25_MNIST - CNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "pygame 2.0.1 (SDL 2.0.14, Python 3.8.5)\n",
      "Hello from the pygame community. https://www.pygame.org/contribute.html\n"
     ]
    }
   ],
   "source": [
    "# 기본\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# 경고 뜨지 않게 설정\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "# 그래프 설정\n",
    "plt.rcParams['font.family'] = 'Malgun Gothic'\n",
    "# plt.rcParams['font.family'] = 'AppleGothic'\n",
    "plt.rcParams['font.size'] = 16\n",
    "plt.rcParams['figure.figsize'] = 20, 10\n",
    "plt.rcParams['axes.unicode_minus'] = False\n",
    "\n",
    "# 데이터 전처리 알고리즘\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "# 학습용과 검증용으로 나누는 함수\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# 교차 검증\n",
    "# 지표를 하나만 설정할 경우\n",
    "from sklearn.model_selection import cross_val_score\n",
    "# 지표를 하나 이상 설정할 경우\n",
    "from sklearn.model_selection import cross_validate\n",
    "from sklearn.model_selection import KFold\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "\n",
    "# 모델의 최적의 하이퍼파라미터를 찾기 위한 도구\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# 평가함수\n",
    "# 분류용\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import precision_score\n",
    "from sklearn.metrics import recall_score\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "# 회귀용\n",
    "from sklearn.metrics import r2_score\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "# 머신러닝 알고리즘 - 분류\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from lightgbm import LGBMClassifier\n",
    "from xgboost import XGBClassifier\n",
    "from sklearn.ensemble import VotingClassifier\n",
    "\n",
    "# 머신러닝 알고리즘 - 회귀\n",
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.linear_model import Ridge\n",
    "from sklearn.linear_model import Lasso\n",
    "from sklearn.linear_model import ElasticNet\n",
    "from sklearn.svm import SVR\n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.ensemble import AdaBoostRegressor\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "from lightgbm import LGBMRegressor\n",
    "from xgboost import XGBRegressor\n",
    "from sklearn.ensemble import VotingRegressor\n",
    "\n",
    "# 차원축소\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis\n",
    "\n",
    "# 군집화\n",
    "from sklearn.cluster import KMeans\n",
    "from sklearn.cluster import MeanShift\n",
    "from sklearn.cluster import estimate_bandwidth\n",
    "\n",
    "# ARIMA (시계열 예측)\n",
    "from statsmodels.tsa.arima_model import ARIMA\n",
    "import statsmodels.api as sm\n",
    "\n",
    "# 시간 측정을 위한 시간 모듈\n",
    "import datetime\n",
    "# 주식 정보를 읽어오기 위한 라이브러리\n",
    "from pandas_datareader import data\n",
    "\n",
    "# 형태소 백터를 생성하기 위한 라이브러리\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "# 형태소 백터를 학습 백터로 변환한다.\n",
    "from sklearn.feature_extraction.text import TfidfTransformer\n",
    "\n",
    "# 데이터 수집\n",
    "import requests\n",
    "from bs4 import BeautifulSoup\n",
    "import re\n",
    "import time\n",
    "import os\n",
    "import json\n",
    "\n",
    "# 한국어 형태소 분석\n",
    "from konlpy.tag import Okt, Hannanum, Kkma, Mecab, Komoran\n",
    "\n",
    "# 워드 클라우드를 위한 라이브러리\n",
    "from collections import Counter\n",
    "import pytagcloud\n",
    "from IPython.display import Image\n",
    "\n",
    "# 출력 창 청소를 위한 함수\n",
    "from IPython.display import clear_output\n",
    "\n",
    "# 저장\n",
    "import pickle\n",
    "\n",
    "# 딥러닝\n",
    "import tensorflow as tf\n",
    "\n",
    "# 딥러닝 모델 구조를 정의하는 것\n",
    "from tensorflow.keras.models import Sequential\n",
    "# 층구조를 정의하는 것\n",
    "from tensorflow.keras.layers import Dense\n",
    "# 활성화 함수를 정의하는 것\n",
    "from tensorflow.keras.layers import Activation\n",
    "# CNN : 커널을 통해 합성곱을 구하는 것. 이미지의 특징이 두드러 지게 처리한다.\n",
    "from tensorflow.keras.layers import Conv2D\n",
    "# MaxPooling : 불필요한 부분을 제거한다.\n",
    "from tensorflow.keras.layers import MaxPooling2D\n",
    "# Flatten : CNN과 MaxPooling을 하면 2차원으로 결과가 나오기 때문에 \n",
    "# 1차원으로 변환하는 것\n",
    "from tensorflow.keras.layers import Flatten\n",
    "# Dropout : 이미지나 영상은 많은 은닉층을 가지고 있어야 데이터 학습 및 탐색이\n",
    "# 가능하다. 은닉층이 많으면 과적합 될 우려가 있다. 이에 은닉층에 있는 노드를\n",
    "# 경우에 따라 사용하지 않게 하여 과적합을 예방할 수 있다.\n",
    "from tensorflow.keras.layers import Dropout\n",
    "\n",
    "# 다중 분류를 위한 원핫 인코딩\n",
    "# 결과데이터의 종류 수 만큼 결과데이터의 컬럼을 늘리는 작업\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "\n",
    "# 저장된 학습 모델을 읽어온다.\n",
    "from tensorflow.keras.models import load_model\n",
    "\n",
    "# epoch마다 모델을 저장하는 함수\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint\n",
    "# 더이상 성능 향상이 이루어지지 않는다면 조기 중단 시킬 수 있다.\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "\n",
    "# 현재 프로젝트를 gpu에 할당한다.\n",
    "# 컴퓨터의 GPU는 메모리를 가지고 있다.\n",
    "gpus = tf.config.experimental.list_physical_devices('GPU')\n",
    "# gpu가 있다면..\n",
    "if len(gpus) > 0 :\n",
    "    try :\n",
    "        for gpu in gpus :\n",
    "            tf.config.experimental.set_memory_growth(gpu, True)\n",
    "    except RuntimeError as e :\n",
    "        print(e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# MNIST 데이터 \n",
    "from tensorflow.keras.datasets import mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# seed 설정\n",
    "np.random.seed(1)\n",
    "tf.random.set_seed(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 이미지 전처리(이미지의 색상 데이터를 0 ~ 1 사이로 맞춘다)\n",
    "# 3차원 행렬을 4차원으로 변경한다.\n",
    "# 2차원으로 되어 있는 이미지 행렬을 3차원으로 변경하는 작업\n",
    "# (가로, 세로, 채널) -> 채널 - 흑백(1), 컬러(3)\n",
    "# 이미지의 개수\n",
    "cnt = X_train.shape[0]\n",
    "# 이미지의 가로길이\n",
    "w1 = X_train.shape[1]\n",
    "# 이미지의 세로길이\n",
    "h1 = X_train.shape[2]\n",
    "# 채널\n",
    "channel1 = 1\n",
    "\n",
    "X_train = X_train.reshape(cnt, w1, h1, channel1)\n",
    "# 값을 실수로 변경한다(데이터가 0 ~ 255 이므로 255로 나눠주기 위해)\n",
    "X_train = X_train.astype('float64')\n",
    "# 255로 나눠 0 ~ 1사이로 맞춘다.\n",
    "X_train = X_train / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 이미지 전처리(이미지의 색상 데이터를 0 ~ 1 사이로 맞춘다)\n",
    "# 3차원 행렬을 4차원으로 변경한다.\n",
    "# 2차원으로 되어 있는 이미지 행렬을 3차원으로 변경하는 작업\n",
    "# (가로, 세로, 채널) -> 채널 - 흑백(1), 컬러(3)\n",
    "# 이미지의 개수\n",
    "cnt = X_test.shape[0]\n",
    "# 이미지의 가로길이\n",
    "w1 = X_test.shape[1]\n",
    "# 이미지의 세로길이\n",
    "h1 = X_test.shape[2]\n",
    "# 채널\n",
    "channel1 = 1\n",
    "\n",
    "X_test = X_test.reshape(cnt, w1, h1, channel1)\n",
    "# 값을 실수로 변경한다(데이터가 0 ~ 255 이므로 255로 나눠주기 위해)\n",
    "X_test = X_test.astype('float64')\n",
    "# 255로 나눠 0 ~ 1사이로 맞춘다.\n",
    "X_test = X_test / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = to_categorical(y_train)\n",
    "y_test = to_categorical(y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### CNN 구성\n",
    "- Conv2d : 케라스에서 컨볼루션 레이어를 추가하는 함수\n",
    "- Maxpooling2D : 데이터 축소\n",
    "- Dropout : 은닉층의 노드를 임의로 꺼주는 것(과적합 방지)\n",
    "- Flatten : 2차원 배열을 1차원으로 변환"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "\n",
    "# 입력층 + 은닉층 : 이미지 데이터를 입력받아 합성곱을 수행한다.\n",
    "model.add(Conv2D(32, kernel_size=(3, 3), input_shape=(28, 28, 1)))\n",
    "model.add(Activation('relu'))\n",
    "\n",
    "# 은닉층 : 데이터를 받아 합성곱을 구하고 MaxPooling을 수행한다.\n",
    "model.add(Conv2D(64, kernel_size=(3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Flatten())\n",
    "\n",
    "# 은닉층 : 최종 데이터를 모아 전달한다.\n",
    "model.add(Dense(128))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "# 출력\n",
    "model.add(Dense(10))\n",
    "model.add(Activation('softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 모델 컴파일\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', \n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 26, 26, 32)        320       \n",
      "_________________________________________________________________\n",
      "activation (Activation)      (None, 26, 26, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 24, 24, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 24, 24, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 12, 12, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 12, 12, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 9216)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 128)               1179776   \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                1290      \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 10)                0         \n",
      "=================================================================\n",
      "Total params: 1,199,882\n",
      "Trainable params: 1,199,882\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 모델들이 저장될 위치\n",
    "# epoch : 학습 횟수\n",
    "# val_loss : 검증 데이터의 손실률\n",
    "path1 = 'models/25_MNIST - CNN/{epoch}-{val_loss}.hdf5'\n",
    "path2 = 'models/25_MNIST - CNN/최종.hdf5'\n",
    "\n",
    "# 저장 콜백 설정\n",
    "# save_best_only : True를 넣어주면 모니터링하는 값이 개선 되었을 경우에만\n",
    "# 저장을 한다. 기본은 False 이며 False시 매번 저장을 한다.\n",
    "call1 = ModelCheckpoint(filepath=path1, monitor='val_loss', \n",
    "                       save_best_only=True)\n",
    "\n",
    "call2 = ModelCheckpoint(filepath=path2, monitor='val_loss', \n",
    "                       save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 조기 중단\n",
    "# patience : 현재 평가 값이 이전 평가 값 보다 좋다고 판단될 경우 앞으로\n",
    "# 몇번을 더 볼 것인가를 설정하는 부분\n",
    "call3 = EarlyStopping(monitor='val_loss', patience=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/2000\n",
      "120/120 [==============================] - 5s 16ms/step - loss: 0.4187 - accuracy: 0.8719 - val_loss: 0.0815 - val_accuracy: 0.9760\n",
      "Epoch 2/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.1182 - accuracy: 0.9654 - val_loss: 0.0532 - val_accuracy: 0.9831\n",
      "Epoch 3/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0811 - accuracy: 0.9762 - val_loss: 0.0400 - val_accuracy: 0.9870\n",
      "Epoch 4/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0650 - accuracy: 0.9808 - val_loss: 0.0367 - val_accuracy: 0.9879\n",
      "Epoch 5/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0562 - accuracy: 0.9826 - val_loss: 0.0286 - val_accuracy: 0.9904\n",
      "Epoch 6/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0492 - accuracy: 0.9850 - val_loss: 0.0310 - val_accuracy: 0.9908\n",
      "Epoch 7/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0440 - accuracy: 0.9862 - val_loss: 0.0324 - val_accuracy: 0.9897\n",
      "Epoch 8/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0379 - accuracy: 0.9880 - val_loss: 0.0310 - val_accuracy: 0.9905\n",
      "Epoch 9/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0351 - accuracy: 0.9888 - val_loss: 0.0295 - val_accuracy: 0.9911\n",
      "Epoch 10/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0329 - accuracy: 0.9897 - val_loss: 0.0281 - val_accuracy: 0.9903\n",
      "Epoch 11/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0307 - accuracy: 0.9898 - val_loss: 0.0302 - val_accuracy: 0.9914\n",
      "Epoch 12/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0268 - accuracy: 0.9915 - val_loss: 0.0283 - val_accuracy: 0.9927\n",
      "Epoch 13/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0257 - accuracy: 0.9917 - val_loss: 0.0259 - val_accuracy: 0.9915\n",
      "Epoch 14/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0252 - accuracy: 0.9920 - val_loss: 0.0278 - val_accuracy: 0.9921\n",
      "Epoch 15/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0231 - accuracy: 0.9923 - val_loss: 0.0313 - val_accuracy: 0.9923\n",
      "Epoch 16/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0212 - accuracy: 0.9925 - val_loss: 0.0288 - val_accuracy: 0.9919\n",
      "Epoch 17/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0200 - accuracy: 0.9933 - val_loss: 0.0297 - val_accuracy: 0.9920\n",
      "Epoch 18/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0192 - accuracy: 0.9936 - val_loss: 0.0304 - val_accuracy: 0.9921\n",
      "Epoch 19/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0198 - accuracy: 0.9933 - val_loss: 0.0273 - val_accuracy: 0.9930\n",
      "Epoch 20/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0174 - accuracy: 0.9942 - val_loss: 0.0317 - val_accuracy: 0.9919\n",
      "Epoch 21/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0165 - accuracy: 0.9944 - val_loss: 0.0285 - val_accuracy: 0.9920\n",
      "Epoch 22/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0160 - accuracy: 0.9948 - val_loss: 0.0329 - val_accuracy: 0.9917\n",
      "Epoch 23/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0161 - accuracy: 0.9948 - val_loss: 0.0271 - val_accuracy: 0.9920\n",
      "Epoch 24/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0150 - accuracy: 0.9948 - val_loss: 0.0329 - val_accuracy: 0.9921\n",
      "Epoch 25/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0155 - accuracy: 0.9948 - val_loss: 0.0283 - val_accuracy: 0.9921\n",
      "Epoch 26/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0127 - accuracy: 0.9960 - val_loss: 0.0292 - val_accuracy: 0.9930\n",
      "Epoch 27/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0138 - accuracy: 0.9951 - val_loss: 0.0277 - val_accuracy: 0.9924\n",
      "Epoch 28/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0125 - accuracy: 0.9957 - val_loss: 0.0293 - val_accuracy: 0.9931\n",
      "Epoch 29/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0130 - accuracy: 0.9956 - val_loss: 0.0286 - val_accuracy: 0.9936\n",
      "Epoch 30/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0113 - accuracy: 0.9962 - val_loss: 0.0301 - val_accuracy: 0.9935\n",
      "Epoch 31/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0131 - accuracy: 0.9956 - val_loss: 0.0311 - val_accuracy: 0.9924\n",
      "Epoch 32/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0106 - accuracy: 0.9966 - val_loss: 0.0315 - val_accuracy: 0.9921\n",
      "Epoch 33/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0123 - accuracy: 0.9955 - val_loss: 0.0313 - val_accuracy: 0.9923\n",
      "Epoch 34/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0128 - accuracy: 0.9958 - val_loss: 0.0304 - val_accuracy: 0.9927\n",
      "Epoch 35/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0115 - accuracy: 0.9961 - val_loss: 0.0305 - val_accuracy: 0.9926\n",
      "Epoch 36/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0115 - accuracy: 0.9964 - val_loss: 0.0324 - val_accuracy: 0.9931\n",
      "Epoch 37/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0107 - accuracy: 0.9960 - val_loss: 0.0280 - val_accuracy: 0.9935\n",
      "Epoch 38/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0107 - accuracy: 0.9962 - val_loss: 0.0313 - val_accuracy: 0.9931\n",
      "Epoch 39/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0096 - accuracy: 0.9967 - val_loss: 0.0315 - val_accuracy: 0.9932\n",
      "Epoch 40/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0089 - accuracy: 0.9969 - val_loss: 0.0336 - val_accuracy: 0.9928\n",
      "Epoch 41/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0097 - accuracy: 0.9966 - val_loss: 0.0308 - val_accuracy: 0.9938\n",
      "Epoch 42/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0084 - accuracy: 0.9971 - val_loss: 0.0336 - val_accuracy: 0.9931\n",
      "Epoch 43/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0086 - accuracy: 0.9969 - val_loss: 0.0295 - val_accuracy: 0.9933\n",
      "Epoch 44/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0084 - accuracy: 0.9970 - val_loss: 0.0316 - val_accuracy: 0.9932\n",
      "Epoch 45/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0084 - accuracy: 0.9970 - val_loss: 0.0318 - val_accuracy: 0.9927\n",
      "Epoch 46/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0086 - accuracy: 0.9972 - val_loss: 0.0357 - val_accuracy: 0.9938\n",
      "Epoch 47/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0089 - accuracy: 0.9970 - val_loss: 0.0308 - val_accuracy: 0.9932\n",
      "Epoch 48/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0076 - accuracy: 0.9973 - val_loss: 0.0298 - val_accuracy: 0.9929\n",
      "Epoch 49/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0084 - accuracy: 0.9970 - val_loss: 0.0307 - val_accuracy: 0.9937\n",
      "Epoch 50/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0097 - accuracy: 0.9964 - val_loss: 0.0334 - val_accuracy: 0.9935\n",
      "Epoch 51/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0080 - accuracy: 0.9975 - val_loss: 0.0318 - val_accuracy: 0.9934\n",
      "Epoch 52/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0070 - accuracy: 0.9976 - val_loss: 0.0354 - val_accuracy: 0.9926\n",
      "Epoch 53/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0075 - accuracy: 0.9975 - val_loss: 0.0316 - val_accuracy: 0.9932\n",
      "Epoch 54/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0071 - accuracy: 0.9978 - val_loss: 0.0318 - val_accuracy: 0.9934\n",
      "Epoch 55/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0065 - accuracy: 0.9979 - val_loss: 0.0340 - val_accuracy: 0.9936\n",
      "Epoch 56/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0057 - accuracy: 0.9979 - val_loss: 0.0385 - val_accuracy: 0.9931\n",
      "Epoch 57/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0072 - accuracy: 0.9974 - val_loss: 0.0338 - val_accuracy: 0.9930\n",
      "Epoch 58/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0076 - accuracy: 0.9977 - val_loss: 0.0326 - val_accuracy: 0.9919\n",
      "Epoch 59/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0060 - accuracy: 0.9979 - val_loss: 0.0342 - val_accuracy: 0.9937\n",
      "Epoch 60/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0065 - accuracy: 0.9979 - val_loss: 0.0336 - val_accuracy: 0.9933\n",
      "Epoch 61/2000\n",
      "120/120 [==============================] - 2s 15ms/step - loss: 0.0068 - accuracy: 0.9976 - val_loss: 0.0350 - val_accuracy: 0.9934\n",
      "Epoch 62/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0074 - accuracy: 0.9976 - val_loss: 0.0340 - val_accuracy: 0.9930\n",
      "Epoch 63/2000\n",
      "120/120 [==============================] - 2s 16ms/step - loss: 0.0064 - accuracy: 0.9977 - val_loss: 0.0346 - val_accuracy: 0.9929\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, y_train, validation_data=(X_test, y_test), \n",
    "                    epochs=2000, batch_size=500, \n",
    "                    callbacks=[call1, call2, call3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABIYAAAJGCAYAAAAu146OAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAAsTAAALEwEAmpwYAAB1SklEQVR4nO3deZwcdYH///en+pi7Mzkmx0wuUE6BoAQEUUARUGABYWGDAssKIuuq6wKKiIoL+wVXV3+eLIfHIuKK3LeArHLKkWg4QggghEDOyeSazNFH1ef3R1V1V/d0kkky091Jv5559ONT9ama6U+6a7qr3vWpTxlrrQAAAAAAAFB/nGo3AAAAAAAAANVBMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHUqXu0GRE2YMMHOnDmz2s0AAAAAAADYacybN2+1tbaj3LKaCoZmzpypuXPnVrsZAAAAAAAAOw1jzFubWsalZAAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKYIhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAAAA6hTBEAAAAAAAQJ0iGAIAAAAAAKhTWxUMGWNSxpifGWP+YwvrTTTG/NYYs8EYs8YY81NjTOP2NRUAAAAAAAAjaVjBkDFmrDHmQkmvSjp7C+s6ku6VlJJ0hKTTJP2dpP9vexoKAAAAAACAkTXcHkMnSvpXSZdJemIL654gaTdJ/2Ct/Yu19g+SviDpHGPM2G1uKQAAAAAAAEbUcIOhOyXtYq29dhjrHifpAWttb6TuAUmupA9uXfMAAAAAAAAwWoYVDFlr11lr3WH+zr0lLSj5+YykNyW9a+uaBwAAAAAAgNEyGnclmyBpTZn6tfLHHQIAAAAAAEANiI/C74zJv2yslA0eRYwx50k6T5KmT58+Cs0BAADYdtZa5TyrTM5TOucpnXOVznrKup6ScUcN8Zga4o4aEzEl445ijql2k0eEtVaelTxr5Vkrm5+WXM/mlxtJjmMUc4xixshxpLjjyDGSMZV9LcL3yvWC0rXKeZ5czypbMh+ul3W9/P/L8/xyc/93z9r8euFyyf+/OkaKGVOYdowcY2RKp42R4/jrOMavd4PfG7bLjf4/PE+up6K2l67rWZv//U7w/DKF5zBSfnnYvnBdEyy3wWvov5aSlf9/L8z7y638lcPlpTv44bsevv1G+YnIMlN23fD1D19zK+Xfg/B9yZcaum74O43xnze6CYb/T2PCdUx+WiWvgRvZFlxv0+998bbjtyXm+K9pzCm81/kymA63iXC7iEXWiTtG8VhQOk4w7ZeJmFHMcRR3jBIx//MmETOKx5xgfSPPSoNZV+mcq8GsN6T0l/nl0HpXmZzNv8+Fcuj7n39fItuBtdHXu/BehNuBMYX3RtH3oeR9Cd9j19ri9yB4nd2i98bK8xTUF9bbHDv0sLR4uS0pVfjbKK6P/r2EywqfCbHgvY3HgtIJSyf/XseGlI6/7QTfJeH/zYu81vn5os+nsC78Gylsq5KKXqfSvy8v+jNe8d/WlmzpYz76t2DM0L8BJ/83osi0/10Sfj4NfX+GNmyTTY1uo2W3Vztk+47OS9Kps6fphFmdW34xdmCjEQz1qnzPoDEq05PIWnudpOskafbs2cPY9AAAtaJ4h6T4i7doh97zl4UHlK5nN3kQ5EUO7DxrlXODncCgrogZOhndgYjuSkT3K6I7evn/R35h8Y5eWG7qC6r0oGbIdJlGZnOeBvM75W5+53wgMh1dNlBSl/OskjFHyXjw2MR0w5Blsfy051llPU+u6x8w59zg4Dk4cM66/sFoLljuBnU511POK+ws5nf4Iwd50Z37aJ1KDtLCg3LXs8HzRNoQtidoQy5oYzjtelbGSIlY5KAp2PlOhAdIMcc/YIpMxxxHiWA9SX7QExwYZdxwOgh/Isu2cIxRJBEz+bAoGhg1JIrrGuKOjDH+TvuQEKKwk+7/HQ09AAin/b8n/0Ao/LuJ/o25wY5+aZ0bvI/hz5QeXIyEaDgSPfAND4hjQUBRdPCvoTvr0WCgEEoU6sP/z5DPCGAU5UO+yEGuVDgAD7fLHUEy5qghEXxexZyiAKcoSJOkSNhYGsCVCxmlcgfkxX/L0QPykB+wFV7jMEzNB2+RIMFxpITj5IPW4QTTW4qtCz9u8vPR73wTqS+tkykEV8Uhr18OBN/lhf0gr7DcLXyGh4Fu+P+JBr+l807QECcIVYwK60VD6HBbjUfmo7+jKDzewqs0nIAtGqRGgzzXFvYBokFf/rsp+L7a9Bs0rKrgvdn89lpue1dkWTbnbfb/uTMYjWDodUl7RCuMMXFJu0p6eRSeD8AIcT1Xg+6gBnIDGswVSte6aow3qiHWoMZYY2E63ijHjMYVqdvO37Gwcj3XP7C0VjnPlbUK2uoEZ6lMPqwYcuBR5qyCZ/2D1UzO7yWQznnK5DxlXC9flwnq0q6nbGRZuF7OtSVneiJnaLwyZ2xsyVkdq+CLXjLGlXFcyeQkZaVg2pqcjHLylPOXmWBaOVm58mRkvbism5DnxeW6wcOLyc3Flc05yuXiyuSc4GHz/490zi06azqcs0jYEi//PhnjKh531ZiwSiQ8NSQ8JeOekglPiZinWJOn5lZXqZgnY6w8a+S6jlzPaNB1tNE1yg36dVlX/nvpGmVzwcM1cj0j2ZhkHcl4knEl48oYVzHHUzzmKeZYxeKuXzquHMdTLObJMZ5MMO84noxxZeXJs56sdSXjycqVtZ6s8UsF61i5wc5jYV6yhd4KMSMT8/9G/Z1YJ79z2iCjRmOCZYUdWCOjmElINiGjhIyNS9afl+dPW5uQ9eLKeXFl3Lg8Ny6biSvnxuW5MUlxNcbjfoiTjCsVjysZS6ghHlNjPKHGhL/ML/2Dpnzgk3AUd5z850F4Bj4MkqJ1g1lX6WxOadfTYDarDYOeujf6dZKVdTw5xsoYT0ZBaayMY/0ymJe8wrysZDwZR3ISfsCSDA+QFB5AOUEQ4+R7sjjBgVP+QMs4Mo7nbw8Ktgn50zasC963cJ38e2j8996zrlzryrWeXOvJs648z1PO5uRZT6515VkveLiFUlaezcnacOc8JkeOjHFkgtIpqXOiy4y/zDFO5MBHkQOhoAeNE9bZyAGVIqV/kBk3Mf81c2KKRUrHiSlu4v5ZfFNYFndiQc+omIyMsl5WGTejtJtW2s0o46aV8TLKRKb9ddLKepngkVXGSyvnZWTkKO4kFHfiSjjxoEwGZUIJJ65ELJxOKBlLKBGLKxlLKukkFHOc4MDLDb7T3JJeLeF74E9bWXleUCfJWi84GHLyY034r3vhgNMEf3thKRN8qwbzjokpFrwmMcUVc2L+tIlF3i9/PubE5CimmInLCV5ba62yXlZZL62szSgbvFYZr1AWXtdoGdS72WC7jykevIZOMB0z4fsYD97HuOImLscJp/0y/76H24DjKGYcxZ1YEHDG/G3FMcH/q7ANxkxMkvLbQtbLKutmlXGzSgfty7hZZbygdIPtIF+XUdbLRc5EOLKSjDWSjGxYSlK+TrIyki3UG4U9j5wghHXyvVTC3kbxkrrwczV8n/1tLJnf3hKxhJJOMj9dVEYeyZi/zVpZ5bxc/uF6rnI2p6yX9ae9nHI2V7ROzsvJtW6+DPfpwu3V2uLp0mX5eWsL68vb7DLP3xksu17h86CwDUXnHeME21Rhm47O+59XhR4v0dc3/y96Miv6Hsj/fE7EEmqMNSoZS6ox1qiGeIMaYmUe8QYlnMQm9zastUq7aW3MblRfts8vM33qzfb685lIfWT5oDuY38bD/1P+77ekrqgMX5PgbyL6HkXn8+0rXR6ZNjKbfb7NtiX4bIm+9uHvjL7e5V7/6PzuYzObfG13FqMRDP1B0mXGmAZrbTqo+5iktKSnRuH5UINyXk4bMhuUdbNFO4XR0vWG1uVLzy8d4+QDiPADsagcoWDCWivXukVf5Dkvp6ybVdZmh3ypbepLLOtlC19owbqSir5US788N/WFG+4Ixp14ZEczrcHcoDJuRoNuSZkbzK+Tf+SCn3EHNZgbzIc9fdkB9WcH8sHPoDuojJeWa7Nb/doZxeUoKccm5KhBjhIySspRUsYmgvlE8AUchhOerM0FB4v+vH/gkfMPOIwbzPsHKDZ/oCIF55UkWVkTTvv1/oHTcN9zI8nxd6ysE0w7Cs7NBDtcfl2hPvw/h10los9nI2W4vFCfPzMRXc8U/6wp83NF/VSMlXWDwGekGUmJ4BGpdJRQTEk1mKSaTUKOiZdvW/ieRNtvC1/+xhTW9g8inPxBhL8z6uR3PPwDt1h+pyhmosv8v3f/QMaTDQ50wh05fzrcuSvMFw6EvMjORnQHRJF5m/8fqWQnZiT4O7ausjYrbxP3dcgGj7JKfyTYfLXp/UHFtXVf+GEcsLWfCJvaQS7deY478fzOWfEZ5WCHPbJjKBXvQFr5Z1GtrPrdrNK28FlX9j0KX5/hvgCepEzwCITbojFGjvwd3rAu3PYk5Q8mSg8w8mLBoxJq4OTmJnfSIzvxMROTMabou39z+wsj9XdYaQ2xBiWdpJKxpD8dC6bjDRoTSyrpNMuTF+xLZJTz+jXgZpXL+vsiOVso8/snXna7Xg/HOHLkBOGOk798p+gzLzxBUuZgrRry+4WR/cDofmFbsiX/Oma8QQ3khu63hftppcFEuM2NdvtL9/WSTnLI/qGMH1TIBp8rxubL/GdMmeAj3J+VJLmSzQ09EA/f13xdyQG5pMJ+rZtVxquNA+Jy22u4TxANYPKf1zL5dba0LPp7wunw7yH8DIoGVuE+fvjZFG5H1RYzsSHHSTkvlw95wuOSzYk7cbUl2tSSaFFrslXJWDL/OpQep5U7jouuE54cGBKAlQnIggVDgjQj4wd4m3jOSvnX9/2rdm3ftWLPVw3bHQwZYzok3SXpEmvto5J+Lenrkm40xlwhaYqkqyX9RyQowg4k62a1Lr1u6GOw/Pza9Fr1Znor1r6EkxiSmIfTCScRnHnK5s/aFJWRIGhH3dncJGtkFJxNtwlZLyHPS8h1E5KXlLUJyWsOyqSsl/TPsHv+vFFSDU6DmuJNSsbikpORNVkpePjTmaCXSrBMfmlNRjmTlVWvP69sEATECg8npphictQgY+L+WUTFg6DAP4vomHhkOhYclJV8uRd9sUe/6AvdiQuJf3imu/iRr7clyyLzntwgoIhcC+2Y4MxxeDbeP/vmOI7ixpHjFM7Qb+qMUThf7kszemY2rAsPJsIDjDBgDOvCgLFcXdyJy7OeBt3B/IF0uTCxXOiY8fwy5+WKv8hNmfZu4qxMftMMe3UFX+zhDkd4JjAMfQpnt72inRL/PY7ndxKjoVF4xja6s1f6KD0LNNz3IVw3enZpW4U7b0XvV/ieltQlYomi5UknKcdxioLqTZ193eQBkOcGPR4KAXS5clPL4k48H/yEZ1HD179awp4Gw9mew+VZLzt0GyzZ9oq2Qw3dVrf2QCT62RVuC9GznqVhSulZz/AR1htjig7gpaFnXsPXpzRgC39uU2e7o+9t9Gx5UQC4mXaPxvYw5ADFekWfS9LmzwSXrhceiJT7vfkycvBX7sAofIShT0OsQYlYomhfZLT+NvK9MFz/xFS4TZTbFsOD6dLPue0R3a7CMjxg3tKBdLl1rGz+hGC5HhFxEx/VzxnXc/NBS/RvfUufD/ntJwhzwqCn6ASgk1DMqVQyPHLC9zS675zzcoUeUWGvKC9TtI8d/dwIvz+iPbliTkwJU/Kd4sTz3yul39k7gvDzIrp9S8WB6nA/n8PfFz05XHriN/rY1AniRCzhhzyJ1kKZLJ6P1iVjyQq/attnix0OgvqyAXeZgDQ6H10+rnFcNf57FTUSPYYSknaXNE6SrLV9xpijJf23pGcldUv6sbX2+yPwXNiMnJdTd3+3lvUt0/K+5dqY2eh3X853Uc3kP1zCQKTc8vCDfiA3oHXpderL9m3yOZviTWpvaM8/utq61N7QrrENY5VqSCkZS26xi+GmljnGkbV2kz1iyvWcKf2gzLpZJZxGNTit/uUT8Zhk47I2Js/zH/6lGI5yrqNsLqZcziidc5TNGqWDSzByrlEuZ5Tz/Esz/EsxYrL5nib+pRnWxorng87YJn+5iH/JRjgdi7lKxDzF41bxmKt4LLiUI+YpFnPlOK4cx/qBiZKKKSHHJGWCXhx+mVDMJP2eHZHSmLjC678bEzG1NMTV2hCWcbUEj9aGmFqS/nRbYzy/PBz7AgB2BMYUgtM2tVW7ORgl/iVfo9HhXUpsrttdjYoFJ1kaYg1Vef7oJWWhHfF1DIWvJwqMMUqYxGYvU4LPMY6cmLND/w3saMIAke1z+231N6u19oiS+WXyb1EfrVsoqWg9bL++bJ+Wb1yuZX3LtKJvhZb3Ldeyjf70sr5lWtW/arNdYMPuquXOQodnNRrjjUolU/npaOjT3ugHPtH5kdoR8TyrDYNZre3Pam1/xn/0ZbUxnSs78OqQgVpzngYzrgZzwUCtGb8uM8yBwhoTjlqDQKS1Ma7xybjamv2QpDFeGKw1OrBrw7AGfS3zs5F1dpY71wAAAAAAdkyjc8oF26w/268FPQv0cs/LWrpxqZZvXO4HQH3LhlyeFTdxTWqZpCktU3TQ5IM0uWWyprRMUWdLpya3TtaY5Jj82dOEkxiRsXi2JOd66ku76k1n1TuY07pI0LOuP6u1fRmt7c9qXbSuP6P1A9kt3gElGXPUGNwtoTERU1MipsaEf5eXMU0JTU41+MviMTUlY2pIOGpKxIoCn3LTLQ1xJWK1NYAyAAAAAACVQDBURa7n6s31b+qF1S/ohe4X9OLqF/X6utfzvX7aEm2a3DpZnS2d2n/i/n7o09qpKS1TNKVliiY0TRiVa5U3pnN6q6dPy9cNamM6p950ThsHc9qYzmrjYHQ+eETqBrKbHwSsMeFoXHNS7c1JjW1JaEp7k8Y2JzQ2rGtOaGxLUmOD6daGuB/yxGP0rgEAAAAAYIQRDFXQ6oHVerH7Rb24+kW90P2CXup5KT9+T1uyTftO2Fcfmf4R7TthX+0zYZ9RG+TKWqt1/Vkt7unTkjX9Wry6X2/19OmtNX65emP5uw/EHJPvZdMW9LgZ15LU9HHN+fnWhoRaG+NqC3rltDcl8iHQ2OakGhNctw0AAAAAQK0gGBolaTethT0L8z2BXuh+Qcv6lknyLwHbbexuOn7X47Vfx37ad8K+mpGaMaKXellrtao3rbd6+rW4p88Pfnr6g0efNgwW365wyphGzRjfrCP3nKQZE5o1Y1yLusY2KdUYD4KehBoTDEYMAAAAAMDOhGBoFPxg3g90w8s3KOf54cuUlinad8K++uRen9R+Hftpr3F7qTHeOOLPm3M9PfPmGt3/4nI9uGBFUc+fmGM0dWyTpo9r1v7TujRjfLNmjG/RzPHNmjaumZ48AAAAAADUIYKhUbD3+L31j3v/o/bt2Ff7TdhPHc0do/ZcWdfTU3/r0QNBGLS2P6umREwf2XOi3r/ruHz409nexADLAAAAAACgCMHQKDh65tE6eubRo/b7MzlPT76+Wve/uFwPvbxS6weyaknGdORek3TsvpN1+O4T1ZSkBxAAAAAAANg8gqEdxGDW1eOvrdYDLy7XwwtXqncwp7aGuD669yQdu+8UfWi3CVwOBgAAAAAAtgrBUA0bzLr606JuPfDScj2ycJU2pnNKNcZ1zHsm69h9J+vQd09QQ5wwCAAAAAAAbBuCoRo0kHF16Z0v6vcvrVB/xtXY5oSO32+KPr7vFB2y63gl44wVBAAAAAAAth/BUA26a/5S3f6XpTpt9lSdMKtL7991HANHAwAAAACAEUcwVIPumr9Mu05o0X+esp+MMdVuDgAAAAAA2EnRDaXGrFg/qKff7NEJ+3cSCgEAAAAAgFFFMFRj7n1hmayVTpjVWe2mAAAAAACAnRzBUI25a/4y7Td1jHbtaK12UwAAAAAAwE6OYKiG/K17o15cup7eQgAAAAAAoCIIhmrI3fOXyRjp7wiGAAAAAABABRAM1Qhrre5+fpkO2XW8JqUaq90cAAAAAABQBwiGasSLS9frzdV9OnF/egsBAAAAAIDKIBiqEXfNX6ZkzNHH3jOl2k0BAAAAAAB1gmCoBrie1T3PL9MRe3RoTHOi2s0BAAAAAAB1gmCoBjzzRo9W9aZ14v5d1W4KAAAAAACoIwRDNeCu+cvUkozpyL0mVrspAAAAAACgjhAMVVk65+r+l5brmH0mqzERq3ZzAAAAAABAHSEYqrI/LepW72COy8gAAAAAAEDFEQxV2d3zl2l8S1KHvmt8tZsCAAAAAADqDMFQFfUOZvWHhSt1/H5TFI/xVgAAAAAAgMoijaiihxasVDrn6QQuIwMAAAAAAFVAMFRFdz2/TFPHNul909ur3RQAAAAAAFCHCIaqpLs3rSdfX60T9++UMabazQEAAAAAAHWIYKhK7n9xuVzPcjcyAAAAAABQNQRDVXLX/KXac3Kbdp/UVu2mAAAAAACAOkUwVAVLevr1lyXr6C0EAAAAAACqimCoCu55YZkk6e9mTalySwAAAAAAQD0jGKqCu+cv04Ezx2rq2OZqNwUAAAAAANQxgqEKe2XFBi1a2asTuIwMAAAAAABUGcFQhd01f5lijtGx+0yudlMAAAAAAECdIxiqIM+zunv+Mn1otwka39pQ7eYAAAAAAIA6RzBUQX9ZslZL1w3oxP07q90UAAAAAAAAgqFKumv+MjUmHB21N5eRAQAAAACA6iMYqpCs6+m+F5fro3tNUmtDvNrNAQAAAAAAIBiqlCdeX601fRmdyN3IAAAAAABAjSAYqpC75y/TmKaEDt+9o9pNAQAAAAAAkEQwVBEDGVcPLlihY/edrGSclxwAAAAAANQGUooK+MPClerPuDphFpeRAQAAAACA2kEwVAF3zV+myalGHbTLuGo3BQAAAAAAII9gaJSt68/o0VdX6e9mTVHMMdVuDgAAAAAAQB7B0Ch74KUVyrqWu5EBAAAAAICaQzA0yu6av1S7drToPZ2pajcFAAAAAACgCMHQKFq+fkDPvLlGJ87qkjFcRgYAAAAAAGoLwdAouvf55bJWOmH/zmo3BQAAAAAAYAiCoVF01/NLNWvqGO0yoaXaTQEAAAAAABiCYGiUvL5qo15aukEnMOg0AAAAAACoUQRDo+Tu55fJGOnv9ptS7aYAAAAAAACURTA0Cqy1unv+Un3gXeM1MdVY7eYAAAAAAACURTA0Cl54Z70W9/TrxFlcRgYAAAAAAGoXwdAouGv+MiVjjo7ZZ3K1mwIAAAAAALBJ8Wo3YGd00ns7tdukVo1pSlS7KQAAAAAAAJtEMDQK9pvarv2mtle7GQAAAAAAAJvFpWQAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKYIhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAAAA6hTBEAAAAAAAQJ0iGAIAAAAAAKhTBEMAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDo17GDIGHOMMeYvxphBY8wrxpjTNrPunsaYB40xfcaYt4wxVxhjEiPTZAAAAAAAAIyEYQVDxpgDJN0p6deS3ivpl5J+Y4z5QJl1WyQ9Imm9pEMl/YukT0u6amSaDAAAAAAAgJEQH+Z635D0G2vt94P5hcaYgyVdIOmpknWPlDRW0lnW2kFJ840xX5f0HUkXjUCbAQAAAAAAMAK22GPIGBOXdLSkm0sW3S7pw2V+xJOUk5SJ1G2UFNvGNgIAAAAAAGAUDOdSsl0kNUlaUFK/SNI4Y0x7Sf3DklZK+i9jTLMxZjdJ35T0w+1sKwAAAAAAAEbQcIKhCUG5pqR+bVCmopXW2rSkUyWdL7+n0KvBz/5HuV9ujDnPGDPXGDO3u7t7uO0GAAAAAADAdhpOMBReAuaW1NuSUpJkjNlF0gOSrpF0iKRj5fc4uqncL7fWXmetnW2tnd3R0THcdgMAAAAAAGA7DWfw6d6gTElaHakfE5Rri1fXv0v6o7X2grDCGPOspDeNMR+11v5hWxsLAAAAAACAkTOcHkNvyB9Qeo+S+j0kLbHWbiypP1DSM9EKa22P/EvKDtrGdgIAAAAAAGCEbTEYstb2SnpW0skli06WdF+ZH1kh6X3RimCA6t3kD0oNAAAAAACAGjCcS8kk6SpJtxhjFkl6TNJJko6SNMsY0yHpLkmXWGsflfRdSXcbY96Uf0v7cZIul7RO0u9GtPUAAAAAAADYZsO5lEzW2rslfV7SJZLmSzpR0tHW2sWSEpJ2lx8AyVp7v/zQ6EhJj8sfdPoNSYcEvY8AAAAAAABQA4bbY0jW2uslXV+mfpkKt7QP6/4o6UPb3ToAAAAAAACMmmH1GAIAAAAAAMDOh2AIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKYIhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAAAA6hTBEAAAAAAAQJ0iGAIAAAAAAKhTBEMAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKYIhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAAAA6hTBEAAAAAAAQJ0iGAIAAAAAAKhTBEMAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKYIhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAAAA6hTBEAAAAAAAQJ0iGAIAAAAAAKhTBEMAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKYIhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdGnYwZIw5xhjzF2PMoDHmFWPMaVtY//3GmIeNMb3B41fb31wAAAAAAACMlGEFQ8aYAyTdKenXkt4r6ZeSfmOM+cAm1j9U0h8kPSbpEEmHSrp1BNoLAAAAAACAERIf5nrfkPQba+33g/mFxpiDJV0g6anoisaYmKSfSfqKtfa/I4te2N7GAgAAAACwsxkcHFR3d7cGBweVy+Wq3RzsQBKJhCZOnKhUKrXNv2OLwZAxJi7paEknlSy6XdIPyvzIhyWNkfTzbW4VAAAAAAB1YP369Vq5cqU6Ojo0efJkxeNxGWOq3SzsAKy1GhgY0NKlSyVpm8Oh4VxKtoukJkkLSuoXSRpnjGkvqT9S0pOSzjLGvGaMWWuMucsYM32bWggAAAAAwE5q9erVmjp1qsaOHatEIkEohGEzxqi5uVldXV1atWrVNv+e4QRDE4JyTUn92qAsjaT2CB5zJJ0r6VRJkyXdY4wZ8nzGmPOMMXONMXO7u7uH3XAAAAAAAHZ0mUxGTU1N1W4GdmBNTU3KZrPb/PPDGWMoFpRuSb0tKUMpSeMkHWyt7ZckY8zLkt6QdJSkB4t+ibXXSbpOkmbPnl36uwAAAAAA2KnRSwjbY3u3n+H0GOoNytKeQWOCcm1JfUbSE2EoJEnW2mWSXpW097Y0EgAAAAAAACNvOMHQG5I8+ZeHRe0haYm1dmNJ/WINDZEkv2dRf5l6AAAAAAAAVMEWgyFrba+kZyWdXLLoZEn3lfmR/5N0uDEmHJtIxpgZkvaS9Ny2NxUAAAAAAAAjaThjDEnSVZJuMcYskvSY/FvXHyVpljGmQ9Jdki6x1j4q6U5Jb0q62xjzFfnh0/8n6R5r7V9GtvkAAAAAAADYVsO5lEzW2rslfV7SJZLmSzpR0tHW2sWSEpJ2lz/gtKy1OUkfk7RK0u/lh0bzJP3jyDYdAAAAAABg0zzPq3YTat6wgiFJstZeb63dxVrbaK09xFr7TFC/zFo7wVp7R2Tdd6y1J1lrW621Y62155UZiwgAAAAAAOxk/va3v+nMM8/UtGnT1NbWpo985CN69dVX88uXLFmiT33qU5owYYKampp04IEHqqenJ7/8F7/4hfbbbz81NjZq8uTJ+tGPfiRJOvvss3XGGWcUPdfixYtljNHixYuL5p944gl98IMfVCKR0DvvvKPe3l597Wtf02677abm5mbts88+uu++4tFxBgcHdckll2jmzJlqaGjQu9/9bj3++OM69thj9alPfWrI//OUU07R+eefP1IvW9UMOxgCAAAAAADYkjvvvFOTJ0/W7bffrscee0ye5+mss86SJK1cuVKHHHKIli9frptvvllPP/20TjnlFGWzWUnSFVdcoc9//vP69Kc/rWeffVb/8z//o6ampq1uw6WXXqrPfe5zeu655zR27FgtWLBACxcu1DXXXKO5c+fqyCOP1Jw5c7R69WpJfs+iE044Qf/7v/+r7373u5o3b54uu+wyZTIZnXnmmbr33nuVTqfzv7+3t1f3339//v+1IxvuGEMAAAAAAKBC/v2eBXp52YaqtmHvzpQu+7v3bPXPffGLX1QikcjPX3755Tr88MO1ceNGXXHFFRo7dqwefvhhxWIxSdKsWbMkScuWLdMVV1yhX/3qV5ozZ44kab/99tumth922GH65Cc/mZ8/4IADdMcd+Qud9N3vflfXXXednnnmGR133HH5EOuVV17RzJkzJUn77LOPJGlgYECf/exn9fDDD+v444+X5IdfU6dO1Qc+8IFtal8tIRgCAAAAAAAjJpFIqL+/X0899ZQWLlyouXPnSvJ7C91777268MIL86FQ1EMPPaRUKpUPhbbHMcccM6RNnudp3rx5euGFF/TKK6/IcRytWLFCknTvvffquOOOy4dCUU1NTTr55JN166235oOh3/72tztFbyGJYAgAAAAAgJqzLT11asUNN9ygz3/+89pjjz209957a9y4cZIka62WLl1aNnyRpKVLl2rGjBkj0oaJEycWzc+bN0+nnHKK4vG43ve+9+ld73qXGhoaZK3NP/fmeiedccYZOu2005TNZtXb26tHHnlEP/3pT0ekrdVGMAQAAAAAAEZET0+PPvvZz+rmm2/WiSeeKEl6+eWX9cMf/lCS1NbWppUrV5b92c0tk6TGxsb8mEChvr6+sus6TvGQyueff75OPvlkff/738/X/fznPx/2c3/kIx9Rc3Oz/u///k9LlizRwQcfvMmAa0fD4NMAAAAAAGBE/O1vf1M6ndYRRxyRr/vTn/6Unz7ssMN04403lv3Zww47TEuXLtUf//jHssunTp2q119/vajumWeeGVa7FixYoMMPPzw///LLL6u7u7voue+77z6tXbu27M87jqPTTz9dt956q26++ead5jIyiWAIAAAAAACMkHe/+91qbm7W17/+dS1YsEA33HBDUc+cb33rW3r22Wd1+umn68knn9S8efN06aWXatmyZdp///01Z84cnXbaafqf//kfvfTSS7rzzjvzvY1OPvlkvfDCC/r1r38ta62ef/75/K3st2TWrFn6/ve/r7/+9a966KGHdO655+YvcZOkc889Vx0dHTryyCP1wAMP6KWXXtK1116rBx54IL/OmWeeqbvuukvPPfecTj311BF6xaqPYAgAAAAAAIyIcePG6X//93/10EMP6cADD9RNN92kq6++Or98//3316OPPqoVK1boox/9qI455hi99tpramtrkyT96le/0nnnnadvfOMbmj17ti699FJ1dXVJkvbee2/dcMMNuvzyyzV27Fh98Ytf1He+851htesXv/iFXNfVoYceqgsvvFDf/va3NXbs2Pzy1tZWPfroo9p77731yU9+UocccohuuukmTZ06Nb/Ofvvtp0mTJun444/Pt3dnYMKBlmrB7NmzbThaOQAAAAAAO7uFCxdqr732qnYzMAzZbFZdXV268cYbh9z1rNq2tB0ZY+ZZa2eXW0aPIQAAAAAAgC2444471NLSoo9+9KPVbsqI4q5kAAAAAAAAm/Dmm29q5cqVuuSSS/TNb35TsVis2k0aUfQYAgAAAAAA2ITvfve7OvbYY3XmmWfqn/7pn6rdnBFHjyEAAAAAAIBNuPrqq4sG0N7Z0GMIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAYMScffbZOuOMM6rdDAwTwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAjJr7779fhxxyiJqbm9Xe3q5TTz1VS5YsKVrnuuuu01577aXGxkbtsssueuSRR4a1DNsvXu0GAAAAAACAndNtt92mOXPm6JJLLtHVV1+tnp4eXXbZZTr88MP1/PPPK5VK6ZZbbtFXv/pV/epXv9Kuu+6q+fPny3H8fiybW4aRQTAEAAAAAECteeCr0ooXq9uGyftKH//2Nv+4tVYXXXSRLrroIl1++eX5+oMPPlgzZ87Uz372M11wwQV69NFHddhhh+n444+XJO299975dTe3DCODmA0AAAAAAIy4119/XYsXL9bZZ59dVN/a2qrjjjtOTzzxhCTpkEMO0YMPPqif/OQnGhwcLFp3c8swMugxBAAAAABArdmOnjq1YtWqVZKkrq6uIcumTJmSH2foU5/6lAYHB3XZZZfpiiuu0MUXX6wvfelLchxns8swMnglAQAAAADAiGtvb5ckLV++fMiylStXqqOjIz9/zjnn6I033tCVV16pb33rW/qv//qvYS3D9iMYAgAAAAAAI27PPffU5MmTdcMNNxTV9/f36/7779cxxxxTVJ9MJnXOOefonHPO0WOPPTbsZdg+XEoGAAAAAABGXCwW01VXXaXPfOYzisfjOumkk7R27Vr9+7//u6ZNm6YzzzxTknTFFVdo6tSpOvDAA7Vu3To9+OCDOuuss7a4DCODYAgAAAAAAIyKs88+W01NTbryyit11VVXqb29XZ/4xCf07W9/W8lkUpI0ffp0XX755Vq6dKmmTJmif/zHf9RXvvKVLS7DyDDW2mq3IW/27Nl27ty51W4GAAAAAAAVsXDhQu21117VbgZ2cFvajowx86y1s8stY4whAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAAAA6hTBEAAAAAAAqAkzZ87Uz372s2o3o64QDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAACwnTzPq3YTtgnBEAAAAAAAGBHHH3+8zjjjjCH1H/rQh/TVr35V1157rQ444AC1trZqxowZuuqqq7b5uW677TYdeuiham9v15QpU3TBBRcom80WrXP33Xfr/e9/v5qamjRu3Dh99atfzS9bs2aNPve5z6mzs1ONjY3ad999tWjRIkmSMUZ/+MMfin7Xt771LR1xxBFF8x/84Ad1++23a8qUKTr66KMlSX/961/1iU98QpMnT1Z7e7tOOukkrVixouh3LViwQCeeeKLa29vV0tKiI444QsuXL1c8HtfDDz9ctO6aNWuUTCb19NNPb/NrtTkEQwAAAAAAYETMmTNH9957rzKZTL5u2bJleuqpp3TmmWfql7/8pS6++GLNnTtX3/rWt/T1r39dv//977fpuX7+85/r3HPP1dNPP62rr75a119/vX7+85/nl99444065ZRTdMwxx+jJJ5/U7bffro6ODklSf3+/jjjiCD355JO6/vrr9dxzz+mf//mflU6nt6oNa9as0U9/+lPddddd+t73vidJuummm3TAAQfo97//vX7/+9/r9ddf15e+9KX8z7z88ss65JBD1NjYqHvuuUePPfaYDj74YE2ZMkVHHnmkbr311qLnuO2227TLLrvo4IMP3qbXaUvio/JbAQAAAADANvvPZ/9Tr6x5papt2HPcnrr4oIu36mdOOukknXfeeXrkkUf08Y9/XJJ0yy23aP/999d73vMePf7440okEv7v33NP3XjjjXrkkUf0sY99bKvbd9dddxX9rlNPPVWPPPKIzj//fA0ODurf/u3fdMUVVxT1Egp7/Fx99dVauXKlXnvtNaVSKUnSvvvuu9VteOWVV3T33Xfr3e9+d77uqquuyrdLkr7yla/owgsvLJo/9NBDdfPNN+frDjjgAEnSGWecoYsuukj//d//Lcfx+/L89re/1VlnnbXVbRsugiEAAAAAADAiWltbddxxx+n2228vCobCYCORSOi1117Tc889p0WLFmnJkiXq6urapudKJBJ6++239fTTT2vRokWaP3++WlpaJElPP/20enp69LnPfa7sz957770644wz8qHQttp1112LQqGwXevXr9eTTz6pV155RY8++qhWr14t13Xluq4efvjhIb2CQieffLL++Z//WY899piOOOIIrVy5Uo899ph+8YtfbFc7N4dgCAAAAACAGrO1PXVqyemnn67zzz9f11xzjVasWKHnnntOt99+u3p6enTyySfr5Zdf1kEHHaTddttN7e3tstZu9XNkMhnNmTNHjzzyiN7//vdrt91207hx4zQ4OChJWrp0qcaNG7fJ4Gfp0qU65ZRTtuv/KUkTJ04cUnfVVVfpiiuu0P77768999xT48ePlyRZa9Xd3a1MJqOZM2eW/X0tLS066aSTdOutt+qII47Q7373O33oQx/SjBkztrutm0IwBAAAAAAARsyxxx6rdDqtxx9/XPPnz9dRRx2liRMn6stf/rIymYzefvttNTY2SpJOO+20bXqO66+/Xn/961/11ltvqb29XZJ/idZTTz0lSWpra9O6deuUTqfV0NAw5Ofb2tq0cuXKTf7+hoaGfMgU6uvrG7JeeLlX6KWXXtLXv/51zZ07V+9973slSffff79++ctf5p9XklauXLnJS9fOPPNMffrTn9aPf/xj/fa3v9VnPvOZTbZzJDD4NAAAAAAAGDGNjY066aSTdOedd+q2227LX0a2YMECHXzwwflQKJ1Ob/OdthYsWKD9998/HwpJ0qOPPpqfPuSQQxSLxfSb3/ym7M8fdthh+u1vfzvkLmahqVOn6vXXXy+qe+aZZ7bYrpdfflmpVCofCknSn/70p/x0KpXSrFmzdOONN27yd3z0ox+V53m6+eab9cILL+jv//7vt/i824NgCAAAAAAAjKg5c+booYce0muvvaYTTjhBkjRr1izdfPPNeuSRRzR37lzNmTOnaJDmrTFr1iw9/PDDuuOOOzR//nx99rOf1erVq/PLOzo6dOGFF+oLX/iCfvSjH+nFF1/UQw89pMsvv1yS9OUvf1nr16/Xscceqz/+8Y96/vnn9d3vflfz5s2TJJ1yyin68Y9/rMWLFyuXy+mHP/yhXn755S22a5999tGGDRt05ZVXasGCBfrBD36ge++9t2idq666SjfddJO++MUv6rnnntPTTz9dNBZSLBbT6aefrgsuuEAnnXSSWltbt+k1Gi6CIQAAAAAAMKKOOuoodXd364QTTsj3EPra176mI444QieffLJOPPFEHXXUUfrwhz+8Tb//3HPP1dlnn61zzz1XH/3oRzV58mSdffbZRetceeWVuuKKK/SjH/1Is2fP1uc+9zlNmDBBktTV1aUnnnhCTU1NOuGEE3T44Yfrj3/8oyZNmiRJ+uY3v6nDDz9cBx54oKZNm6bFixfr/PPP32K79t57b11zzTW65pprdNBBB+nPf/6zvvvd7xat8/GPf1z33nuvnn76aX3wgx/USSedpHQ6XbTOpz71KS1fvnxU70YWMtsyyNNomT17tp07d261mwEAAAAAQEUsXLhQe+21V7WbgRpzxx136Atf+IKWLFkyZByjcra0HRlj5llrZ5dbRo8hAAAAAABQUx577DE1NjaWfYz2YMy14Prrr9enP/3pYYVC24u7kgEAAAAAgJoye/ZszZ8/v+yyMWPGVLYxFfTCCy/oySef1HPPPacbbrihIs9JMAQAAAAAAGpKc3Oz9txzz2o3o+IOOeQQzZgxQ3feeac6Ojoq8pwEQwAAAAAAADWgr6+v4s/JGEMAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAANWHmzJn62c9+Nqx1jTH6wx/+MMot2vkRDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQpwiGAAAAAADAiDj++ON1xhlnDKn/0Ic+pK9+9au69tprdcABB6i1tVUzZszQVVddNWLPnU6ndemll2qXXXZRMpnULrvsoiuvvFKe5+XX6enp0TnnnKNJkyapublZH/jAB/LLN7dsZxavdgMAAAAAAMDOYc6cOfr85z+vTCajZDIpSVq2bJmeeuopXXPNNTrnnHN08cUXa7/99tOf//xnnXvuuXrve9+rj33sY9v93Keddprmz5+v//qv/9Kee+6pZ555RhdeeKF6enr0ve99T5J01llnyXVdPfjgg8pkMrrvvvvyP7+5ZTszgiEAAAAAAGrMiiuvVHrhK1VtQ8Nee2ry1762VT9z0kkn6bzzztMjjzyij3/845KkW265Rfvvv7/e85736PHHH1cikZAk7bnnnrrxxhv1yCOPbHcw9Mc//lH33nuvXnzxRe29996SpH333VctLS0666yzdPHFF2vixIl69NFH9etf/1r777+/JOmggw7K/47NLduZcSkZAAAAAAAYEa2trTruuON0++235+tuueUWnXXWWZKkRCKh1157Tb/5zW902WWXacmSJVqxYsV2P+8f/vAHHXTQQflQKHTqqafKWqunn35aknTIIYfosssu01NPPTXkd2xu2c6MHkMAAAAAANSYre2pU0tOP/10nX/++brmmmu0YsUKPffcc7r99tvV09Ojk08+WS+//LIOOugg7bbbbmpvb5e1drufc9WqVerq6hpSH4/HNWHCBK1du1aSH1JddNFFOvzww3XggQfq+9//vg4++OAtLtuZ0WMIAAAAAACMmGOPPVbpdFqPP/64brnlFh111FGaOHGivv3tbyuTyejtt9/Wfffdpx/84AfaddddR+Q529vbtXz58iH1ruuqp6dHHR0d+fV+9rOf6c0339Qee+yhD3/4w1qyZMkWl+3MCIYAAAAAAMCIaWxs1EknnaQ777xTt912W/4ysgULFujggw9WY2OjJP8uYuElXtvrwx/+sJ555hm98krxuEx33HGHGhoadOihhxbVT506Vb/4xS/U1NSkZ599dtjLdkYEQ6Nh6Tzphd9VuxUAAAAAAFTFnDlz9NBDD+m1117TCSecIEmaNWuWbr75Zj3yyCOaO3eu5syZkx+Ient9/OMf12GHHaaPfexjuvXWW/XSSy/pl7/8pc477zxdfvnlGjNmjCTplFNO0cMPP6xFixbp5z//uQYHB3XAAQdscdnOjDGGRsMLt0jzb5L2O63aLQEAAAAAoOKOOuoodXd36xOf+ES+h9DXvvY1vfXWWzr55JPV2tqqSy+9VOPHj9fg4OB2P58xRvfcc4++9rWv6Qtf+IJ6enq0++6763vf+57+6Z/+Kb+e4ziaM2eO0um09ttvP91zzz3aZZddtrhsZ2ZGYpCnkTJ79mw7d+7cajdj+z35I+nhb0hffVtqTFW7NQAAAACAGrVw4ULttdde1W4GdnBb2o6MMfOstbPLLeNSstGQ6vTL3qEDXwEAAAAAgM177LHH1NjYWPbxmc98ptrN26lwKdloSAW3yNuwVOrYo7ptAQAAAABgBzN79mzNnz+/7LJwvCCMjGEHQ8aYYyRdJWlvSYslfdNau8URlo0xl0i6UtJu1trXt7GdO5awx9CGZdVtBwAAAAAAO6Dm5mbtueee1W5GXRjWpWTGmAMk3Snp15LeK+mXkn5jjPnAFn6uVdIF29nGHU/bFL8kGAIAAAAAADVsuD2GviHpN9ba7wfzC40xB8sPfZ7aws/9RdLR297EHVA8KbVM9C8lAwAAAAAAqFFb7DFkjInLD3ZuLll0u6QPb+bn9pJ0tqRLt6N9O67UFHoMAQAAAACAmjacS8l2kdQkaUFJ/SJJ44wx7aU/EIRJ/yPpckmrt6+JO6hUF8EQAAAAAGCLrLXVbgJ2YNu7/QwnGJoQlGtK6tcGZarMz3xD0kZJV29ju3Z8qU4uJQMAAAAAbFYymdTAwEC1m4Ed2MDAgBKJxDb//HCCoVhQuiX1tqSUJBljPiLpc5L+0Q4jtjLGnGeMmWuMmdvd3T2M5uwgUp3SwFop01/tlgAAAAAAatSECRP0zjvvaM2aNcpms/QewrBZa9Xf36+lS5dq4sSJ2/x7hjP4dG9QplR8WdiYoAx7DskYM0PSbyWdZ619ZzgNsNZeJ+k6SZo9e/bO8xeQ6vLL3uXS+HdVty0AAAAAgJo0ZswYNTQ0qLu7Wz09PcrlctVuEnYgiURCkyZNUipV7mKu4RlOMPSGJE/SHioOhvaQtMRauzFSd7akDkm3G2NKf89rxpgbrLVnb3NrdySpTr/csJRgCAAAAACwSY2NjZo2bVq1m4E6tcVgyFrba4x5VtLJkp6MLDpZ0n0lq18r6a6Sus5gvRMkPb/tTd3BhD2GGIAaAAAAAADUqOH0GJKkqyTdYoxZJOkxSSdJOkrSLGNMh/ww6BJr7aOSVkR/0BizLphcaK1dMhKN3iG0TfFLBqAGAAAAAAA1ajiDT8tae7ekz0u6RNJ8SSdKOtpau1hSQtLuksaNThN3UMlmqWksPYYAAAAAAEDNGm6PIVlrr5d0fZn6ZSrc0r7czy2WNGTAobqQ6iIYAgAAAAAANWtYPYawjVKdXEoGAAAAAABqFsHQaEp10mMIAAAAAADULIKh0ZTqkvq6pVy62i0BAAAAAAAYgmBoNKU6/bJ3eXXbAQAAAAAAUAbB0GgKgyEuJwMAAAAAADWIYGg0pbr8kmAIAAAAAADUIIKh0ZTvMcSdyQAAAAAAQO0hGBpNDW1SQ4oeQwAAAAAAoCYRDI22VCc9hgAAAAAAQE0iGBptqU56DAEAAAAAgJpEMDTaCIYAAAAAAECNIhgabakuqXeF5Gar3RIAAAAAAIAiBEOjLdUpyUobV1a7JQAAAAAAAEUIhkZbqssvuZwMAAAAAADUGIKh0Zbq9EvuTAYAAAAAAGoMwdBoywdD9BgCAAAAAAC1hWBotDW2S4lmgiEAAAAAAFBzCIZGmzHBLeu5lAwAAAAAANQWgqFKSHXSYwgAAAAAANQcgqFKSHURDAEAAAAAgJpDMFQJqU6pd7nkudVuCQAAAAAAQB7BUCWkOiUvJ/V1V7slAAAAAAAAeQRDlZDq8ksGoAYAAAAAADWEYKgSUp1+yThDAAAAAACghhAMVUK+xxDBEAAAAAAAqB0EQ5XQPF6KJbmUDAAAAAAA1BSCoUowxr+cjB5DAAAAAACghhAMVUqqi2AIAAAAAADUFIKhSkl1cikZAAAAAACoKQRDlRJeSmZttVsCAAAAAAAgiWCoclJdkpuR+nuq3RIAAAAAAABJBEOVk+r0Sy4nAwAAAAAANYJgqFLywRADUAMAAAAAgNpAMFQpqS6/pMcQAAAAAACoEQRDldLSITlxegwBAAAAAICaQTBUKU5MaptCMAQAAAAAAGoGwVAlpTq5lAwAAAAAANQMgqFKSnXSYwgAAAAAANQMgqFKSnX5wZC11W4JAAAAAAAAwVBFpTqlbL80uK7aLQEAAAAAACAYqqhUp19yORkAAAAAAKgBBEOVlOryS4IhAAAAAABQAwiGKinfY4g7kwEAAAAAgOojGKqk1kmScegxBAAAAAAAagLBUCXFEn44RI8hAAAAAABQAwiGKi3VSY8hAAAAAABQEwiGKo1gCAAAAAAA1AiCoUpLdREMAQAAAACAmkAwVGmpTim9QRrcUO2WAAAAAACAOkcwVGmpLr/sXV7ddgAAAAAAgLpHMFRpqU6/5M5kAAAAAACgygiGKi0fDDHOEAAAAAAAqC6CoUprm+KXBEMAAAAAAKDKCIYqLd4gtXRwKRkAAAAAAKg6gqFqSHXSYwgAAAAAAFQdwVA1pLoIhgAAAAAAQNURDFVDqpNLyQAAAAAAQNURDFVDqlMaWCtl+qvdEgAAAAAAUMcIhqoh1eWXvcur2w4AAAAAAFDXCIaqIdXpl1xOBgAAAAAAqohgqBrCHkMMQA0AAAAAAKqIYKga2qb4JT2GAAAAAABAFREMVUOyWWoaS48hAAAAAABQVQRD1ZLqIhgCAAAAAABVRTBULalOLiUDAAAAAABVRTBULalOegwBAAAAAICqIhiqllSX1Nct5TLVbgkAAAAAAKhTBEPVkur0y97l1W0HAAAAAACoWwRD1ZK/ZT2XkwEAAAAAgOogGKqWVJdfMgA1AAAAAACoEoKhagkvJaPHEAAAAAAAqBKCoWppTEnJNoIhAAAAAABQNQRD1ZTq5FIyAAAAAABQNQRD1ZTqpMcQAAAAAACoGoKhakp1EQwBAAAAAICqIRiqplSntHGF5Oaq3RIAAAAAAFCHCIaqKdUpWU/auLLaLQEAAAAAAHWIYKiaUl1+yeVkAAAAAACgCgiGqinV6ZfcmQwAAAAAAFQBwVA15YMhegwBAAAAAIDKIxiqpqaxUryJHkMAAAAAAKAqCIaqyRi/1xA9hgAAAAAAQBUQDFUbwRAAAAAAAKgSgqFqS3URDAEAAAAAgKogGKq2VKfUu0zyvGq3BAAAAAAA1BmCoWpLdUpeTurrrnZLAAAAAABAnSEYqrZUl19yZzIAAAAAAFBhww6GjDHHGGP+YowZNMa8Yow5bRPrGWPMBcaYVyPrfm7kmryTSXX6JeMMAQAAAACAChtWMGSMOUDSnZJ+Lem9kn4p6TfGmA+UWf0ESadLukjS+yT9RNL/Z4z59Eg0eKeT7zFEMAQAAAAAACorPsz1viHpN9ba7wfzC40xB0u6QNJTJes+L+lQa20mmH/ZGPNuSf8k6Rfb2+CdTvN4KZbkUjIAAAAAAFBxW+wxZIyJSzpa0s0li26X9OHS9a21iyOhUOhVSRO3tZE7NceR2qbQYwgAAAAAAFTccC4l20VSk6QFJfWLJI0zxrQP43e8V9LLW9e0OpLqIhgCAAAAAAAVN5xgaEJQrimpXxuUqc39cDAO0ack/WATy88zxsw1xszt7q7TW7anOrmUDAAAAAAAVNxwgqFYULol9bakHMIYc4Kk+yV92Vr7aLl1rLXXWWtnW2tnd3R0DKM5O6FUp99jyG7ypQQAAAAAABhxwwmGeoOytGfQmKBcW1IvY4xjjLlS/t3LzrTW/nTbm1gHUl2Sm5b6SztlAQAAAAAAjJ7h3JXsDUmepD0krY7U7yFpibV2Y5mf+aWk/SS9z1r71na3cmeX6vTLDUullvHVbQsAAAAAAKgbW+wxZK3tlfSspJNLFp0s6b7S9Y0xcyR9RNJHCIWGKdXllwxADQAAAAAAKmg4PYYk6SpJtxhjFkl6TNJJko6SNMsY0yHpLkmXBOMInSTpcUljjDFjSn7PUmttdiQavlOJ9hgCAAAAAACokOGMMSRr7d2SPi/pEknzJZ0o6Whr7WJJCUm7SxoXrD5R0umS3izz2GPkmr4TaZ0omRg9hgAAAAAAQEUNt8eQrLXXS7q+TP0yFW5pL2vtR0amaXXEiUltUwiGAAAAAABARQ2rxxAqINXJpWQAAAAAAKCiCIZqRaqTHkMAAAAAAKCiCIZqRarLD4asrXZLAAAAAABAnSAYqhWpTinbJw2ur3ZLAAAAAABAnSAYqhX5W9ZzORkAAAAAAKgMgqFakeryS4IhAAAAAABQIQRDtSLfY4g7kwEAAAAAgMogGKoVbZMlGXoMAQAAAACAiiEYqhWxhNQ6iR5DAAAAAACgYgiGakmqkx5DAAAAAACgYgiGagnBEAAAAAAAqCCCoVqS6iIYAgAAAAAAFUMwVEtSnVJ6vZTurXZLAAAAAABAHSAYqiWpLr/csLy67QAAAAAAAHWBYKiWpDr9kjuTAQAAAACACiAYqiX5YIhxhgAAAAAAwOgjGKolbVP8kmAIAAAAAABUAMFQLUk0Ss0TuJQMAAAAAABUBMFQrUl10mMIAAAAAABUBMFQrUl1EQwBAAAAAICKIBiqNalOLiUDAAAAAAAVQTBUa1Kd0sAaKTtQ7ZYAAAAAAICdHMFQrUl1+SWXkwEAAAAAgFFGMFRrUp1+STAEAAAAAABGGcFQraHHEAAAAAAAqBCCoVqTmuKXDEANAAAAAABGGcFQrUm2SI3t9BgCAAAAAACjjmCoFqW6CIYAAAAAAMCoIxiqRalOLiUDAAAAAACjjmCoFqU66TEEAAAAAABGHcFQLUp1SX2rpFym2i0BAAAAAAA7MYKhWpTq9Mve5dVtBwAAAAAA2KkRDNWiMBha80Z12wEAAAAAAHZqBEO1aMr+UtM46d5/kzZ2V7s1AAAAAABgJ0UwVItaxkuf/J3Uu0L6zWlSpq/aLQIAAAAAADshgqFaNe1A6e9/IS2fL93yT5Kbq3aLAAAAAADAToZgqJbteax07H9Jrz0o3XeBZG21WwQAAAAAAHYi8Wo3AFtw4DnShqXS49+TxkyVDv9KtVsEAAAAAAB2EgRDO4KPfEPasEz64//z71j23jOq3SIAAAAAALATIBjaERgj/d2P/MGo7/6i1DpZ2u2j1W4VAAAAAADYwTHG0I4inpT+4UZp0t7S786Slv212i0CAAAAAAA7OIKhHUlDm/SpW6Xm8dJNp0pr3qx2iwAAAAAAwA6MYGhH0zZZOuNWyc1Kvz5F6uupdosAAAAAAMAOimBoR9Sxh/TJm/27lf3vP0iZ/mq3CAAAAAAA7IAIhnZU0w+WTvmZ9M5c6bZzJc+tdosAAAAAAMAOhmBoR7bX30kf/4606D7p/i9L1la7RQAAAAAAYAfC7ep3dO8/T9rwjvTkD6UxXdKHLqx2iwAAAAAAwA6CYGhncOS3pA3LpUcul1Jd0qw51W4RAAAAAADYARAM7QwcRzrxp9LGFdJd/yK1TpTe9ZFqtwoAAAAAANQ4xhjaWcST0j/8WurYU7r5LGn5C9VuEQAAAAAAqHEEQ6NgcNEirf3tzZV/4sYx0qdu8cub/l5au7jybQAAAAAAADsMgqFRsPam32jFt76llf/5HVnPq+yTpzqlM26VcoPSTw+WHrxU6l1R2TYAAAAAAIAdAsHQKJh82Tc19owztOaXv9TSf/2SvIGByjZg4l7SZ/4ovecT0tP/Lf1wlvTAxdKGZZVtBwAAAAAAqGnGWlvtNuTNnj3bzp07t9rNGDFrfvUrrbzq22rcd19Nu/qnik+YUPlG9PxNeuL70vO/lYwjve8s6YP/Jo2ZWvm2AAAAAACAijPGzLPWzi63jB5Do2jcWWdp6k9+rPSrr2rxP8xR+vXXK9+I8e/y71j2hXnSrNOleTdIP9xfuudL0tq3Kt8eAAAAAABQMwiGRlnbkUdqxo03ykuntfj0T6rv6aer05CxM6UTfiR98S9+r6H5N0k/fp901+elNW9Wp00AAAAAAKCqCIYqoGnffbTLzb9VYvIkLTn3M1p3x53Va0z7dOn470tfnC/NPkd64XfSjw+Q7vycf9kZAAAAAACoGwRDFZLo6tKMm25Sy0EHavkll6j7Rz9SVcd3GtMlHfsd6UsvSO//rPTSbdJPZku3nyd1v1q9dgEAAAAAgIohGKqgWCqladdeqzGnnKzVV/+3ln3lYnmZTHUb1TZZ+thV0r++IB3yL9LCe6SfHiTdeo604kWphgYnBwAAAAAAI4u7klWBtVY9116n7h/8QM2zZ2vqT36sWHt7tZvl61st/fkn0rPXS5mNUqpL2vUI/7HL4VLbpGq3EAAAAAAAbIXN3ZWMYKiK1t97n5ZfcokSXV2adt21Sk6fXu0mFfSvkV6+U3rjUenNR6WBtX79xL0LQdGMD0gNbVVsJAAAAAAA2BKCoRrWP2+e3vncv0iOo6k//ama3/feajdpKM+TVrwgvfEn/7Hkz1JuUHLiUtfsQlA0dbYUS1S3rQAAAAAAoAjBUI3LLF6sJZ/9rHLLV6jzO/+p1Mc+Vu0mbV52UHr7mUJQtHy+ZD0p2SrNOLQQFE3cSzKmqk0FAAAAAKDeEQztAHJr1+qdf/m8Bv7yF3VceIHGn3uuzI4SqgyslRY/UQiKel7361smStMO8nsSdc2WOvfn0jMAAAAAACqMYGgH4aXTWn7J17Th/vvVfuqpmnjxxYq1tlS7WVtv3dv+uERvPia985y05g2/3jhSx55S1wFBWHSA1LGXFItXt70AAAAAAOzECIZ2INbz1P3DH6nn2mvlNDcrdfzxaj/tNDXt855qN23b9a+Rlv5FWjpXemeutHSeNLDGX5ZoljrfK3W9z+9VNHW2fye0HaW3FAAAAAAANY5gaAc08PzzWvvbm7XhgQdkBwfV+J73qP2005Q67rgdsxdRlLXS2jeld+b5YdHSedLyFyQ37S9vnRSERAdIk2dJHbtLqamS41S33QAAAAAA7IAIhnZg7oYNWn/3PVp3881Kv/ZaoRfRP5ympvfswL2ISuUy0soX/Z5F78z1A6NwrCLJ71k0YTdpwu7ShD38sGjCHtK4XaV4snrtBgAAAACgxhEM7QSstRqYP1/rfndLcS+ifzhNqWN3gl5E5QyslVYtlLoXSatfLZTr3y6sY2LSuF2Kw6IJu/shUmOqem0HAAAAAKBGEAztZOqmF9GmpDf6vYnyYdEiqftVac3fJC9XWK+tU+rYQ5r0HmnyftLkffzQKJaoXtsBAAAAAKgwgqGdVNleRPvso/bTTt15exFtjpuV1rzpB0ZhWNS9UFr1SmH8oljSvzPa5H39x6R9/MCoaWx12w4AAAAAwCghGKoD5XoRNb3vfYpPnKj4xA7FJ05UYuLEYH6i4uPHyyTqpOeMm5VWvyatfEla8WLh0b+6sM6YaZGgaF8/LGqfyYDXAAAAALaKtVayVoZjCdQQgqE6ku9FdNttSi96VblVq5RbvVpy3eIVjVFs/HjFJ3Yo0REJjKJBUmen4mN30p401kobV0orXpJWvFAIjXpel6znr5Nskybt7YdGLR1Sy3i/bJ4QzE/wHw0pyZjq/n8AAAAAVIx1XWWXr1D27SXKvP22sm+/rcySt5V5e4myS96Wl04rOXWqktOnKzlzhhIzZig5fYY/PWWKTCxW7f8C6gzBUJ2zrit3zRplV63yg6JV3X7ZHZSrVinbvUru6h4/MIlIvutdaj7oQLUceKCaDzxQ8Y6OKv0vKiTT7w94vfJFPzRauUDqXS7190jpDeV/JpYMAqPxkcAoKJsn+PVN7VJje6FMNBEmAQAAbCObySi7apWyy5Ypt3y5ssuXK7e6R4nJk5ScOVPJmTOVmD5dTpK71442L5NRbuVK/z1YuVK5VavktLQq0TlFic5OJaZMkdOyYw5x4Q0O+oHP228rs8QPfDJvv63skiXKLFsmZbOFlRMJJbu6lJg+Tcmp0+Q0N/lB0VtvKbNkiezAQPG6U6cqOWOGkjNmKDFjemG6DkMja63s4KC8gQF5/f3y+vtl+/vlDQz4Pa8aGmSSDTINSTkNDf58Q4OcZFKmsXFUXi/rurKuK+VyUjy+U3yWEAxhWGwup1xPTz4sSr/xhvqfe04Dc+fJ6++XJCV32UXNBx2k5iAoSkyaWOVWV1B20L/8rC949K+W+rrLzHdLfT1Stm/TvyuWLA6KmtqlxjFl6qLlWH860UyohLrk9fUps3SpYmPaFZ/YIVOFvwN3/XpllryteMcExSdOpIs4UAe8TEbe+vVye3vlbdggd8MGuRt65W5YL29Dr9zeDbKDacXa2xWfMF6xceMVHz/O75k9fryctrYR+7yymYxya9fK7elRrmeNcj2r5fasUW5NjyQp0dWl5NSpSnR1KdHVJaexcUSet9KstXLXrfNDnxUrlF3mBz/Z5cuUC6Zz3d1DTmg6LS3y+iL7X46jRGdnPiiKPhJTJu/wB9/e4KDcNWuUW7NW7poe5daskduzRtZ15bQ0y2lpkdPSolhQOs2FOqelRSYe3/JzlIQ+2eUr/PdkRVCuXCm3p2eLv8cZM0aJKVMKjyA0ik8JygkTRv39sJmM//e7fr3c9Rvkrl/n/03n59f7jw3r5a1b729/3d3F/4+2NiWnTVNi2jQlp4fldCWnTVN88qa3KWutcqu6lXlrsTJvveUHS4vfKoRGg4P5dU0iocS0aUpMm6pYa5tMY4OcxiY5TY0yDY3ly8YmOY0NQ0rrevL6+vxHf39xWa6uZJlNp6V4TCaekInHZRJBGY9LiXihPh6XScSleHGd4jHZdEbeQBj2BMHPwEA+BIoGQNssCG5MPjRKykkG04lEPuCxuZwf+OSyUs4NpnMly/z5aHs6LrhAE877zLa3r0YQDGG72FxOgwsXqv/ZZ9X/7HPqnzdP3saNkqTkjBl+UHTQgWo+6CAlJk3avudyXf8De+06/8N640Z5/QPyBvplBwb86cGBwnT4oTIQfNCUzNtsVqaxUU5jo5zmpuBDslGmqVFOU3NhurGp/PKmRjlNTXKam2Wam+U0t/hftM3NMsnk5nf0Mv1+WNTfIw2skwbXDaNcKw1ukLSZv0snMTQsKhcghXVNY6XWSX5ZR4GSl077OzArVii3cpW8vo2KpVKKjR3rP9rbFRs7Vk5DQ7WbWrO8wUFlly4Nuke/o+w7frfoeEeH/5g4sTA9fvx279CFO03Zd/yu2OEZuuzbbyvzzjtyVxfGBYuNGaOG3XePPHZTw267j9ig+zaTUfrNN5V+9VWlX31Vg4sWKf3qa8qtWJFfxzQ0+DuF02coOW2akjOmKzHdP+OXmDx5WDvcGHnWdWXTadlMRl4mIxt9ROvTGX/H0PNkc67kubKuF5SuX++6kuvJeq7klix3Pdng0mP/u8D449IZScb4dWF9OG1UqDeOv1487n+fJJMyyYRMMunv3CaT/g54fllY568XrqN4fLPfRVvcz/M8eRs3yu3dKK9vo7zeXj8A2bjRL3s3ytvYm592NwZ1vb1yN26Ut3GjnJaWwoHStKlKTC0cMMXGjh3xENe6rn8iK+j57Pb0yMtkJM9KnidZT9az/nvleX699fxp1ytMh/WuJzs46O9/9G6Qt35DUQhk0+nNtic8EPE2lO9ZbBIJPyQaVwiLYuPHKT5ufD5Iio0dKzs4oFzPGv8Af3WPcmt68qGP27NGuZ6ezT6HJNlo7wVJsQkT/B4MXV1KRAKj5NQuxTs7t+oMuJfJBK9Jr7zeMCDb4G8L6zfI690gm80Gr32Z1zmc9oK/nTLTXl9fEAAtLzpQDl/nMFCIT5mixJTOQsAwZYrikyfLaWyU29vrH3AvXjzkEQ2NTDKp5IzpxYHRLrvk9w38HgrJ/IHlaJ2MsJ435HMqt26d3DVr/NCnZ43ctWvyoU9urV+6a9bkT9xuK9PQUBQUhcGRicX8Kwo2Efo4qZQSkycrPnmSEpOn5MvE5EmKT56i+MQOeRs3BoHeskKvrnzAt3zothyPKzFpkv9edk5RbEy7f2AebiPWDt2WrI38vfvrhNuTzeUKQe769f7f8hZeLyeV8vcVx4xRbExK8UmTg8+y6YXPtPb2kf9Ms1a5Vav87XbJW8q+FQRG7yyV198nOzAoL52WHRiQzWRG9Lkl+cdKQ8LD4LinISmb88MS/5GVsjn/bz1fl5NyWdlsrrgum5V1XX87C46nnKYmOS3NMk3Nhfl8XbhOsKzZXybHkU2n/dcgnZHNpOUNDhamw/rBQXmZYDqdDtZL+8eDsVgQVAUhVizmB1mxYD4e85fFCoGWCecTcTXPnq2mWbNG/LWvNIIhjCjruhpc+IofFD33nPrnzpXX2ytJSkyfnr/0rOmAAyQZuev8kMddF32sj9StD9ZZ739JDGObNA0NfoAT/UBpapJpbgoCnSDoSSTyH6TewKAfKvUPyBsMpgeCLouDg373zq35e3Cc4EOr+GFaoh9ohbY4zcEHXlPhg87/AGzJzztNTf4OSLa3bHjk9fbIW7NK7jp/J8Hb4J/BdHv75PUNyh3IyMsYuRlHbtbJTxvHykk6ijUn/Q/6tpRiqXY5YyfIGT9ZsfGdciZMlTNmjGJtbXJaW+W0tirW1uZ3z6yhQMlaK2/DBmVXrvS7K69cqeyKoFzph0C5lSvlrls3rN9nmpoUG9uueHshLMqXY9v9M8Bjx8oZM8bf0cgWdtwKB53ZyE5d6QFptnBAGv0SzWb8sxNDvlyz/o59+OUaWSbJP7CYMF7xCR2KT5gQPMYrlp/2H05T07Bey1x3t7LvvBOEMO/kQ5js228rt2rVkNfKaWgo/9o6juLjxw8NjCYWz8fa2pRdudLvjv32O/51+GH5ztLiAwFjFJ8yWclp05WYNtUvu7rkrl2bD2zSr71WtGOc6OoqCosad99dyZkzNznYvrVWuRUr8sFP+tVXlV60SOk33/TPFklSIqGGXXfN/87kjBly16xR5q0lQbfyt5RZ8nbxwWO0O/n0Gf7ZxDA46uryD+ilwnscbi/hdDZb2IayWX87yha2J8kWQuvwc6Ql8jnU0LBdf7fW8yJnDcucWezrk/Xc4AxhsnAWMRE5m5hIFM4cJkvONCYSkuP4n799kbOU/X3FXcj7+/3l/eUfdnAw/9qFr9eQMfVGU/ga19C+1KhIJBRrbZXT1pYvnbZWxVr97wuvt9f/3FiyZOjZ9ebm/Jnvwt/ytHxIEQ0m8oHPqm7luoNL37tLLn/v7vbHTvS8bf//GD/EM47jh3mOIyeZ9L8DUynFUm1yUsH3YapNsdQYv64tpdiYVFDvH0A6bW35Eww2m/V786zxQ5ywV8+QoCdYtqUDvFh7+yYDpdj4cf5n7rhxik2Y4F+mY61y3auVXfqOskuX+sH+O8H0O0uVXb688LkWvA7xiROD0KhL8Y4O2YFIQLah1w9+NvhBWWlQM+RlTSRkGhr81zZ4jf2HkTGRaSc2ZB3j+GGpaWws6lGSD4A6p2x3yGitlbt6tTKLFyudD4uCAGnJkuJLgsr9/6K9EZKJQm+EZLLQOyGZ9HtIDNk3yAz5nPeCsug92ZREwn+vx41TfOzYYLsY64eK48YWlgWliccLn9sln91u2fqSXiLZrD8OaZnQJzFp4ohcFuZu3Fh0GWC0V1h22TJ5G3qlWKwQpke3pfBvOLqdGQ3ZzpxUm2Jj2ovCHmfMGH86NUax9jH5ZU5b2w7Rgyw8+eENDvpBSKT0Bgb8ZQMDsoNpeYMDMrF4ofdYuK8QDYGam3eI/zdGBsEQRpV1XaUXLVLfs8+q/7m5flC0fv1mf8ZpbfV3eNrb/Q/ncDqcDw7IndbWojTZBL14RuU6UmsjH6aDfpCU76kUOUDJzwdfnuGBTfSAJfoz/f1b3NkoZRojPZWSSf/M7XDOWjY2BjvtzYo1N/pBUIMjm+6X17sh2CEYlDeYk5ex8nLDuAwm5ijW0uzvDIfvzZhwZ2Tc0PcumN5SMOFlMn6PsPCsc3hWehPTudU9QfizsuzOaWzCBP/Oe5MnKz5pohKTJik+abISk/w6p6XVD9LWrfN33NcGweTatfkyty6cXrfJs7LDZkyZM/2JogPjfNfbZOQAejP18jz/rGH3auVWr/YPLtasKXtQ6rS0KD5hQlFgFBs3Vu669UH483b5IGbSJP8yhPyBnH8Al5w2TbHx42WM8S9lWL3aP0Dr7vbHL+uOHsD5y9yeoeOWDXmZmpuDngZTlZw6zQ9SpvmPRGdnPkDZFOt5yi5bVgiKXn1Vg6++qsybi/MBgUkklIwEO7HWVqVfe02Dr76q9KuvFb3X8SlT1BgGS3vs4fdEmjlzWO3IrVoVBEVL8qFRZskSZd96q/isruPIJJP+AcH2HNxujuMUzs6VC6wbG+Vl0pEDg+AzLTg4KBoHoZqMGRq850P3oNdn9G+sIRnpcdMwZJlT9DfZ4J8tdBz/781x/IOQoJTj+N81QWlisaHLSw5SrS2ctQ4f1l8w5BHescbmcuUPHssFzEUhYljmttwbdHPLjf954Z8UaFOsLQh+WlvzIdDWBI3ewECkp2E0cPZ7HhZ9jwXhb2xMu9zgM63c30Rs3LhCwDyx0GMxEdZNmOAHEuFBZD5sKAQ/xhj/fQt7c1WZtdb/Tl692u8JsmaNTFOTH/aMH+/3Shjhu8da1/V7gbzzjjJhWLR0qX9yYOlS5bq7/YPFlB9+xcakghNJQSiWSvkH2mFdJByLjRmzQ/fAtbmcssuXK/Pmm3LXbyj0Qgh7F4a9D4rmM4XeCfkTRmkp55Z8/0d6AiaSQ/cNSnsKJpPBCapx/iWJ48bJaW2tie0WwI6NYAgVZT1P6Vdf1cD852US8eLgoN1P7Ud6Z6fW2Wy2cJlbX38hcAovievvL75cLn85XL+8TCbYOY/slAVnL/35VL7cqkHRMv2ya5fIW/6avBVvyu1eIm/1Urk9K+St7Za3Ya3cjPxeR1lHbtrxeyJlTH7abiZYMnGjWFNMseaEnKaEbNbKS7tyB3PyBnOyuS0fEJtETE5zg2JNDYqNaVO8Y5wf+EyeokTndMWnzlCia7o/1ssIDwhnczm/6/Hatf5j/XrJieV34pyGhjI7d8lhX94xku3MrVmTP6gKQyO3Z3UhQAqWeevX+2fvp0/PX+7hBzLBdFfniO7YF8YtK4RGbu8GJSZP9sOm6dP9M5uj8Dp5mYwyb7xRCItee82/FGz5ckl+OJ2/BG333dW4xx5q2G03xVKpEW+LtTbSw8gfV8BLp/NhoVN6yVA4nS+DA4dEYbmM/AB7SGhdpsdN/8DQnjYDA37Py+bmMt3HhzdtYk6hp1smW+jtFu3plu8RV2aZ6wY9JoOgJ39GMxIE1ViPRWwf63l+j5Z3wsFb/cDIXb/eD7BLA5+JE/3LVHeCAT8BAKg2giEAW89zpY2rpPVv+wNqZwek3GBR6fVvDK7d9i9nc3v75G4ckLtxUG5fxn8M5OQNujJxKZbw5CSsX8Y9OQlXsbgrJ+7KiecKy5JWsbgnM5yOYU7CH7g7P4B39FGmriElNbRJjSl/Otnqn1GuAzabrVhgVavcDX7PufjkyXX9OgAAAKC+bC4YYmRMAOU5MSk1xX9sapXgMSL9v6w/UKisK3m54OEGj5yU7Y+MubQ+eESmo/XrlhTqvC1dxmf8gCgMiqKhUWmZbJHijf6d4RJNkTKcDpbFkjU50He99dQrJxb0sAMAAADgIxgCUBuMkWJx+R9LI3Q5k7V+D6cwMEpv8O/6ll4vpXuD6Q1Dy40rpdWvFeq2GC6V/l+c4tAoHpaNUiwRPJJ+6YTT8aBMSk5kOqx3En69l/Pb4+UkNzqdLQRqQ6azhYAt2erfoa55XHAHu3HF0+Fd7RwGIgQAAADqAcEQgJ2XMVKy2X9spufTZlnrXzqX7vV7LWUHImU4PVh+WS6yTqZfctN+mJMdkD+AU84v3Yxf72YKQU64fDicRCFkcmKbmI7785k3pIG1/sNuapwnE1yeVyZAamiTYg1SPHjEkn7gFU8G9dHp0nUaIoFYXDKxurmMDwAAAKhVBEMAsDnGFHr+VJrnFYdHnuv3IHLihTAovG3rtvzu9Ho/IOpfKw2sCabXDJ3u7/F7UA2s9XtRjSTjREKiILxy4pFHmXljInc7s0WFZMssK5k3ThBaNflloql4Pt7oXxYYjzzy88Fy65Vc8pjbzHyZukSL1DJBaumIPCb442IRlgEAAKCCCIYAoFY5juQk/R44o/G7m8b6j3Fb8XPW+r2a3LSUywTlYGQ6eBRNZ4rXGVag4m4iYIn2ogoCsXwwVjq/iWWeG7RtUMpslPpX+72+cmm/l1e4zM1s/eu6OdEQLDeoSJoVWScWCYwmFIdGzZH5hjb/tcwOBj3TBoPXODJAfG5w08vdjB8sJpojj6aSsbOCMtlcZlytZj+k25ZQLJy31h+3q6FVSrb5/6eGVj80G8lwzM0Gl44Gl5Cme/2AM90ryfg94ZrHFx7J5pF77uHyvGB8tfC1cQt/A0Pqvch0zv8/FL1/wWWrNTjOGQAAqE0EQwCA4TPGD6riyREbCqpmRQOkfOASzA+nZ1N0vvSyOTfn98bq6w4eqyPTkfm1b/nTmd5t/3/EGoIeT5EeUrGkHypk+iKXRfZt5vLCSjJBYNTmj4nV0BqUbUPrcumSsCccKyxSlxvcuqePNwUhURAYtUyIBEclIVJDyn/twufKbBwaPuUfG8vXuxmVDQm39zUMA6LS0CgM9vLLGv3QMvw5aTNha5nwVQrGRYuMnRaOi5avKxlHrWh8tfDS0i0FWZtZbkzhktXw/x1vpPcdAADDNOxgyBhzjKSrJO0tabGkb1prf7eJdSdK+pGkYyXlJP2vpAuttVu5dwYAQJU4scIYVSMtFpdaJ/qP4cgORMKj1X64UHSJW1NhOnpgvDUHx2FvsKLxskrHzgrKTL8fIjmxrQ/JwjrJD6bCwCQThCeZjUHdhsJ0ZqN/t8F0b6HOTfuBRkOb1DCmEBy1TpTGvyuYL73bYFtxvax/yWR/j99zrL8neKwpTK9d7M+n12/D+9wQec7gkeosDrniDZHwMHhEL60Me5oV1Ud6n1mvEFyG71FusHi8s6Ix0Qb8/1e0vuwlmCqZ38wlmp47SgHXdoolI4FoSThadJlokx92l94MIB9uJYqnS5fFkv775GYj48ZFp8vVZQs9KsP6WDLopddSpgdfmbr8es2FwNd6wXtjtzydn7fFobAxkkxJWa7eGVpXpGR7sOW2jzLrWG8T7fQKy4esE87b4nWK1vOKf+em1nES5XtPViJsDP+Won+fuZK/3yF/62WWGVMI0Rvagl6ZrZFgPdJLMwzeY9y5FKhnwwqGjDEHSLpT0qWSHpB0gqTfGGPesdY+VbKuI+leSaslHSH/IoVfyL+r9T+PVMMBAKgbiSapfZr/GC3R3mBN7aP3PCPFzQ6zp8kIyWWCsbd6CkHS4IZC76boIzwIi+/s3eoibBAQhYPnu9EyEogUDbqfLQy4v6XfvdnlXuGS1fDyyVy6/GWV0frBdZH6TND2oH3hHR9HnIkMxB/pQRVLFAcCmT7VXNCGwl1GN3f5bfheupngEupwOl0cFOaXpQtBoXW3rV1OItKOYBy8MFQf7mXRsYbioCjeELmRRHBjifyNJKL1kenoDSnynwm5kktkt+JyYydWJhiPnJDYbF3kxETYMzF8bHI+VrhhR6yk1+NIf9d4XnBzkj7/fcr0RR4bC8uir0v0MuOylx1HXz8vGJsyUdL7NiiLQsJW/7ssrI9t50VF4SXP0TbmQ1m3EMJ6bkloW64+vIQ68n0R3pU3vAtv/m68pcuCeesNvUlK6fa6yWWR7Xsnv2OvsVv6spVkjLlTUo+19pxI3R2SXGvt35ese5KkX0qabq3tDepOlHSLpEnW2rWbep7Zs2fbuXPnbsN/AwAAABhB0RsAhAca+YONSOAVHgDFEv7BRfSSulgQtobTwz2wsNb/3UWXe4a9QkrqMkHvPjcb9OJRcHlg0JOn7HRJb598j59I77Foj6KypTe0rrTX0JCD6TIH16XrGGdoO8PLHaPzZdcpmY+uW7TciSwv+X1etsxrvqm7kpapc7OFbSH/3jf4dfGGyLLIdH5ZQ2T8t5JLQYtCqciyeNPmD+RzmcJlrvlemJFLXIvqNhZCilw49l5k3L2iumBcvpG4BDnfM7KkV2l+/LVy4wxWWPTy13LhbvTOq+H7KEXCn5IAKNu/fe0xztBgzJQGaUFPxkxfMK7eMMPmeFNhDMB4U0noFA2lcmVCoCq+R6PpI9+QDruo2q3YbsaYedba2eWWbTEONMbEJR0t6aSSRbdL+kGZHzlO0gNhKBR4QJIr6YOS7tlykwEAAIAqGs0bAGyJMYUz1cD2iCel+Dh/jLTR4OYiN5yIBEfGKbmUeBN3Hy0dg29LhoRFpT1mor2Q3ELvPzdctqX50p4nJT2/yvUGi/aCzPRL7jp/WsFNFpIt/o0jki1+oJfvndMSmS5Tn2ge+vqFlx1vbQ8ma/0wKh/+RS7VTvcWgqswJAwv9c4NlOmJVe6OspvoxWUiZRjGhpdJh+FWOB3+v8rV53t2hZf3RufjxT29osvCnsVernAjlNxg8U1SSm+aUhSEBuWMQ7fu9d4BDaef2C6SmiQtKKlfJGmcMabdWrsuUr+3pPujK1prM8aYNyW9azvaCgAAAACoFbG4/0i2VOb5HEeSw5hIW8uYQuCkSdVuTeWFlwaStW/ScOLZCUG5pqQ+vCQsVWb90nXD9UvXBQAAAAAAQJUMJxgKL4YuHQ2t5PYUReuXGznNlllXxpjzjDFzjTFzu7u7h9EcAAAAAAAAjIThBEPhWEGlvX3GBGXpYNK9ZdYN1x/Sk8hae521dra1dnZHR8cwmgMAAAAAAICRMJxg6A1JnqQ9Sur3kLTEWruxpP710nWDAax3lfTyNrYTAAAAAAAAI2yLwVBwd7FnJZ1csuhkSfeV+ZE/SDrOGBMd2uljktKSntrGdgIAAAAAAGCEDffegFdJ+nwwHtCexpivSjpK0neMMR3GmKeMMYcH6/5a/hhDNxpj9jXGHC3pakn/Ya1Nj/j/AAAAAAAAANtkWMGQtfZuSZ+XdImk+ZJOlHS0tXaxpISk3SWNC9btk3S0pInyexr9TNKPrbXfH+G2AwAAAAAAYDvEh7uitfZ6SdeXqV+mwi3tw7qFko7Y3sYBAAAAAABg9Az3UjIAAAAAAADsZAiGAAAAAAAA6hTBEAAAAAAAQJ0iGAIAAAAAAKhTBEMAAAAAAAB1imAIAAAAAACgThEMAQAAAAAA1CmCIQAAAAAAgDpFMAQAAAAAAFCnCIYAAAAAAADqFMEQAAAAAABAnSIYAgAAAAAAqFMEQwAAAAAAAHWKYAgAAAAAAKBOEQwBAAAAAADUKWOtrXYb8owx3ZLeqnY7RsgESaur3QggwPaIWsG2iFrC9ohawvaIWsG2iFrC9jhyZlhrO8otqKlgaGdijJlrrZ1d7XYAEtsjagfbImoJ2yNqCdsjagXbImoJ22NlcCkZAAAAAABAnSIYAgAAAAAAqFMEQ6Pnumo3AIhge0StYFtELWF7RC1he0StYFtELWF7rADGGAIAAAAAAKhT9BgCAAAAAACoUwRDAAAAAAAAdYpgaIQZY44xxvzFGDNojHnFGHNatduE+mKMSRljfmaM+Y8yy84ItstBY8xfjTEfrkYbsXMzxnQaY240xqw2xqw3xjxijNm/ZB22RYw6Y8wBxpgHg21xnTHmPmPMniXrsC2i4owxjjFmgTHmiZJ6tkdUhDFmP2OMLfOYGVmH7REVY3xfCra5tDFmuTHmE5HlbI+jiGBoBBljDpB0p6RfS3qvpF9K+o0x5gPVbBfqgzFmrDHmQkmvSjq7zPK/k3StpKvkb59/knSvMWbXCjYT9eHHkjZI+rikIyWtlfSwMWaixLaIitpX/vfyR+Rvj5L0oDGmUWJbRFWdJmnvaAXbIypsnKT1knYpebwjsT2iKn4i6Z8lXSJpP0n/IOlNie2xEhh8egQZY+6U1GOtPSdSd4ck11r791VrGOqCMeZsSZdL+n+STpf0hLX265Hl8yXdZq29IlL3F0n/Z629qLKtxc7MGLOHtXZRZL5B0hJJX7PW/pxtEdVijJksabmkg6y1z7EtohqMMU2SFkh6XVKztfaDQf18sT2iQowxp0i6ylq7+yaWzxfbIyok6EjxgKQ9rLUryiyfL7bHUUWPoRFijIlLOlrSzSWLbpdENzdUwp2SdrHWXlu6wBjTJWmWhm6fd4jtEyMsGgoF82lJb0mayLaIKosFZTfbIqroUklPSnoqrGB7RBWMl7Sq3AK2R1TB5yRdv4lQiO2xAgiGRs4uksIzQFGLJI0zxrRXvEWoK9baddZadxOL95aUlfRaSf0iSe8a1Yah7hljmiXtLullsS2iCowxcWPMeyT9XNJ/W2sXi20RVWCMeZ+kz0q6uGQR2yMqbbykA40xG4KxXG4yxkwPlrE9otKOlDTXGPOrYFzAt4wxlxljHLE9VgTB0MiZEJRrSurXBmWqgm0BSk2QtNYOvXZ0raS2KrQH9eU/5V++c5/YFlFhxpgnJaUlvSS/59oFwSK2RVRUcAnZTZIutNYuK1nM9ohKu1XSEZIOk/SvkvaR9IQxZozYHlFBxpiUpMmSviLpb/LHBLxS0kWS/k1sjxURr3YDdiJh9/TSHhu2pASqIaah26bkb5dsmxgVxpik/IEEPybpo9banDGGbRGV9kn5g6y+S9IXJf3ZGPMh8bmIyvuppAXW2l+VWcb2iIqy1r6mQg+M+caYR+TfwOQs+QfcbI+olLADxaPW2n8Ppp8zxrTKD4e+LLbHUUePoZHTG5SlPYPGBOVaAdXTq/K91saIbROjwBgzTdLjkvaSP9Dvq8EitkVUlLX2LWvtX621t8ofC3CM/LuesC2iYowx/yJ/LIzPbGIVtkdUlbW2R/7YV/uJ7RGVlQnKB0vq/yi/J1FMbI+jjmBo5LwhyZO0R0n9HpKWWGs3Vr5JQN7rklqCwdui9pA/7gswYoIxCp6S9JikI0oGEmRbRNVYawcl/UX+mFdsi6ikiyTNlLTGGGONMVbSZZIODaaniu0R1ZeQP5YLn4+opG5JfRoa/lj5x9dzxfY46giGRoi1tlfSs5JOLll0svxxNYBqeln+GC+l2+cnxPaJkXedpLustV8uMyA62yIqxhhjSuYT8s+Gvy62RVTWcZLeW/K4VtL8YPpXYntEFRljJks6VNKfxecjKigYO+hPkk4pWfRx+Td2YnusAMYYGllXSbrFGLNI/pnykyQdJf/2ekDVWGutMebbkq4wxqyQPwjr+fK7Z15T1cZhp2KMaZH/uXedMWZmyeK0tXY52yIq6FFjzDWSXpB/B56LJDXIvyUun4uoGGvtkLPawXbXZ62dH8yzPaJigs/Gp+T3opwh/zjmb5Ju5vMRVfBtSX8yxvw/Sb+T9H5Jl0qaw/ZYGQRDI8hae7cx5vOSvibpR5L+Kuno4La4QLX9WP7I/T+Sf03uE5KOtNZuqGqrsLMZL7836m1lls2TNFtsi6icRyVdIalT/tnGByWdY60N7yDKtohawvaISnpL0v+TNEn+XZXvknSJtTYc74XtERVjrX3CGPP3kv5D/kmctySdb629J1iF7XGUmaF3fQMAAAAAAEA9YIwhAAAAAACAOkUwBAAAAAAAUKcIhgAAAAAAAOoUwRAAAAAAAECdIhgCAAAAAACoUwRDAAAAAAAAdYpgCAAAAAAAoE4RDAEAAAAAANQpgiEAAAAAAIA6RTAEAAAAAABQp/5/pll64bpoLLMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 오차 정도와 정확도를 그래프로 그린다.\n",
    "plt.plot(history.history['accuracy'], label='accuracy')\n",
    "plt.plot(history.history['loss'], label='loss')\n",
    "plt.plot(history.history['val_accuracy'], label='val_accuracy')\n",
    "plt.plot(history.history['val_loss'], label='val_loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}