe8ACm7yayrjt3m7FqbNGlS0I4zQXgxRSur8QKk7Fq1rlQ+jzTKWRbOqhs5cmSwj93hq1evjj1GXPbI2rVrg23u5/3339/bXE0b2HqxyTQycODAYJvlLa5IbLNwWKK0bnRe0JQljI8++ihod91113mb+2LNmjVBu7jFTe3iwi1btvQ2yzJWguL3JVWT5oVsrUSWe1/SIoyVjZV3eM7kTFi7EKWdWxmekzl0wH5vvt6cNWSPHVcNO0nC4vvQSo48X3BWXr5zTBs2jKA8VESWF5OUUVaZyNMjhBBCiFSghx4hhBBCpAI99AghhBAiFVR6ReY4bOox67/77LNPsI91Xl692eq4nILKcSU2bqBdu3beZt3RpuN26dLF26yZcjVYex5JlUzTAsdsWP2e40VslV0mLjbKVuHOVcsFgAMOOMDbe++9d0HHK0XiyivYa8DxHvvuu6+3bUwPx8lYnX7YsGHeHjNmjLdtDE6DBg28PXHiRG/36dMnaDd16lRvH3HEEd7mkhZAOH/w97Vp6Xy+NhaP79O4qsDAlliV6nwP8fzJc5Cdj/i62evB8xjHP9nrwfMxx/HY43Fquj0Gw7Fg/D323HPPoN2iRYu8bce3jcdMGzaWje/VQmNweK4u9D1JY8KWlKkq5OkRQgghRCrQQ48QQgghUkG1WXDUctFFF3n78ssvD/axW5PdaWeccUbQbuHChd7mdPO2bdsG7bhi6axZs7xtpZMhQ4Z4+7jjjvO2lbdECF9fK4ewTGFlx0Kwbld2obKkwosTAtVbmqhoeMxxOYX3338/aMfXpHnz5t62ZQHYdc7SBgDccsst3j755JO9PXbs2NjPuuGGG7z985//PGjXtWtXb99xxx3evummm4J2HTp0yHtONpWbK7jbtN569ep5m0sp2MWQc+M9qcp7ZWMlyLj0YCsxcMkBW/IjLpU8SQbjubnQSudJVeuTfi+S5LLqIqVUFbY0QVLlZSZOBqsIeYvndx6vlY08PUIIIYRIBXroEUIIIUQqKLq8VahbjKt6AqGrmRcOBcKMjH79+nm7V69eQbsDDzzQ2+x2sy55dvF2797d29ZFylkFnN1iSXILpklWycHSIi8CCoSLUlo5sRCs25Zd3tx/SZVnSx2WYTgL55hjjgnatW/f3tssl9gsRu4nlouA8DpzJtfzzz8ftGOJhGWr/v37B+1Ybp4zZ463bTVprs5bt25db3MGHxDeL3YRWv7+LMl+/PHHQbtcNqmV7KoSriANAPPmzfM2f0+uzgyEc6nt57gsKruQKG/zXG0rXvNcmDQeWcbi3wV77jNnzvT24sWLg30cfpBG7CLB/LsTt3hzElY+jMsCTQpZKU/4QjGQp0cIIYQQqUAPPUIIIYRIBXroEUIIIUQqKHqgg41hidMQbTwHr67829/+Nth31llnedtW4mRYu+QYgsmTJwftOCWVdWeuGguE6bO24iWT5viRfDRs2NDb9rpx7AjrwTYF18Zh5bBxV9yOq7La+I008cgjj3j7D3/4g7c5LR0I46GSKmXzGLGVluPKN9hyElxJneGyEEB85V5bkZnjBVasWOFtO69wVd/Zs2cH+zjNl7+/jTfMzQNPPfVU3nOrCmy1Yo7p4PR1jnMEkitP87Wz14Dh1HYecxzLB4TlA5JiSXj+5H61K6nzOdl5oFatWrHHTwO2ZEFcLKldxSAuBteWL+A+KnQFAhsbV1XI0yOEEEKIVKCHHiGEEEKkgkrXYeLcbLbqKbs/edFBIHTlsvvTutY//PBDb3NKJy9cCISu7Mcff9zbV155ZdCOUzCXLFnibU6RBeIXeEwr7Gq2abErV670Ni/4aFNQrVs+h3Vrx6XWVqfqucXm7rvvDraHDx/ubU7hti5wlh459djKtTyurLxs+y3HiBEjgm0rT+Ww8wCnXjMLFiwo6Hicyg6EVah5rAPhdbvuuuu8zfclsGVR1FtvvTXvZ1YFdhzwIs2vvvqqt+3cx7KFrbTM6ew8x9kyBTwv8jiz8qlNo85hJRWWWfnYtsIz/w5YCdZW4k4b69atC7YreuFrlrQK/Y2TvCWEEEIIUYnooUcIIYQQqUAPPUIIIYRIBZUe0xMX72I1R17Z+d133w328erIvFzFRx99FLTjlEbWOG3KOi9fwWnpL730UtCOy+Cz/eyzzwbtFMcTjy2Xz6nkrNnbmI24mB4bq8P3QNLq7qVM7969g+233nrL2xz7YeNnOKaD4zZsfABj49k4rbhnz57eHjlyZNCO461OPfVUb3fr1i1o165dO2+fe+653ublS+y5M3ZO4NgPXgUeCO+5v/zlL962afTVERsvwennPOZsf3Gavk0x57GVtHQBx+pwP9gyE3zv8bFtTA/HkPH8bn8juKzJsmXLgn0cj5RGbDwVl3EodDkQJmk5KY4Lsynw/FuoZSiEEEIIISoRPfQIIYQQIhUURd4qT8q2lbA4ffaUU04J9rGcxO7Pli1bBu04PXrp0qXeHj9+fNDuhBNO8DanSPLq60DoMoy0kmkgAAAgAElEQVSrFCuSsZVj41JLkyQVxt5ffO+xO7w8K7jXVDp37hxs83dnOcOWD1i4cKG3uYq2TYG/5pprvG3TkHlczJ0719u2AjO7wVnqsmnJPL5ZErHnzlI2yy9WFh0wYIC377333mAfy9lxMml1Jem6sXxv0++TxgVLH9yvTZs2DdqxtMaSlq2KzGOdZTArifA+Ht82VZ6lOjsPxMmdacFK1wyPI9uO+7zQMh98b1jpk+8v+3taVcjTI4QQQohUoIceIYQQQqSCoshbSZJWnPRlZY977rnH29aVzZk4y5cv97aNDudsAXZzW3c1u8ZZBmvdunXQjl11XElYFI69pqNGjfI2Z/7YbIw4uDowEMot3Edt2rQp03mWEv379/c2LzhqF2HlqsZvvPFG7PFOPPFEb9txwOOHpTSWjS12cVmG7wmWN2wmD4/pjh07ettK3ny/Wfll7Nixec/BStkVXd22IjjppJNi97GEcdtttwX7eLxMmzYt2Mf3B899Vkrj65E0hlnGsgtYMjxvs1xiF7dlWfSBBx6IPV4a4YVf/3975x5vVVWu/+cVFQUUVFAEuYgiF1EQUPOatzAv5THvR0u7WVmdk93sZP7yWJplncw8pdmhsFLT0rwrJFnmBQXkLl5QREWuKkoqKo7fH2vuwTNe1pysvdmXtfd8vp8PH9655lhzzTXHHGPN/T7v+w4gla34966o2nqt8Lzh5W6+9+bPn9/oY7cE8vQIIYQQohTooUcIIYQQpUAPPUIIIYQoBa1ekTkPTrEEgOnTp0fbp7qNGzcu2qzz+rgB1hpZr/ZVWlnn5zQ9nzY9a9asaPuKsHn4SpZlr9Z81FFHJdsc98H3QN++fWs63pAhQ5JtLj/Aqdtjxoxp1Hl2JPia/PnPf472Zz/72aTd/vvvX9PxOKZj8ODByT6/XW98+MMfzt3HsSQcf1Jr6m69wuf/ve99L9nHsTVcZR5IY3L42vi4jbzV033aOMcIdevWLbcd31+9e/eOto8H9NWlxTrOPPPMZLtPnz7R5n7w1duvuOKKaPPqBFzCAgAmTJgQ7RNPPDHaV111VdKO52euoN2WyNMjhBBCiFKghx4hhBBClAIrWkhsvcZmywHUlkssmosBIYRezX1Q9WWbof7sOKgvOxbN3p/qyzYjty8b9dAjhBBCCNFekbwlhBBCiFKghx4hhBBClII2f+gxs+3MbEb2b4mZvUTbuUvlmtlAM5uTs+8iMzsiZ99ZZtbHvXaqmZ1vZoeYWW15u6LRNLWvRfvFzNZm/TvTzKZrfLVPsvlxrpnNyvpzXzNbaGY9q7T9qJl9K+c4mmPbABqHc7Ox+DUza/Pf/7agzev0hBBWAhgFAGZ2IYDVIYQfb+Qx/1+1182sE4CzAMwBwAu5HAXgCgAfAbAawEMb8/miOhvqazPbNISQvyhPM2NmnUIIazfcUmwEb4UQGvr8SAA/APDBtj0l0RjMbD8AxwIYHUJYkz3o5P6REkK4DcBtVY6zKYBDoDm2LeBxuD2A6wBsDeC73Ki15+C2oF086ZnZ7mb2aPakOsvMGiqgdTKza7Kn14lmtmXW/rdmdmJmLzSzH5rZdACnARgL4A/Zsba0SrXAUQBeAfB5AOdm+w7KvEmTs8+8z8z60/GvMrOpZvaUmR3b2teko0DXcgqAH5nZKDN7JLvmt5jZNlm7+81sbGb3NLOFmV313jCzM+j1q7MHXpjZajP7iZnNBFAf1bLKw9YAXgUAM+uWjanpZjbbzI5raGRmF5jZk2b2TzO73sy+3mZnLABgRwArQghrACCEsCKE0PBH45epD4cC0Zt+ZWbz+L4Rbo5tg+9SekIIywCcDeBLVuEsM7vNzCYDuM/MuprZ+Gz+fLxhbFaba7O2d2beozlmdkqbfrkaaBcPPagMlJ9lT6pjATQsBTsYwP+GEHYH8BqAE3LevzKEMDqE8HsAUwGcHkIYFUJ4C8BeAGaGEJ4DcBWAn2b7HgDwcwATQgh7AvgDKt6gBgYC2AfAMQCuMrMtIJrKTgD2DyF8FcC1AM7LrvlsuL9EqrDevWFmwwCcAuCA7PW1AE7P2ncFMCWEMDKE8M8W+C4iZctskpwP4NcAGkoCvw3g+BDCaACHAvhJNgHvjco4HomKB3ZstYOKVmUigH7ZH3i/MDP21K3I+vCXAPIeThvG98ew/hwr2oAQwrMAOgFoKLU8GsCJIYQPAjgfwOQQwj6ojM3LzKwrqv8OfxjA4mw+HQHgHtQ57eWh52EA3zaz81DJv38re/25EMKMzJ6GyoNINf5YcOwPA7g7Z99+qLgBAeB3AA6kfTeGEN4PITwN4FkAQ4u/gijgphDCWjPrDqBHCOHv2esTABy8gfdWuzcOBzAGwGNmNiPbHpS1Xwvgz1WPJFqCt7IfuKGojLVrM++qAbjEzGYB+CuAvgB2AHAAgFtDCG+HEN4AcHtbnbioEEJYjcp4OhvAcgB/NLOzst03Z/8Xzb83SUaueyaFEF7J7HEAvpXNnfcD2AJAf1Sfa2cD+FCmphwUQljVBufeKNo8pqcaZnY81v2F/5kQwnWZe/QYAHeZ2edQedBYQ29bC2DLnEP+q+DjxiHfQ1SEL3CkgkdNp6h/GngP6x7So1ct594wVDx0/1XlOG9rAm4bQggPWyUepBeAo7P/x4QQ3s3kSnlL65RszNwP4H4zmw2gYXGnhjl4LfJ/T2oZ36IVMbNBqPTZsuwl7iMDcEII4Un3tif8XBtCmGxmo1EZz983s/tCCBe19PlvDHXp6Qkh3JL9dTgqhDA166BnQwhXALgVwJ4bcfg3AGwFAJlnYdMswDbZl/EQgFMz+3QA7I49ycw2MbNdUPEi+BtENJLsr4RXSev/OIAGr89CVP7aBIC4wl3OvXEfgBOtErAHM9vWzAa0/DcQRWQxH50ArATQHcCy7IHnUAAN/fMggI+Y2RZm1g2VAFrRhpjZEFsXRwlUYiCbWmXYz7GilTGzXqjIjFeG6tWJ70UlVsuy9ntl/68311olE/rNLHTkMlRksrqmLj09VTgZwMfN7F0ASwBcgkpQZFP4LSoxOG8B+AkqrvUGbgfwpyxw68vZv9+Y2TdQcet+ktouAvBodh6fDyG83cTzESlnotI/XVDx5jVc8x8DuNHMzgZwJ7Vf794IIbxiZt8BMNEqaZnvAvgiVA6+Ldgyc5MDlb8gz8ykzD8AuD3zGkwFMB8AQgiPmdltAGYBWIqK+7zuXeYdnG4Afm5mPVDxuD6DitTVlAfSZI5VXE+r0TAON0OlD38H4H9y2n4PwOUAZmXz53Oo9HW13+G9UYn5eR+VefYLLfotmoFSL0NhZr8G8OsQwiONfN9vAdwRQvhTi5yYECXGzLqFEFZnD77/AHB2CGF6W5+XEKL90148PS1CCOEzbX0OQoj1+JWZDUclxmeCHniEEM1FqT09QgghhCgPdRnILIQQQgjR3OihRwghhBClQA89QgghhCgFeugRQgghRCloVPZWz549w8CBA1voVDae995LF4d94403or355usWBX7//feTdp06dYp2ly5dWujsmsbChQuxYsUKa+7jtre+fOWVV6K9xRbrCve+/XZaHundd9+Ndt++fVvo7JrOtGnTVoQQejX3cVuzP3n8rFmzJtnH15+TJLI6Z5FNNln399a//rWuGGz37t2TdjyGud893bp1y/2slqJMY5PH31tvvZXs475ke+3atPA5b/PY5Pe0JS0xNuuxL8tAUV826qFn4MCBmDp1avOcVSPwGWZ5kxoPTAC47777or3TTjtF+80330za9ejRI9pjxoxBHkWTeEsxdmzLrLfYmn3Jkx0/YBaxcuXKZPsPf/hDtIcNGxbt+fPnJ+1eeumlaF966aWNOs/WwMxapEBia/Ynj59nn3022cfXv6jf+Y+LKVOmRPuYY45J2v3tb3+L9tCh65a38z+UH/jAB6Jd9HDUnHSEsVkr119/fbRnzpyZ7OMHTrb9GF61al2NyYsvvjjaW21VHwWaW2Js1mNfloGivmzTOj38F2PR036tDxjf+c53ku1Zs2ZFe9ttt432dtttl7RjbwEP7qaeR63fq6NS5Ekrgn/8zjjjjGTfZpttFu1DDjkk2i+//HLS7sEHH4w29/M3vvGNms6h1gfsMvG9730v2V62bFm0/Q8b/wHBfcMPqgAwY8aMqrb/o+Ovf11XMJ1/PPhhCAC++MUvRnvixInRvuCCC5J2Bx10EMQ6XnzxxWjzfAmkD62XX355tNmbBwB77LFHtK+99tpoc58AwKabrvu5+frX1y3IzuMZAHbddddoDx8+PNpdu3at/iWEaATl+0UWQgghRCnRQ48QQgghSoEeeoQQQghRCto0pqfWeBcfrHr33XdHm+NAvNY8bdq0aL/zzjvR9jEnHIdw2mmnRfukk05K2h1++OHR9lkmTBnjeJii73/VVVcl2zfeeGO0OW7HZzw8+uij0f7LX/4SbR+D07t372hz/48cOTJpd/zxx0f729/+drQ5y6/McDyHD1bm+DgOXAWAI444ItqTJ0+ONo8xII3P4ABlH9S62267VW3HGV9AOh45duTqq69O2pUxpmf27NnR5hgpIM2+8300ZMiQaH/84x+PNieIAGkMzs9//vNo+9jJ559fF1u69dZbR9vP23z8SZMmRXvLLbdM2n3+85/P3dee8b9PeRlxtcZKAmkQ+d577x1tn9TTs2fPaPP90K9fv6TdFVdcEe1tttmm5vOoB8r96yyEEEKI0qCHHiGEEEKUgjaVt7jIlU8pnjdvXrS9lMTuUF8Ai9lnn32i/fTTT0e7c+fOSTt293H6uj+nXr3W1Trq06dPtEeMGJG0+8///M9oexdvGXjmmWeS7fPOOy/ar7/+erKP5aQiaYnvgVdffTXaq1evTtoNHjy46vt9oUJObT/uuOOi7fv8sMMOyz2njgwXBfQSIvch1+UB0n567bXXos1ucyBNdedxNWfOnKQdS1U8Nn368tKlS6PNhS29VFAWuP/uuuuuaO+8885JO5ZIvETEEiLXJPJlBTjEgKUYX85g++23jzb3uS8LkSdV+fIUv/rVr6LNc257xI+xPGoNneDfTwDYfffdo33kkUdG+8ADD0zacZ9z3bvf/e53STueP738zdRj+Zb6OAshhBBCiBZGDz1CCCGEKAVtKm9xppQvHc+SkXf9cYbHkiVLos1ZXUDqyh00aFC0vWuc3b8cpX7iiScm7ViOY4nNZ5d99rOfjfbNN9+MssFZO0Dq5t5hhx2SfeyG50wSlg+BtM9YUvHSJ2eSsLzCMgmQZibwObDbFiivvMUSImdrAel49FIES8c8XnxmEMsbfP05y9Ifb8GCBdH28wXfOyx5+7XZWJrjDKKOxsKFC6PNY8SPl6JrxdIXX1/uLyCVTljO8GOOrz3LKEVZSCxVsjwGACtWrKh6fsD6IQz1Dkt8Xu7j8VbrUkicaQykyzDdc889jT4/nwW5yy67RJtXQvj+97+ftKsXSYupvzMSQgghhGgB9NAjhBBCiFKghx4hhBBClIJWj+nh9DZOUfYVH1mT9VUjWYfmuJ2LL744acep0/weHzew3377VT3XIp24S5cu0fbpuFx5lKuhAumKxB0VjrMC0uvmrz3H6nCKpO/zPffcM9qcuupjhDhOg2N6ilZo5n0ch1Jmxo8fH20f68HxbL5kwKJFi6LN8R0cjwOkcSFsL168OPd4zz33XLTHjRuXtOP3ccwf31MAcN1110WbK/p2NDhmhuOufCVrjgspKhnB7XyMJd8D3M7H1fB9w+183AfP1Wz7dnxf+tIJHMPZ3vBlWDjmqeh6cDVsPwdzpXSmqPozx1P5+KxRo0ZF+/LLL4+2j+mpR+TpEUIIIUQp0EOPEEIIIUpBq8tbL7zwQrTZxendruxmYynJ72O5pKgyJMOLUgKpjFUkb7CLjz/XSzZMGeWt5cuXJ9u8kKh3p7K7lmVCL0sMGzYs2ix9+YrXefeNT5nmY/A58P1ZZtilPnr06GQfl5O47bbbkn233357tD/84Q9H+ze/+U3S7phjjon2HXfcEe1ly5Yl7Viu/OAHPxhtXxbh2GOPjTbfEzNmzEjalaVCOqeHsxzlJRGe07wEzPNanoQFpDIIL0DrU9FZZuN5tmhOYOnTy298b3BIAdC+5a2ilHUvMzE33XRT7j5eMJbxi73yNS5Kif/CF74QbS7L8tOf/jRpd+6550a71nR7D98fRan9tSJPjxBCCCFKgR56hBBCCFEKWl3e4oU/WfbwGSLsai1yY7E7jo8HpC5eljO8GzevimiRm5Fdv0WSGGcQlYVVq1Yl2+zm9q7svCyObbbZJmnHEuTQoUOjzZVBgdRdW1RxN28hSi+zlpULL7ywpnZ+cVnOtuKMTC9Rc2Vglpx8Je4XX3wx2ryAsJfVWC7jqtxlheckHlf+vuc5zc9j3JYrYPt5kefWvOrB/n0sYeUtMAqkWUj9+/dP9vHczxXE2zteguTfGv7OPNYA4IEHHoj2Zz7zmWTfJz7xiaqfVVS5uqhSNmeD8b3BqxsAqbzF/c+SKJD+7hbdN82BPD1CCCGEKAV66BFCCCFEKdBDjxBCCCFKQavH9LBGy+mNXpNlza9v377JPo4PYL3Pa5Csf/Iqsz7tj+NF+HO9lsjaKseO+DgQ/l6+Ym0Z4NWPgVSz52rKQBpfVVSmgFNXhw8fHm3flxxfwMfw7TjWijVp/7mimEmTJiXb3J/cZ34scUwHj1MfD8axfhyz50sQPPbYY9HmmJ6Wjg+oVzg+h/vEx1LweNl2222TfRzvwfOdrxjMY4bnyKL0aq6a7ecL/lyem30sJs+tPia0vcHxUz6mJ69S9syZM3OPd8011zTPidXAgAEDou1LSaxcuTLaHLvn7w3uv1q/f1ORp0cIIYQQpUAPPUIIIYQoBa0ub3GlXXZbedcqu8aLJCJ2eXtXNrvJ+BhFKZJ8Tt5lyjIW7yuqGN3e3a5NwS8WyunJPi2W3dcsQXrJkCUovoe8y5v7j13tvs9ZRuH7xssmZaXW6ql+zG2//fZV3+dLN/iSBA0UlQxg+dq7vL1s2oBP0S5Kw+1I8L3P85Eff0ULMfO1431e3uKxWSRpMXwefn7n43HYgx/D3I7Lk7RHvKSTx7333hvt888/P9nH18NXUefVCnju86EjeWnlXnbmchQ8h/syIVwpnRcEP+yww3LPvYimVnVm5OkRQgghRCnQQ48QQgghSoEeeoQQQghRClo9podT2Dht0eu1rBN67Y7jZIo0Pt7HcSZ+uQrWq1nz9zoxa9l8fj4OhD/Xa6FlwMcxFZUf522+bj5Wh7V9Tmn1ZdQ5poBjEri//DnyvVeWlOaNgdOXOV4LSPuQ9X3f79yO9Xwf38HlHziWa9iwYUk7v8p2A2WJ4fHwuOCYQx/Tw7Ek/t7na8f7/NjkfTzO/LXn+ZP7v6hMBB/Pt+P4r/YY0/PUU09F+5e//GW0b7jhhqQdl3kpguexxx9/PLcdjzcff8nwPTR48OBkH8+tHJ/nf8dfeumlaB9++OHRPvDAA5N25513XrT97/ORRx4Zbb7XmlqOQp4eIYQQQpQCPfQIIYQQohS0urzF7muu0OglEXbD+vRWdsmynOHdqey649Q879Lj9Ncidx+7V5cuXZp77uz+ZSmgLHj3JLshfV8uW7asqt27d++kHbvGfXkDJk8u8+c0duzYaM+aNSvaPnW0qFJqWWGp0Y85lqxZfuASFEA6vvka+/HC/cnH8H3h5cuy4b8/X0ee03xaOo8rn4rO2zx+fBkAble0Wjbvy1s5HMivil90PH9O9QhLWADwrW99K9osz/n0bZ4zd95552g/8cQTuZ81dOjQZJuPydfUhwewPMVhGywzA/ny5PTp05N2HM5y6KGHRvu5555L2p155pnR7tmzZ7LvjDPOiPYFF1xQ9Xs0Bs3iQgghhCgFeugRQgghRClodXmL3WLsWvOVWDkrxGd+sEu9KFuAXeC8qJ2PMGdXMLvMvPuMP4uP7V337O4rqjDbUfFVWbn/fKXQKVOmRDsvKw9I3eucFeSzURjuV+/G7dOnT7SLMh0WL14cba4YXWa4P/1CwSxZcx/6/mSK2vEY5EwQL+d4aaZseNkqT+7zcxXLFn5e5GPy8XzGHsP3hu8Tnhf4eF7S5Hmc51l/PD5Gvcpb77//fvwNOOecc5J9fD1YSvLhEnkV7XfYYYekHcuEXjLMWxnAz9U8nvma+orqfE4csuIrMnOWJZ/DbrvtlrTj4/s5gCs5cxXqolCEIuTpEUIIIUQp0EOPEEIIIUqBHnqEEEIIUQpaPKbH67Cs17G+7ON2uGLj5MmTk328kjO/z+uWrBXz5xalu/I+HyPEKzlzdcmHHnooacfnVBRz0pHg6+s1etZaOZYGSLVnTqv09w3H9HA/eA2Z4xK4/7wmfcABB0Sb+9zHRnAV4DLF9BSlg3IKNMfwAOn9zrYf33npxn5s8j3BcQ8+5bWoqm8ZKBovPMb8/c39UBQXVXQ/5MXxFK1wz59bVEk/Lx3eH69eY3pWrVqFu+++G8D6ZTOGDBkSbS6B4ucqvr5FsS95MVP+s9n26fEc48XX18/p3C98TxWlwC9fvjzaPuaIY4SKYmRvu+22aP/bv/1b0k4VmYUQQgghCD30CCGEEKIUtLi85d1s7IJil5lfxG6PPfaI9h133JHsY9cYS1pFblKuauldvOzu4+N5eYurZu61117Rnjp1au7nlgVOWyxykw8YMCDZzkuz9DJH3r3iXd7s5i5y3RdVjmXKIk82Bpa3fAVzrpbN6a++P3nM8TF8OjRX6eYx513gXAm6jPh+yLun/dzHFX79Yq98zCIpiY/JUkSRJMK2lz75+AMHDqx6PkC+NO6P6eeI1mTzzTePqwH468GLeHLlYg9/74MOOijaixYtStrxGHjxxReTfXwNeE7z44jlKD5fL83x9WVJy8+X/L7+/ftX/RwgvYd81X6uPH3LLbdEW/KWEEIIIUQBeugRQgghRClocb8fV0IGUjcbu898FghXbKzVdethlzp/Vq2LgHqXPLt/ubJwkVutLItUcuXpHXfcsaZ2QCp/Fi1qyLIjyxycXeWPz/3iZdbZs2dXbedd4WWsqL0huEp1UaYe91mRTFhr5h+7x71cybIYH69I8m7qgoX1SNGcxuPKLxzJY/XRRx9N9rF8VJQZxPv4s7yUxvM4SyI+W4n7luUcL8XwMfx8wXM1Z/21Nl27dsV+++1XdR+PIx/ewfBvIctCflFRhmVLID+j2H9uXviJv4Z5YSU+m5YXjuY+8ZXcWY7zv5n9+vWL9rXXXhvtq666KmnnJbM8yvGLLIQQQojSo4ceIYQQQpQCPfQIIYQQohS0eEzPqlWrkm3W61gX9DE9rA36lEaG9d+ieACOLfLpjXxOeZqmh9v5NFvWqP25sx7uq1e2Z4oqsXIaq1/RPK+Sqo8H4FiEhx9+ONoN6aANcHwBa8heu+bqoBzn4c8nb7XqMrNkyZJo+1RxjoHiMexjeni7qAo69wensvo4P27HfcvV28sEz1089/nK1RwH4e91nieLyklwKY+iseSr8Oa9zr8ZPL7nz5+ftPNzBMMxI20Z01MEX2+u9v7MM88k7fh3YsGCBdF++umnk3Yc/8Sp/gAwZcqUaPP16NWrV9KOK53z2OExD6RzK98nPiZy5syZ0eaYLD9vcDtPz549q567j+k599xzc4/ByNMjhBBCiFKghx4hhBBClIIWl7eKFrhjV/jw4cNrPgZLKUULibK0xG68oirR7DYvSndlfBooV5H1586uYO9abM+wVOklPZYsfAkDdrez69JLi3z8okUIuc/YherTXdl1y4ug8sJ/1c5DpPe3h2UFTof290ReurgfL9zvPM58aizfE3x+ZZG3/PXluaqo4jXLvl4y5LYssfgyEQzPwf543H9+cWiG5+eichI8l3pZtOj4rcmaNWsSSYphGbaozAdLTuPGjYv2mDFjknYsOfkFebkvuSq+7yO+9jyOisII+Fp7CZKlS5an/e/AwoULo82lKYD0N4Lvw5tvvjlpJ3lLCCGEEILQQ48QQgghSkGLy1s+a4PdlewW46qTGyJvQUufLcDbfB7+/XmL33m5LE/e4qqTADBv3rxo+8wEzuzqSPIWyxpFWRveRc2yE2cw+Mw2dt2yW9dXc82TQ7xswhkB7O71i/iVpaJ2Y5gxY0a0vVzCY8m7zhmWDfMWQwTys4t8v3AGCVe6HTFiRNKuI1VhZrwMm5f96scVy1tFC4nyuPJSMc9/RdIzH4/7wX8uzyU8d/gsLA6P8J9VT/LWs88+W3UfLzLKc5X/PeE5iSUdXy2e7wE/Pni+mzRpUrT9bxBnUbNU5eU3HpcsVfnfVg4xYLnM/97vueee0fbzRl4m3pw5c9AUNKMLIYQQohTooUcIIYQQpUAPPUIIIYQoBa0e08MaH6fs+ZS4omOwdllUCZi1bE5v5JWAgTTdkfcVxf4wrEcCwO9///to+1R8H4PSUeCYGx/7tMMOO0Tbp9ZyhWbe51NQuUorX8OimBu+H3w7jkvg8/WVwYvSs8sKa/g+JZzHSFF8B8dccJwJ30dAGnPC7/H3GMf0+LIDZcCX4eDrwzERvGI1UHztuaov958v0cGxJRy34+OM+H08Hn35AR7ffAx/r82aNQt5FFXxb03WrFmzXvp4A3yOnNbuq2bzvc3zol9VnOOffB9x32677bZV3wOkY5tjHX0cF4/FvMrd/hz5t9//lhatkM6fzfNzU8e5PD1CCCGEKAV66BFCCCFEKWhxecu7XfOqbfpFQPOqLvttdpPmLV4JpNUgfaofw+5ZL7HkyVu+giSfkz93L9V1FPIWbQVSFzpLXUB6P7DtUyn5OrLL20uV7Brmfb4fOC1y5MiR0X7iiSeSdi+88AJECrvH/Rjh65ondXlYOvHpqrvuumu0eS77qIYAACAASURBVOx4CYCPwbJ5WfDXl8ccX1OuPg7UPm5ZYvDlH3ju5jFXFG5QJNPklRfx0jOfnz/3ot+C1mSTTTbJXWiVrxuXgfDSFN/b3A/+N5O3/e8MX28+npcB+Xr7xcLz4PnYlwrg8y1azJv3FY1tfl9TF+yWp0cIIYQQpUAPPUIIIYQoBS0ub3lYZvBuLKZIVmBXGLvjfIYOt2PbSyJ8TuwG9C7SvIrMXorh95VF3uIquH4xuSVLlkTbVxvlCpvcL5y1A6T9wjKYv4c4S4Hf493f7KLnc/DnXrS4YlngzEegeBFWvveLMmh4H7vlvfScJ2t61zZLCE8++WTu53ZUvLzF14rvdT/+uL/8NeXxyOPAyy98jLy5uWifv594DLNUOWTIkKQdSydeQiqqBt6avP/++7nnwvc2X0N/3Xi+43HjvzPLUUWLPvvxzLCUlPf7WW27luPx77P/jk899VS0/T3KWV/8HX3oTK3I0yOEEEKIUqCHHiGEEEKUAj30CCGEEKIUtHhMT1FKXF4qH5DGVhTF1vDxilLMOUbIVwDl9xUdj8mrSAmk2qVP7+yo7L333tH+6Ec/muwbN25ctOfOnZvsY22YNdqi68b3g+8j3i6KAeFYgQMOOCDaZ511VtKOv1dZ8amrfI3z4tyA9Sv8MnmxP16n5/HI90RR9eD77rsv93M7KkUVmXmu8vGHfB19HB33Cx+ve/fuSTuu+MxzoY+x5FIHfL4+Rmjo0KHR5ngOfzyOF/FxmkVxZ61JUUxPUZxpLfgxwL+nPtYxLwbHv15U4b4W/HzAx+djc78C6Zzi7y/uS74nmxq3JU+PEEIIIUqBHnqEEEIIUQpaXN7yrnF2fxUtMsYu2aLqmkVpdexO48XqfFosu2SL3PV8fE7RHjRoUNKOU6D98eqlUmhzs8suu0T7O9/5Tm47v/geu3XZjenvG5Y52MXp5VPezkuXBICVK1dGm6t1f/7zn88997LiFxssWlSSrzkvHOlLNfAY5PFXtMgv96GXxlki6dmzZ5VvUS64j/Iq2APpdXv++eeTfdxHRVV3+fhFKescslBUuoPniFrTkv33KqoA3ppsttlm66VgN5A3joq+S9GcxnOf76M8/DHyynwUlYNhav2t9r+L/P29bMffn8ufFC1SXoQ8PUIIIYQoBXroEUIIIUQp0EOPEEIIIUpBi8f0eM2fdUdOM/UUabKshRataM7bHANQdE5FS0jwdlG6XFH8UFEJ8PZM3nIBQHrti/qV+7KoTEFe+Xa/j+MBfLuicgRMrSUMOjJF96yPreG0Z44F6t+/f9Ju2bJl0eb0VU5r9sfgce8/9+WXX452rSXyOxJ+ruISDZwCXHQPP/PMM8k2X/uiWBImb24uoijeq+h8eWz6khRFsZmtSefOnTF48ODcfQ3wtSqKxyla2oWpNYXfz4v1EHPql/7hEgZf/epXo/2Nb3yjSceXp0cIIYQQpUAPPUIIIYQoBS0ub7344ovJNksORRWZ2bW64447JvvyXHzencouTq4aWlT9skhiKVrRnWG3pU+9rpfVf5sblhSKXNK+UmjeqvbePc3lDfg9RasrF61cXGu648ZWKO0IcHkGIL2nfTVd7t+lS5dG248lHt9FFcz5+NzOS1g85lg6mz9/ftKOXeUdCT+uOO13u+22q+kY48ePT7a5in3eSuoe7gffjsc0S5q+EjTvK/oslu38vMohBm1J165dse+++wKoXXKbPn16sj169Oho81j0ciRXli+rFF8LmtGFEEIIUQr00COEEEKIUtDi8ha7uIHUre0rvTL7779/tH2FX5Y62O3qFxLliPW8TC5/DJbfvMuUswx8NgrDlWi5CikATJ48Odqf/OQnc4/RnvHyI7tavdzH16pI3uKMjnnz5kXbZ35w5gO7yf05FVUDFyljxoxJtrkqqh8jPH5YUvYyxcKFC6PN/eRlMJbH//rXv1Y9ByB/scUddtgBZcCHETzwwAPR9plNefiFHv12vTF79uxo870GpBlAvOBxe4DlLE+fPn2q2qJ25OkRQgghRCnQQ48QQgghSoEeeoQQQghRClo8pufiiy9OtjnlrlevXrnv48qsDz30ULLv6quvjvby5cuj7Vfk5SqiHD/kU9Y5nZnTJ30lzZNOOinaPlWXOeOMM6L997//Pdk3cuTI3Pd1FIrSJQ899NBkm2OeeHV67jsg1eg5dst/Fvcl3w++MvawYcNyz5EpY3Vfz4ABA5LtJ554Ito8/oA07oZTaE855ZSkHa+kPXPmzGj7+AuO+brzzjujvcceeyTtjj766Ggfe+yxVb5Fx2bIkCHJNqfm+5TwPGpNqS5qVzReilbZbsrxjjrqqGj70gQjRoyo6fiifMjTI4QQQohSoIceIYQQQpQCa8zCbGa2HMDzLXc6ogoDQgj5OmATUV+2GerPjoP6smPR7P2pvmwzcvuyUQ89QgghhBDtFclbQgghhCgFeugRQgghRClo1w89Zna+mc01s1lmNsPM9jWzhWbWs0rbj5rZt3KOc4iZ7V9tn2g9zKy3md1gZgvMbJqZ3WVmuzXyGD3M7JyWOkeRj5mtzcbhXDObaWZfM7N2PceUAeq3OWZ2k5l12UD7+81sbGZXnW9F/UP9PtPMppflN7DdTkhmth+AYwGMDiHsCeAIAC/ktQ8h3BZCuLTKcTYFcAiAUnR4vWKVghy3ALg/hLBLCGEMgP8C0NjFk3oA0ENP2/BWCGFUCGF3AB8CcBSA7/pG2ZgT9UNDv40A8A6Az7f1CQGVOUEPzS1KQ7+PRGWu/UFbn1Br0J5vqB0BrAghrAGAEMKKEEJD5cMvZ0+us81sKACY2VlmdmVm/9bMrjKzKQBuRGWQn5s99R7UBt9FAIcCeDeEcFXDCyGEmQD+aWaXZX+FzjazUwDAzLqZ2X3Uz8dlb7sUwC5ZX17W+l9DAEAIYRmAswF8KfvxOsvMbjOzyQDuM7OuZjbezB41s8cb+s/Mds9em5F5cAdnbe/M/iKd03APiBbhAQC7Zt7vOxpeNLMrzeysojea2Vez/pljZl/JXrvUzL5IbS40s69n9jfM7LGsn/87e22gmT1pZtcCmAOgX/N/RVGFrQG8ChTOrTCzC7L++aeZXd/Ql+2J9vwX10QA/8/MngLwVwB/DCE0lD9eEUIYnckcXwfwmSrv3wnA/iGEtWZ2IYDVIYQft8aJi6qMADCtyusfAzAKwEgAPQE8Zmb/ALAcwPEhhNcz9/ojZnYbgG8BGBFCGNVK5y1yCCE8a2adAGyfvTQawJ4hhFfM7BIAk0MInzKzHgAeNbO/ovIHyM9CCH8ws80BdAJwNIDFIYRjAMDM6nv573ZK5oE7CsA9TXjvGACfBLAvAAMwxcz+DuCPAC4H8L9Z05MBHGlm4wAMBrBP1v42MzsYwKLs9TNDCI9s3DcSG2BLM5sBYAtUnAiHZa+/jepz61gAJ6AyF28GYDqqz9l1Tbv19IQQVgMYg8pfk8sB/JH+Erk5+38agIE5h7gphLA2Z5+oHw4EcH0IYW0IYSmAvwPYG5WJ8hIzm4XKQ29fNF4KE63LpBBCw3oi4wB8K5t070dl4u0P4GEA3zaz81CptfEWgNkAPmRmPzSzg0IIq9rg3DsyDT9+U1F56Pi/JhzjQAC3hBD+lc3NNwM4KITwOIDtzayPmY0E8GoI4QVU+n8cgMdR+fEcisrDDgA8rweeVqFB3hoK4MMArs3CDPLm1gMA3BpCeDuE8AaA29vqxDeG9uzpQfbQcj+A+81sNoAzs10Niy6tRf53/FfO66JtmAvgxEa0Px1ALwBjQgjvmtlCVH44RZ1gZoNQGYPLspd4zBmAE0IIT7q3PZHJzscAuMvMPhdCmGxmo1Hx+HzfzO4LIVzU0udfIt7ynlEzew/pH8UbM7ZuQmVs90bF8wNU+v8HIYSruaGZDYTm5lYnhPBw5tXphco467Bza7v19JjZEDPjFUFHoemVL98AsNXGn5XYCCYD6GxmZze8YGZ7AngNwClm1snMegE4GMCjALoDWJYNykMBNKyKqb6sA7K+ugrAlaF6BdR7UYm9s6z9Xtn/gwA8G0K4AsCtAPY0sz4A3gwh/B7AZajIZKJleR7AcDPrnMmPh2+g/QMA/s3MuphZVwDHZ68BlQedU1F58Lkpe+1eAJ8ys24AYGZ9zWx7iDbBKrGvnQCsRP7c+iCAj5jZFlm/tcuVfduzp6cbgJ9nA/I9AM+gInU1pSNuB/CnLGDryyGEBzb0BtG8hBCCmR0P4PJM2ngbwEIAX0Glr2cCCAC+GUJYYmZ/AHB75uGbCmB+dpyVZvagmc0BcHcI4Rtt8HXKSoNMshkqY/J3AP4np+33UIn1mGWVDJ3nUBm7JwP4uJm9C2AJgEtQkTMvM7P3AbwL4Ast+i0EQggvmNmNqAQTP4eKDFXUfrqZ/RaVP0gA4NeZtIUQwlwz2wrASyGEl7PXJprZMAAPZ8+9qwGcgYpnULQODeMVqHjezsxiXPPm1sey2J5ZAJaiIju3O6lZy1AIIYQQYoOYWbcQwmqr1HL6B4CzQwjT2/q8GkN79vQIIYQQovX4lZkNRyXGZ0J7e+AB5OkRQgghRElot4HMQgghhBCNQQ89QgghhCgFeugRQgghRCnQQ48QQgghSkGjsrd69uwZBg4c2EKnIqqxcOFCrFixwpr7uPXSl++//360X3311Wi/8cYbSbvOnTtH+9133432e++9l7TbbLPNor3FFuuKiPbo0SO3XWsybdq0FSGEXs193HrpT+4P7sM1a9Yk7bLaLACAd955J9q+X7beeutob7nlllXf31Z09LFZNlpibKov24aivmzUQ8/AgQMxderU5jkrURNjx45tkePWS1++9dZb0f7Tn/4U7cmTJyftdt5552gvW7Ys2suXL0/a7bjjjtEeMmRItI877rikXZ8+fZp4xhuHmTW1angh9dKfK1asiPbf/va3aD/77LNJu8033zzazz+/7pL07ds3afehD30o2rvvvnu0ix5aOSO1JR+OOvrYLBstMTbVl21DUV9K3hJCCCFEKVBxQtGqsDQFAPvuu2+0jzjiiGhvtVW6fNakSZOivXLlymh72WrkyJHRfvrpp6N97rnnJu3Gjx8f7a5du9Z07mWCZcdNNkn/Nlq0aFG0jzzyyGTf/Pnzo929e/doe88Me4S23XbbaL/55ptJu29+85tVz+/UU09Ntq+//vpos3fH1yGrB1lMiJbmwgsvTLYvueSSaO+yyy7Rfu2115J2PF5Wr14d7VNOOSVpd80110Sbx+I999yTtFuyZEm0u3TpUsuptzjy9AghhBCiFOihRwghhBClQA89QgghhCgFiukRrYqP7Tj44IOjvemm627H008/PWnHqctr166NNmvGALD//vtH+84774w2p68DiuPZED6Oh/nABz4Q7U9/+tPJPs6eO++886LN6eYA0Lt372hznA3HEgFp1t11110X7U6dOiXtLrvssmjvtNNO0VZMjygjEydOTLZ5vuOMWT8vckxdv379ov3nP/85aXfttddGu1evdZnhfn7n+XnQoEE1nXtLI0+PEEIIIUqBHnqEEEIIUQokb4k2havx7rbbbtF+7rnnknYsS0yZMiXaQ4cOTdqxPMIp05y+LjaOPfbYI9pcSgAAxo0bF22WlrzMxMUJuYwBu96B9P4444wzos2udwB46KGHon3yySdH28t0rVW4UKwvVeZJpv/xH/+RbF9xxRUtdk5lwV/rbt26RZslLd9HPC55LPL7gVTG5jH11FNPJe24NIXkLSGEEEKIVkQPPUIIIYQoBZK3RJvCWTjsWp03b17SbsGCBdHmas3ejcvr3HC15rZaYLQjsOeeeybbL7/8crTffvvtZB9Lj+xG95WW86QOn+V1yy23RJvvD+9u54qxM2bMiDZXogUkb7UmXtJkpk2bFu2bb7452ceS9TnnnBNtztoE1s/gE+tgWQlIFwJmOZnlYyC9xjx+/XjlxaGLKi2/8MIL0d5nn302dNqtgjw9QgghhCgFeugRQgghRCnQQ48QQgghSoFiekSb8uSTT0a7c+fOVV8H0tWAhw8fHu2//OUvSTte9Zs1f78auyiGVy336f5FFVg51objCNgG0hiBf/3rX9HmqtxAfmqsj8fZfvvtoz158mTkUVRpWjQvRTFTnA7t7yFewfuII46INpe0EMWsWrUq2eYxxrE6frwxHN/j47PyjuHb8Urt9YJmACGEEEKUAj30CCGEEKIUtHt5qykpqLVWCvXwomsnnHBCs55TWRk8eHC0WVLxblLuI05f92mrnOLMktZFF1208SdbIn73u99Fm0sEAKkcwemvHpYrvRud+5Nd5ZyWDqQudv4sPjaQ3geLFi2Ktk/d7dmzZ+75io2nKE2d4T73qeicDn3ggQdG+7TTTkva/exnP2vKKZYC/xvH44N/k/y153HJx/C/kXn7/O+dL2lRD8jTI4QQQohSoIceIYQQQpSCdi9v1Sofsdu1SM7idp/+9KeTfU888US0L7300mg/9thjTTonxrsjy8JXvvKVaE+YMCHaW2+9ddIu7/p49yzLKCtXroz2wIEDN+Y0Swcv+OolRO4LlimANEurSOqoVVLmdrW+hzNGHnnkkWTfscceW9MxRNPgua9oHvzhD38Y7TVr1iT7tt1222izlMrzAwDceOON0fYZhr5idwN+Humo4QdFvye1hl/wMYrCDYqoVe5sTeTpEUIIIUQp0EOPEEIIIUqBHnqEEEIIUQrafUxPrdSqY37iE5+INmvLANC/f/9o82reX/va15J2P/rRj6Jd60rAZa0UO2zYsGhzirnX+Vnb52vlY3oYTn/m1FexYXh15G222SbZx3E7PhWdx1ZemqzfZtuPAx63Pp0975z4GHfeeWfSTjE9LUvRPDtv3rxoz5kzJ9o+/oarNfM91KdPn6Qdp7b7sgoTJ06M9oc+9KFol2We9RXQGe4jXw2by0Lw79/SpUuTdltuuWW0eQ7wfV50Hm1FOe4AIYQQQpQePfQIIYQQohSUUt7ysCt/yZIl0eZF2gBg2bJl0ebF76ZMmZK069q1a7RZvgGAf//3f482uw+96/aoo47KPd+OSpcuXaLNLm4gv1JokTvVp1OLYvi68r3PC4wC6WKGfiHXouqstVBriqtvx9ssiXDqvWh5ivqcKyh379492jznAqmMyXKLr+7LkpYvcTFu3Lho/9///V+0P/WpT+WeX0eGrymHBPj+4rHTr1+/aHOVcyCVt/g9flxK3hJCCCGEaCP00COEEEKIUqCHHiGEEEKUgtLE9LCO6dPIOSaHdWO/IjOXt2et0uuiHMfTt2/fZN+sWbOi/frrr1f9XADYeeedq77ekWH93qeWcv8VpZ0WrSAsivGxFQ34+7soLZm3W7IEvR/D77zzTtV2y5cvb7FzEBWKVuNmHnrooWi/+eab0faxdzwPFB37jTfeiLZPWd91112jffHFF0f7oosuStqdcsopuefbnuncuXOyzeOS43teeeWVpF3v3r2jve+++0b7rrvuStrttNNO0S4qy8KxP/WCPD1CCCGEKAV66BFCCCFEKejQ8lat1VxZ3mK3qIdX6ubqot69x7LK9ttvn+xjyYwrkXJ1UQAYNGgQgPXdlB0NrvTJqdDeLZpXcdenRHKFUT6eLyvArltRgUs3MC29EjWP06LVnGutqs4VYvO+k2g+8vrC9+Vrr70WbZ77vAydV36A5Swgvyqw38flF1hWA4CxY8dWPff2jv/d4O/NpUF8KjpfRy6bcsEFFyTtuI+KVnSvx7Ih8vQIIYQQohTooUcIIYQQpaBDy1t5btdHH3002eaKnUcffXS0vTt16NCh0Wa3HUtdQJrlNXv27GTfdtttF23OUvDuyAY5rqWlhbZmwoQJ0ebrxlWtgXyZw2d08PVit+vVV1+dtJO8tT552VtFWVhe2mW5sdbsLT5GUfZP0fG4r1nKXrx4cU3nIJpOnuzo5z6WmVjeKrqHeA72C5PyHOHnapa5WdrZa6+9knYnnXQSOiI+m43HB0t//veFVwkYPXp07vFZkvSLljJcebtekKdHCCGEEKVADz1CCCGEKAV66BFCCCFEKWh0TE+DftseYk24Suudd94Z7QULFiTtzj///GhPmjQp2pyiDqTpr6xp+nYcq+MrwnKFZU4PbMnqtfXMlVdeGW2Okyq6HrXGdnBq5m9+85uk3fjx4xt1nmWA4yJ4fBdVdy2KqSqqzszvY7so/bVoNee8uKCyjqtaKUo9Lqq0Wwt33313ss3zMY91juUD0vumR48euefHMUL+PuRtno9XrFhR07m3d3juA9LyHfy740t+7LjjjtGu9Te+qB3HCNUL8vQIIYQQohTooUcIIYQQpaDR8la9yVrs4rzhhhuSfSxpsRy1yy67JO3YjctS1dNPP520W7NmTbS33nrraLP71B9/5cqVyb68qsP+GB0Vv4BqXhqrlyXyJBXv8ub3sSzDbnIAeP7556M9YMCAms69o8N9U6u0UVRNt2iuyKvC7D83T54qSm33LnuGU9j79OmT266MNFXOyusLLyn369cv2lyd2ac8s/zCx37rrbeSdkWLkXKaOo/9Z555puq5djS8vMXjlK+jH195cpQfy/x75UsJMH5FgnpAnh4hhBBClAI99AghhBCiFDS5IrN3i7GrsSlR30Xv8ZVi//SnP0V7+vTp0fau9iFDhkSbF7Z8/PHHk3YsM7Gb9KWXXkrasUuPFw718LXo1atXsm+HHXaINmczPPfcc0m7Bqmho2WfcHYckGYL+OudB7tn/YJ2efeUX0iWpc9zzjmnps/t6PCityx1+HFVNG55m9/nj8HjjPcVZeGwFLrNNtvkfItimWbOnDnRlrxVPO9yv9QqfY0ZMybaXnrmCso8r/E8CKRyl9/H8Dn5yr9837Bsy/NvR8ZLTtwXRfLvTjvtVPV1f32LquczRfvaCnl6hBBCCFEK9NAjhBBCiFKghx4hhBBClIImx/R4LXhjq3d6fvnLX0abV+IG0vgMTrHzMSEct1F0fq+//nq0OW7Hp++9/PLL0eb4Bx+HwFU/ecVnIE11HzRoUNXPBdaluhfpr+2RBx98MNnm63HggQdG+x//+EfSjvuIV7vn1HMg1ez33nvvaPsq3PPnz2/MaZcCvsZFcTtcusHHXPD9WlQZOS9N3X8Wx2PwWPKlD3i7KIV22bJluftESt6ceeuttybbp556arQ5RbkoZorj8nzFb76H2OZUdiCNOeJ4L9+W71cfA9hRKYrpKYoT3W233aq+7mNzeGwX/UYppkcIIYQQoo3QQ48QQgghSkGT5S1P3kKBXIEYSCWoF198MdpezuA09VGjRiX72F3JKajeNc6VODkNkt8PpG5SdsexhAWk6dVcdXnmzJlJO164ceedd0728fHZJctp83xOHS1l/amnnkq2Fy1aFO3DDjss2ixNAakc1bt372j7+27kyJHR5rTmJ554ImnHfSkqsLzF48WPK5YkvQSSJ1sUycs8/opkc3aj8zkAtafk+sUt2xt5C4QWpZ5zOy8RFXHvvfdG+0tf+lK0WeYHgLFjx1Y9vq+gnHd/+fCAokrLDM+ZXu5k+D7kubkj41PMeYwVjQ9fYqUBP964ojYfz/9eFUnNbYU8PUIIIYQoBXroEUIIIUQpaLS81eAmu/DCC5PX2W3IbkcvP3CWErvFfDuWprxrPK9Kp3fd8vv4c4tcwSzHeRchy2zs4h08eHDSbsSIEdH27j6WtHgfZ3wB61zDebJhe8VndLDUuO+++0bbu2DZzT137txo+0rL7E494IADon3NNdck7ThzThTj5eC8xUKB2rM4uT+Lsrf8YpR555SXKeYzMJcvX17T+bU1ebI2X5+mXGsPz0eHHHJIsm/q1KlV93GleyBd3JMlLT9Pc/Yd95+X9jnDit/jM1z5s/x9k5f1xbJMR8ZnTW1smIT/LeTr2N5CMOTpEUIIIUQp0EOPEEIIIUqBHnqEEEIIUQoaHdPToJ2eeeaZyeu82vmTTz4ZbV8BNS+1sCi90evwrGXzZ/Xt2zdp98wzz0Q7L5YIAAYMGBBtXu2cY3OANIWTNU2uEAwUp8Vy+iRr2T4eqUFDbe5K123NRz7ykWT7L3/5S7S5urK/pvfff3+0+Rr6VEquKMolEXxslNeoRXr9OXXVXzu+932F27yUaL/Keq2xankrsPt4EX/8vPN56KGHavrctqYo7rAWGiq6A8C0adOifc899yTtrr/++mhzKQgAOPvss6P97LPPRvuVV15J2uXFDPlKy9xH3H++mnJeyro/Hs+fvvI9z/f8uX5u5tIoHQk/vxVVWGf8NW5gq622SrZriTkD1u+XekCeHiGEEEKUAj30CCGEEKIUNEreWrt2bZSdvPvs5JNPrukYnBLOLliWs4DUDeklsbyqkT6lkSvysjvOu8L5fXkVYP25s7vPS1CcZuvde5xS3aVLl2h7t2CD3Lexbu564wMf+ECyvc8++0R7/Pjx0T733HOTdpyCyemunC4LADvttFO0L7jggmgffPDBSbuvfe1rjTntUsDXcvHixdH2448XdfWp4zwWWJrwchaPLR6PeSnq/hheUmF5vCgteeDAgbn76pFbbrkl2eaFmLnMBc+lQL582LNnz2SbrwcvFgoA8+bNi/Z2220XbV/VmOdWlq38/Jl3Tl4qyZPL/FzI0qqXMXlOL0qp7qgp7P43slY5OU/eKlrAlPu5PSzoKk+PEEIIIUqBHnqEEEIIUQoaJW9tsskmUZLhTA+/zS4un13DkhO7VouqhnJGFZC6V1k+8sfgduyC8+3yFmPzMhi7UFmO87IaywR5WSUe365hQUwvH7R3Fi5cmGxz1t8JJ5wQ7QULFuS2+9jHPhbtBx98MGnHbvjTTz892rfddlvS7u677472UUcdVcupd3gmTpwY7auvvjraJ510UtLui3TU3QAACv5JREFUM5/5TLRvvfXWZF///v2jXeRSzxsX/j15FYi9HM7ufF4o2MufLCnXK2+++WbMuPrud7+b7GNpka+1z3bMkwJ99WPOjPWZttyWFwr2Va5ZWsyTuoD8zF0vg/E8y3OrPx7PjV7C4vdxn/u5v6OFDzTA8yCQv1CtJ0/e8uMm73h5768n5OkRQgghRCnQQ48QQgghSoEeeoQQQghRChoV02NmMYbGryzOGh/H0rC+DqRpli+99FK0i3RdXwGUt1l39np13urNRSnm/Lle7+XvyHFLvh3HF/jPyqsi6o/RsBq5Xy23vTN8+PBkmysoc9yAv79OO+20aHOcRtGKzxwH5OOu/L0iUj73uc/l7uP4vR49euS2y6umDKTjzI99hvfxOPWpsS+//HLV97eHGB7P6tWr8fDDDwNYvxo9x2pwurWvaszXimNpfFwM94OP9eC5m/tv6dKlSbu8kh++zzneI69ytz8Pfo8/v6K0dG7L92FZxr2Pu+Lfl6LxlrcCgO+vvFio9lDpXp4eIYQQQpQCPfQIIYQQohQ0esHRPNjdxWnqPmVdlBtfcZdlCnbX+9R2rgjKbnefmsnSKsugvhKtr9Asiise57Urkhx4X1NTg/l+4RRl74avtbRDrd+xLenevTuOPvpoAMANN9yQ7Js7d260iyoSs3TO+3xKcVHJD26bZ3u4X3wf8Xmw5OSlKd7Os/35+nNnuY/vvX79+iXtHn/88Srfov3j5S3u56IK1XkV0f31zVvhgEsq1Cv1OeqFEEIIIZoZPfQIIYQQohTooUcIIYQQpaDZYnqEaAqsPa9evTraXNoeAGbMmBHt448/PtqTJk1K2nG5gK222iraixYtStrVazxHW1LrNeGV7H3sVd7K6n7ZiaJYDSav1IR/z6BBgzZw1hv+rHqhc+fO8fv885//TPbxsh8TJkyIdkOKewPcL83xnbkfilY0r0fyymIA65fG6Cjwck9AfqwOj2UgnTMZH8fF9wOPc78aez2imV8IIYQQpUAPPUIIIYQoBZK3RJsyduzYaM+fPz/a3mU+atSoaHMquq/IzJV5WVI58sgjN/5kBYC0qrqnVjmqKFWa4eqx/B6/YveSJUuizWULvLu+PchbRRx33HFV7SIWL14cba6CD6TjZcGCBck+7suikgNclZsrHnsZjCUW7suiFPiiz+X7y8s3vM3tfOmKvn375h6/PcMyP5D2BYcO+ArdfruWdixv5clo9YQ8PUIIIYQoBXroEUIIIUQpkLwl2hSunMqucb+g5IgRI6JdayVWlreKFjgU61NUuZgzXnz21qpVq6LNLnCfjVcreVWHvWx1+OGH5+5j8hZU7Mg0LF7sbdFx8fc5LxLMoQKcFQvkZ1/5lRX69+8fbQ5FGDZsWONPtpWRp0cIIYQQpUAPPUIIIYQoBXroEUIIIUQpUKCDaFN23333aL/wwgvR5urMADBlypRoc6xI165dk3acksoxID61XRRTlEZ+7bXX5u6bM2dOtB955JFoc0wBkKacc8qyT3nlfaNHj472uHHjcs+B8SnqTV3tXYj2BMdHAmmMXlPGwE9/+tONPqd6QZ4eIYQQQpQCPfQIIYQQohRYYyqUmtlyAM9vsKFoTgaEEHptuFnjUF+2GerPjoP6smPR7P2pvmwzcvuyUQ89QgghhBDtFclbQgghhCgFeugRQgghRCmo64ceMzvfzOaa2Swzm2Fm+zbjsQ8xszua63iiaZjZ2qxv55rZTDP7mpnV9X0pasPMepvZDWa2wMymmdldZrZbI4/Rw8zOaalzFJpnOxpmtl3WjzPMbImZvUTbm2/4CB2buq3TY2b7ATgWwOgQwhoz6wmgLjrMzDYNIbzX1ufRQXgrhDAKAMxsewDXAdgawHe5ka55+8IqxUBuATAhhHBq9tpIADsAeKoRh+oB4BwAv2j2kxSaZzsgIYSVABrm1AsBrA4h/Lhhf2tfVzPrFEJYu+GWrUM9/0W9I4AVIYQ1ABBCWBFCWGxmC83sv81supnNNrOhAGBmXc1svJk9amaPm9lx2esDzeyBrP10M9vff5CZ7Z29ZxczG2Nmf8/+Mr3XzHbM2txvZpeb2VQA/9l6l6E8hBCWATgbwJeswllmdpuZTQZwX0Ef7569NiP7a3Vw1vbOzHs0x8xOadMvVz4OBfBuCOGqhhdCCDMB/NPMLsv6ZHZDv5hZNzO7j8b1cdnbLgWwS9a3l7X+1+jwaJ4tAWb2WzO7ysymAPiRmY0ys0ey+fIWM9sma3e/mY3N7J5mtjCz15tjs9fPoNevNrNO2eurzewnZjYTwH5t8qXzCCHU5T8A3QDMQOWvwl8A+GD2+kIAX87scwD8OrMvAXBGZvfI3tcVQBcAW2SvDwYwNbMPAXAHgP0BTAPQH8BmAB4C0CtrcwqA8Zl9P4BftPV16Wj/UPkrxL/2GioegbMAvAhg2w308c8BnJ69vjmALQGcAOAaOmb3tv6uZfoH4D8A/LTK6ycAmASgU9bHi1D54d0UwNZZm54AngFgAAYCmNPW36ej/tM827H/AbgQwNcB/Dbrh07Z67Oory8CcDld/7GZ3RPAwsyuNscOA3A7gM2y138B4BOZHQCc3Nbfv9q/upW3QgirzWwMgINQ+avxj2b2rWz3zdn/0wB8LLPHAfiomX09294ClQG2GMCVZjYKwFoAHFMwDMCvAIwLlb9uRgAYAWBSxTuPTgBepvZ/bMavKGpjUgjhlczO6+OHAZxvZjsBuDmE8LSZzQbwEzP7IYA7QggPtPqZi2ocCOD6UHF3LzWzvwPYG8DdAC4xs4MBvA+gLyoPRaIF0TxbKm4KIaw1s+4AeoQQ/p69PgHATRt4b7U59nAAYwA8lvXjlgCWZe3XAvhzs3+DZqBuH3oAIJsY7wdwf/Yjdma2a032/1qs+w4G4IQQwpN8DKtomksBjERFznubdr+MyqDdC5VBawDmhhDy3HH/2oivI2rAzAah0q8Ng4evedU+BvBE5rY9BsBdZva5EMJkMxsN4GgA3zez+0IIF7X0+YvIXAAnNqL96QB6ARgTQng3c6tv0RInJlI0z5aGWq7re1gX9hLHXwjhOj/HotKPE0II/1XlOG+HOorjYeo2psfMhjTohhmjUFzZ8l4AX7bskdPM9spe7w7g5RDC+wA+jspfFQ28hkon/sDMDgHwJIBeVgnug5ltZma7Q7QKZtYLwFUArgyZj9RRtY+zB6VnQwhXALgVwJ5m1gfAmyGE3wO4DMDoKscTLcdkAJ3N7OyGF8xsT1TG3Clm1inr74MBPIrKOF2WPfAcCmBA9rY3AGzVuqdeHjTPlo8QwioAr5rZQdlLHwfQ4PVZiIr3BqA/WqrNsQDuA3CiVRJQYGbbmlnDuK1b6vahBxWteYKZzTOzWQCGo6JP5vE9VLTiWWY2N9sGKjrjmVlA1VC4p90QwlJUshf+F5W/RE4E8MOs/QxUtGjRcmyZBcHNBfBXABMB/HdO27w+PhnAHDObgYrb/FoAewB4NHvtuwC+34LfQTiyh9bjARxhlZT1uQB+gEp23iwAM1F5MPpmCGEJgD8AGJt5Gj4BYH52nJUAHrRK4LMCmZsfzbPl5EwAl2V9PgqVuB4A+DGAL5jZ46jE9DSw3hwbQpgH4DsAJmbHmYRKfF5do2UohBBCCFEK6tnTI4QQQgjRbOihRwghhBClQA89QgghhCgFeugRQgghRCnQQ48QQgghSoEeeoQQQghRCvTQI4QQQohSoIceIYQQQpSC/w9DNZR+UBn4ggAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 720x720 with 25 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H84pfqDyxTDC"
      },
      "source": [
        "# 모델 설정\n",
        "## Sequential API 사용\n",
        "- 가장 간단한 케라스의 신경망 모델"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z4GGZFMhwgIU"
      },
      "source": [
        "model = keras.models.Sequential()\n",
        "model.add(keras.layers.Flatten(input_shape=[28, 28]))\n",
        "model.add(keras.layers.Dense(300, activation='relu'))\n",
        "model.add(keras.layers.Dense(100, activation='relu'))\n",
        "model.add(keras.layers.Dense(10, activation='softmax'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rov224kZn6cQ"
      },
      "source": [
        "- Flatten: 2차원 배열의 이미지를 28 * 28 = 784 픽셀의 1차원 이미지로 변환\n",
        "    - (input).reshape(-1, 28*28)을 계산\n",
        "    - 전처리 수행 작업\n",
        "    - batch_size를 제외한 샘플의 이미지만을 입력\n",
        "- Dense(완전 연결층) 은닉층 구성\n",
        "    - 활성화 함수: _Relu_\n",
        "    - 뉴런 개수\n",
        "        - 은닉층_1: 300개\n",
        "        - 은닉층_2: 100개\n",
        "- Dense 출력층 구성\n",
        "    - 뉴런 개수: 10개\n",
        "    - 활성화 함수: _Softmax_\n",
        "        - why? 클래스들이 배타적 이므로...\n",
        "\n",
        "> `activation='relu'` == `activation = keras.activation.relu`\n",
        "활성화 함수에 대한 전체 리스트는 [Link](http://keras.io/activations) 참고"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "T0_AwBqzojho"
      },
      "source": [
        "## 모델 층 구성 정보 확인"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zA3x-n2WxDVj",
        "outputId": "aba3ea7d-4f98-4ddd-b6b8-016218d7004e"
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "flatten (Flatten)            (None, 784)               0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 300)               235500    \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 100)               30100     \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 10)                1010      \n",
            "=================================================================\n",
            "Total params: 266,610\n",
            "Trainable params: 266,610\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NkMBBiPgyIiY",
        "outputId": "2619ee35-4533-45d1-e5bc-fcbd798a8fc6"
      },
      "source": [
        "model.layers"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<tensorflow.python.keras.layers.core.Flatten at 0x7f103d72d690>,\n",
              " <tensorflow.python.keras.layers.core.Dense at 0x7f103e8c9510>,\n",
              " <tensorflow.python.keras.layers.core.Dense at 0x7f1039a1e590>,\n",
              " <tensorflow.python.keras.layers.core.Dense at 0x7f1039976390>]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xLNh_NhLyfB5",
        "outputId": "345fd9b2-c2e9-413f-b69e-e84e67e1c682"
      },
      "source": [
        "type(model.layers)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "list"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 38
        },
        "id": "QB5WmrXvySG9",
        "outputId": "2cb07698-e41e-4727-941d-71f73a854e67"
      },
      "source": [
        "hidden1 = model.layers[1]\n",
        "hidden1.name"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'dense'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VmeZ-JT7vgAU"
      },
      "source": [
        "### 층의 파라미터 확인 방법\n",
        "- `get_weight()` 사용"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tSqYqbCWvt26",
        "outputId": "f4d5ef24-4600-4f10-bc3d-1a6cbca567c7"
      },
      "source": [
        "weights, biases = hidden1.get_weights()\n",
        "weights"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[ 0.03770764, -0.05693615, -0.04917307, ..., -0.05639156,\n",
              "         0.0594293 ,  0.03320393],\n",
              "       [ 0.03421903, -0.01009894,  0.03445695, ..., -0.05623742,\n",
              "        -0.02886158,  0.05174698],\n",
              "       [ 0.02896158,  0.04800321, -0.0506296 , ..., -0.06521712,\n",
              "        -0.02440413,  0.03367826],\n",
              "       ...,\n",
              "       [ 0.04398489, -0.06262626,  0.00150014, ...,  0.02040758,\n",
              "        -0.07016663, -0.01243699],\n",
              "       [-0.06560042,  0.03983253,  0.07334235, ..., -0.05007394,\n",
              "        -0.01552234,  0.04555964],\n",
              "       [ 0.02411408, -0.0115964 , -0.04629562, ...,  0.01302569,\n",
              "        -0.03757626,  0.06807001]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RLMdw1d7wU-y",
        "outputId": "79740a1c-3abb-4de3-b05c-438190525b27"
      },
      "source": [
        "weights.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(784, 300)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7UROHHxOwXFk",
        "outputId": "2ed179ca-fdac-4684-d9ab-61d648c37e55"
      },
      "source": [
        "biases"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UbEE9g9awaw9",
        "outputId": "aa70fa7d-f060-4ecf-bb91-17f8dc3276d6"
      },
      "source": [
        "biases.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(300,)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BvqPoduOwtgs"
      },
      "source": [
        "- 가중치 행렬의 크기(784, )는 입력에 크기(28*28)에 달려있음\n",
        "- Dense 층은\n",
        "    - 연결 가중치(weights or kernel)를 무작위로 초기화\n",
        "    - 편향(bias)는 0으로 초기화\n",
        "    - 다른 초기화 방법을 사용하고 싶으면\n",
        "        - `kernel_initializer`와 `bias_initializer` 매개변수 설정할 수 있음\n",
        "        - 매개변수 리스트는 [link](https://keras.io/initializers)를 참고"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8uKaKCIsxQFl"
      },
      "source": [
        "## 모델 컴파일\n",
        "- 손실 함수: 훈련하는 동안 모델의 오차를 측정\n",
        "- Optimaizer: 데이터와 손실 함수를 바탕으로 모델의 업데이트 방향 결정\n",
        "- 지표(Metrics): 훈련 단계와 테스트 단계를 모니터링 하기 위해 사용"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qVuHp6BBx7Ri"
      },
      "source": [
        "model.compile(optimizer=keras.optimizers.SGD(learning_rate=0.01),\n",
        "              loss=keras.losses.sparse_categorical_crossentropy,\n",
        "              metrics=[keras.metrics.sparse_categorical_accuracy]\n",
        "              )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xWPndDktt6nz"
      },
      "source": [
        "- `optimizer=keras.optimizers.SGD()` == `optimizer='sgd'`\n",
        "    - SGD, Stochastic Gradient Descent, 확률적 경사 하강법\n",
        "    - 역전파 알고리즘을 수행하여, 하이퍼파라미터 학습\n",
        "    - SGD는 default 값으로 learning_rate=0.01을 사용\n",
        "- `loss=keras.losses.sparse_categorical_crossentropy` == `loss='sparse_categorical_crossentropy`\n",
        "    - 레이블이 정수 하나로 이루어져 있고, 클래스가 배타적이므로 손실함수로 `sparse_categorical_crossentropy`를 사용함\n",
        "    - 만양 샘플마다 클래별로 타깃 확률을 가지고 있다면(e.g. one-hot 벡터), `categorical_crossentropy` 손실함수를 사용해야함\n",
        "- `metrics=[keras.metrics.sparse_categorical_accuracy` == `metrics=['accuracy']`\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XI9I4l9O0lMS"
      },
      "source": [
        "## 모델 훈련\n",
        "- 훈련 데이터를 모델에 로드, train_images, train_labels\n",
        "- 훈련이 진행되는 동안 모델이 이미지와 레이블을 맵핑\n",
        "- vaildation data를 이용하여 훈련 평가 병행하여 진행"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gC0dGbUE3l43",
        "outputId": "aba72d9c-c870-4398-f67b-cc90d0d84af1"
      },
      "source": [
        "histo = model.fit(train_images, train_labels, epochs=30,\n",
        "          validation_data=(valid_images, valid_labels))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "1719/1719 [==============================] - 8s 4ms/step - loss: 0.9918 - sparse_categorical_accuracy: 0.6847 - val_loss: 0.5131 - val_sparse_categorical_accuracy: 0.8298\n",
            "Epoch 2/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.5088 - sparse_categorical_accuracy: 0.8212 - val_loss: 0.4509 - val_sparse_categorical_accuracy: 0.8484\n",
            "Epoch 3/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.4480 - sparse_categorical_accuracy: 0.8419 - val_loss: 0.4215 - val_sparse_categorical_accuracy: 0.8584\n",
            "Epoch 4/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.4210 - sparse_categorical_accuracy: 0.8522 - val_loss: 0.4054 - val_sparse_categorical_accuracy: 0.8598\n",
            "Epoch 5/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3996 - sparse_categorical_accuracy: 0.8599 - val_loss: 0.3795 - val_sparse_categorical_accuracy: 0.8708\n",
            "Epoch 6/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3847 - sparse_categorical_accuracy: 0.8634 - val_loss: 0.3786 - val_sparse_categorical_accuracy: 0.8712\n",
            "Epoch 7/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3701 - sparse_categorical_accuracy: 0.8693 - val_loss: 0.3730 - val_sparse_categorical_accuracy: 0.8670\n",
            "Epoch 8/30\n",
            "1719/1719 [==============================] - 8s 4ms/step - loss: 0.3563 - sparse_categorical_accuracy: 0.8740 - val_loss: 0.3597 - val_sparse_categorical_accuracy: 0.8762\n",
            "Epoch 9/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3439 - sparse_categorical_accuracy: 0.8771 - val_loss: 0.3540 - val_sparse_categorical_accuracy: 0.8776\n",
            "Epoch 10/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3346 - sparse_categorical_accuracy: 0.8797 - val_loss: 0.3698 - val_sparse_categorical_accuracy: 0.8696\n",
            "Epoch 11/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3264 - sparse_categorical_accuracy: 0.8835 - val_loss: 0.3434 - val_sparse_categorical_accuracy: 0.8810\n",
            "Epoch 12/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3157 - sparse_categorical_accuracy: 0.8884 - val_loss: 0.3362 - val_sparse_categorical_accuracy: 0.8778\n",
            "Epoch 13/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3121 - sparse_categorical_accuracy: 0.8888 - val_loss: 0.3337 - val_sparse_categorical_accuracy: 0.8774\n",
            "Epoch 14/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.3055 - sparse_categorical_accuracy: 0.8913 - val_loss: 0.3229 - val_sparse_categorical_accuracy: 0.8838\n",
            "Epoch 15/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.2970 - sparse_categorical_accuracy: 0.8925 - val_loss: 0.3317 - val_sparse_categorical_accuracy: 0.8776\n",
            "Epoch 16/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.2913 - sparse_categorical_accuracy: 0.8953 - val_loss: 0.3344 - val_sparse_categorical_accuracy: 0.8820\n",
            "Epoch 17/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.2884 - sparse_categorical_accuracy: 0.8955 - val_loss: 0.3229 - val_sparse_categorical_accuracy: 0.8850\n",
            "Epoch 18/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.2749 - sparse_categorical_accuracy: 0.9022 - val_loss: 0.3171 - val_sparse_categorical_accuracy: 0.8870\n",
            "Epoch 19/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.2705 - sparse_categorical_accuracy: 0.9023 - val_loss: 0.3285 - val_sparse_categorical_accuracy: 0.8736\n",
            "Epoch 20/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2651 - sparse_categorical_accuracy: 0.9039 - val_loss: 0.3144 - val_sparse_categorical_accuracy: 0.8886\n",
            "Epoch 21/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2642 - sparse_categorical_accuracy: 0.9065 - val_loss: 0.3250 - val_sparse_categorical_accuracy: 0.8852\n",
            "Epoch 22/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2609 - sparse_categorical_accuracy: 0.9054 - val_loss: 0.3141 - val_sparse_categorical_accuracy: 0.8862\n",
            "Epoch 23/30\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.2516 - sparse_categorical_accuracy: 0.9083 - val_loss: 0.3151 - val_sparse_categorical_accuracy: 0.8872\n",
            "Epoch 24/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2523 - sparse_categorical_accuracy: 0.9099 - val_loss: 0.2982 - val_sparse_categorical_accuracy: 0.8950\n",
            "Epoch 25/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2442 - sparse_categorical_accuracy: 0.9122 - val_loss: 0.2998 - val_sparse_categorical_accuracy: 0.8898\n",
            "Epoch 26/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2428 - sparse_categorical_accuracy: 0.9122 - val_loss: 0.3110 - val_sparse_categorical_accuracy: 0.8910\n",
            "Epoch 27/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2358 - sparse_categorical_accuracy: 0.9144 - val_loss: 0.3073 - val_sparse_categorical_accuracy: 0.8876\n",
            "Epoch 28/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2345 - sparse_categorical_accuracy: 0.9159 - val_loss: 0.3137 - val_sparse_categorical_accuracy: 0.8862\n",
            "Epoch 29/30\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.2252 - sparse_categorical_accuracy: 0.9194 - val_loss: 0.3129 - val_sparse_categorical_accuracy: 0.8874\n",
            "Epoch 30/30\n",
            "1719/1719 [==============================] - 6s 4ms/step - loss: 0.2199 - sparse_categorical_accuracy: 0.9206 - val_loss: 0.3013 - val_sparse_categorical_accuracy: 0.8914\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "crZ6M-z03wH5"
      },
      "source": [
        "- 시각화 도구를 이용하여 손실함수와 정확도 확인"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZGqoGNBT6JwM"
      },
      "source": [
        "import pandas as pd"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 324
        },
        "id": "ya1JkJDK6MJT",
        "outputId": "b0c20ddb-1661-445a-af53-44c39aac35a2"
      },
      "source": [
        "pd.DataFrame(histo.history).plot(figsize=(8,5))\n",
        "plt.grid(True)\n",
        "plt.gca().set_ylim(0,1)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "odrL2DYO6kMy"
      },
      "source": [
        "### 학습 결과가 만족스럽지 못하다면?\n",
        "1. 학습률 튜닝\n",
        "2. 옵티마이저 변경\n",
        "3. 층 개수, 뉴런 개수 활성화 함수 등의 모델의 하이퍼파라미터 튜닝"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mJ_Q96a305h-"
      },
      "source": [
        "## 모델 평가\n",
        "- 모델을 상용 환경으로 배포하기 전에 테스트 세트로 모델을 평가하여 일반화 오차 추정\n",
        "- `evaluate()` 메서드 사용"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0a-j3zDB1NqC",
        "outputId": "6f29777e-3e1c-4480-97c7-14389a2c6f96"
      },
      "source": [
        "model.evaluate(test_images, test_labels)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "313/313 [==============================] - 1s 2ms/step - loss: 0.3371 - sparse_categorical_accuracy: 0.8822\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.3371058702468872, 0.8822000026702881]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 29
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jHDQezaj1TZ3"
      },
      "source": [
        "## 예측 만들기\n",
        "- `predict()` 메서드를 사용하여, 새로운 샘플에 대해 예측을 만들 수 있다.\n",
        "> 다른 새로운 샘플이 없으므로, 여기서는 테스트 이미지로 사용"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zm8sV9n21rnH",
        "outputId": "9eb63bce-c8cc-42d7-b58e-46ddad860b37"
      },
      "source": [
        "new_image = test_images[:3]\n",
        "model.predict(new_image).round(2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.01, 0.  , 0.98],\n",
              "       [0.  , 0.  , 0.98, 0.  , 0.02, 0.  , 0.  , 0.  , 0.  , 0.  ],\n",
              "       [0.  , 1.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  ]],\n",
              "      dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8U6o34P72EZx",
        "outputId": "e8b49021-9bb2-4c3e-8267-52c9a75fe229"
      },
      "source": [
        "new_labels = model.predict_classes(new_image)\n",
        "new_labels"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/sequential.py:450: UserWarning: `model.predict_classes()` is deprecated and will be removed after 2021-01-01. Please use instead:* `np.argmax(model.predict(x), axis=-1)`,   if your model does multi-class classification   (e.g. if it uses a `softmax` last-layer activation).* `(model.predict(x) > 0.5).astype(\"int32\")`,   if your model does binary classification   (e.g. if it uses a `sigmoid` last-layer activation).\n",
            "  warnings.warn('`model.predict_classes()` is deprecated and '\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([9, 2, 1])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JN8LMIWe2Q3G",
        "outputId": "17c99fb2-ddb2-4165-935e-a869f29bec83"
      },
      "source": [
        "new_pred = np.argmax(model.predict(new_image), axis=-1)\n",
        "new_pred"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([9, 2, 1])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 39
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YHEmASDV436J",
        "outputId": "c72dcd46-f822-430c-de23-3a5e1a8ea141"
      },
      "source": [
        "np.array(class_names)[new_pred]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array(['Ankle boot', 'Pullover', 'Trouser'], dtype='<U11')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    }
  ]
}