CustomCNN2 (1).ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"CustomCNN2.ipynb","version":"0.3.2","provenance":[],"collapsed_sections":[]},"kernelspec":{"name":"python2","display_name":"Python 2"},"accelerator":"GPU"},"cells":[{"cell_type":"code","metadata":{"id":"q8epoVRfDQK_","colab_type":"code","outputId":"c6ef9efb-e94f-4915-bcde-4759b56c14c6","executionInfo":{"status":"ok","timestamp":1559892069238,"user_tz":-120,"elapsed":3426,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["import tensorflow as tf\n","device_name = tf.test.gpu_device_name()\n","if device_name != '/device:GPU:0':\n","  raise SystemError('GPU device not found')\n","print('Found GPU at: {}'.format(device_name))"],"execution_count":1,"outputs":[{"output_type":"stream","text":["Found GPU at: /device:GPU:0\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"N9HHs4owD4qB","colab_type":"code","outputId":"13a33ab3-c5ca-4274-f597-a8ca9519c67c","executionInfo":{"status":"ok","timestamp":1559892084295,"user_tz":-120,"elapsed":18451,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":151}},"source":["!git clone https://github.com/Eoli-an/influence-release customCNN\n","  "],"execution_count":2,"outputs":[{"output_type":"stream","text":["Cloning into 'customCNN'...\n","remote: Enumerating objects: 222, done.\u001b[K\n","remote: Counting objects: 100% (222/222), done.\u001b[K\n","remote: Compressing objects: 100% (211/211), done.\u001b[K\n","remote: Total 620 (delta 172), reused 61 (delta 11), pack-reused 398\u001b[K\n","Receiving objects: 100% (620/620), 113.98 MiB | 30.50 MiB/s, done.\n","Resolving deltas: 100% (448/448), done.\n","Checking out files: 100% (57/57), done.\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"T17cO8ejEQrz","colab_type":"code","colab":{}},"source":["import os\n","os.chdir('./customCNN')"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"yuBWxJ6HTz38","colab_type":"code","outputId":"d89fbd6d-3280-4dfe-b769-22f2edd3ea00","executionInfo":{"status":"ok","timestamp":1559890420419,"user_tz":-120,"elapsed":3127,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":202}},"source":["!git pull"],"execution_count":23,"outputs":[{"output_type":"stream","text":["remote: Enumerating objects: 7, done.\u001b[K\n","remote: Counting objects:  14% (1/7)   \u001b[K\rremote: Counting objects:  28% (2/7)   \u001b[K\rremote: Counting objects:  42% (3/7)   \u001b[K\rremote: Counting objects:  57% (4/7)   \u001b[K\rremote: Counting objects:  71% (5/7)   \u001b[K\rremote: Counting objects:  85% (6/7)   \u001b[K\rremote: Counting objects: 100% (7/7)   \u001b[K\rremote: Counting objects: 100% (7/7), done.\u001b[K\n","remote: Compressing objects:  25% (1/4)   \u001b[K\rremote: Compressing objects:  50% (2/4)   \u001b[K\rremote: Compressing objects:  75% (3/4)   \u001b[K\rremote: Compressing objects: 100% (4/4)   \u001b[K\rremote: Compressing objects: 100% (4/4), done.\u001b[K\n","remote: Total 4 (delta 3), reused 1 (delta 0), pack-reused 0\n","Unpacking objects:  25% (1/4)   \rUnpacking objects:  50% (2/4)   \rUnpacking objects:  75% (3/4)   \rUnpacking objects: 100% (4/4)   \rUnpacking objects: 100% (4/4), done.\n","From https://github.com/Eoli-an/influence-release\n","   b20569e..60a3b07  master     -> origin/master\n","Updating b20569e..60a3b07\n","Fast-forward\n"," scripts/load_mnist.py | 1 \u001b[32m+\u001b[m\n"," 1 file changed, 1 insertion(+)\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"ptpTHgq9H5kz","colab_type":"text"},"source":["# Training CNN and saving the influence of every training example(for the test image with test_idx)"]},{"cell_type":"markdown","metadata":{"id":"b1KMH6OfPT1P","colab_type":"text"},"source":["handling imports"]},{"cell_type":"code","metadata":{"id":"Jwijlos1PWB3","colab_type":"code","outputId":"45a30c2d-68be-4f22-e11d-5a8e9706c209","executionInfo":{"status":"ok","timestamp":1559892087578,"user_tz":-120,"elapsed":18750,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["from __future__ import division\n","from __future__ import print_function\n","from __future__ import absolute_import\n","from __future__ import unicode_literals  \n","\n","import numpy as np\n","import IPython\n","\n","import math\n","import copy\n","import pandas as pd\n","import seaborn as sns\n","import matplotlib.pyplot as plt\n","import sklearn.linear_model as linear_model\n","import sklearn.preprocessing as preprocessing\n","import scipy\n","import scipy.linalg as slin\n","import scipy.sparse.linalg as sparselin\n","import scipy.sparse as sparse\n","\n","sns.set(color_codes=True)\n","\n","import tensorflow as tf\n","\n","from influence.all_CNN_c import All_CNN_C\n","from scripts.load_mnist import load_small_mnist, load_mnist\n"],"execution_count":4,"outputs":[{"output_type":"stream","text":["Using TensorFlow backend.\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"4T6reorcP134","colab_type":"text"},"source":["loading the dataset"]},{"cell_type":"code","metadata":{"id":"833iet1mhjmL","colab_type":"code","colab":{}},"source":["import h5py\n","path_to_matrices = \"data/default_labeled_balanced.hdf5\"\n","dataset = h5py.File(path_to_matrices, 'r')"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"irE1Ly5MiGwU","colab_type":"code","colab":{}},"source":["matrices = np.array(dataset['dense_matrices'])\n","labels = np.array(dataset['label_vectors'])"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"iOItgUL0jLTx","colab_type":"code","colab":{}},"source":["randomize = np.arange(len(matrices))\n","np.random.shuffle(randomize)\n","matrices = matrices[randomize]\n","labels = labels[randomize]"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"3cnJzb_yu478","colab_type":"code","colab":{}},"source":["#from one hot to integer coding\n","i = 0\n","new_labels = []\n","for label in labels:\n","  index = np.argmax(label)\n","  new_labels.append(index)\n","  i+=1\n","\n","labels = new_labels\n","labels = np.asarray(labels)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"ydsj_Pz8kMLK","colab_type":"code","colab":{}},"source":["training_test_split = 0.8\n","index = int(training_test_split * len(matrices))\n","\n","train_matrices = np.expand_dims(matrices[:index], axis=3)\n","train_labels = labels[:index]\n","\n","validation_matrices = np.expand_dims(matrices[index + 1:], axis=3)\n","validation_labels = labels[index + 1:]"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"bw_C6gDNklif","colab_type":"code","colab":{}},"source":["from tensorflow.contrib.learn.python.learn.datasets import base\n","from influence.dataset import DataSet\n"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"gGPvQnuDkm50","colab_type":"code","colab":{}},"source":["train = DataSet(train_matrices, train_labels)\n","validation = DataSet(validation_matrices, validation_labels)\n","test = DataSet(validation_matrices, validation_labels)\n","\n","data_sets = base.Datasets(train=train, validation=validation, test=test)"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"nEFQ4rFoP3b5","colab_type":"code","outputId":"b1e247f8-8903-4458-fb76-d7dc7d17ad49","executionInfo":{"status":"ok","timestamp":1559890482568,"user_tz":-120,"elapsed":896,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":118}},"source":["data_sets2 = load_small_mnist('data')  "],"execution_count":12,"outputs":[{"output_type":"stream","text":["ok\n","('Extracting', u'data/train-images-idx3-ubyte.gz')\n","('Extracting', u'data/train-labels-idx1-ubyte.gz')\n","('Extracting', u'data/t10k-images-idx3-ubyte.gz')\n","('Extracting', u'data/t10k-labels-idx1-ubyte.gz')\n","(59500,)\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"s59NeNywoJcy","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":50},"outputId":"c216d13d-7560-4da2-eb8e-3321b3dae6ee","executionInfo":{"status":"ok","timestamp":1559892175572,"user_tz":-120,"elapsed":1263,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["print(data_sets.train.labels.shape)\n","#print(data_sets2.train.labels.shape)\n","print(data_sets.train.x.shape)\n","#print(data_sets2.train.x.shape)"],"execution_count":21,"outputs":[{"output_type":"stream","text":["(484,)\n","(484, 16384)\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"nPoovhz_P8-I","colab_type":"text"},"source":["defining the CNN"]},{"cell_type":"markdown","metadata":{"id":"OtpRfmVwQFQK","colab_type":"text"},"source":["training the CNN"]},{"cell_type":"code","metadata":{"id":"qGNV3H83P-kA","colab_type":"code","outputId":"beab5617-0a91-43f7-affc-6b768bf04018","executionInfo":{"status":"ok","timestamp":1559892185908,"user_tz":-120,"elapsed":7367,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":171}},"source":["num_classes = 4\n","input_side = 128\n","input_channels = 1\n","input_dim = input_side * input_side * input_channels \n","weight_decay = 0.001\n","batch_size = 50\n","\n","initial_learning_rate = 0.0001 \n","decay_epochs = [10000, 20000]\n","hidden1_units = 8\n","hidden2_units = 8\n","hidden3_units = 8\n","conv_patch_size = 3\n","keep_probs = [1.0, 1.0]\n","\n","\n","model = All_CNN_C(\n","    input_side=input_side, \n","    input_channels=input_channels,\n","    conv_patch_size=conv_patch_size,\n","    hidden1_units=hidden1_units, \n","    hidden2_units=hidden2_units,\n","    hidden3_units=hidden3_units,\n","    weight_decay=weight_decay,\n","    num_classes=num_classes, \n","    batch_size=batch_size,\n","    data_sets=data_sets,\n","    initial_learning_rate=initial_learning_rate,\n","    damping=1e-2,\n","    decay_epochs=decay_epochs,\n","    mini_batch=False,\n","    train_dir='output', \n","    log_dir='log',\n","    model_name='mnist_small_all_cnn_c')"],"execution_count":22,"outputs":[{"output_type":"stream","text":["WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n","Instructions for updating:\n","Colocations handled automatically by placer.\n","WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n","Instructions for updating:\n","Use tf.cast instead.\n"],"name":"stdout"},{"output_type":"stream","text":["/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gradients_impl.py:110: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.\n","  \"Converting sparse IndexedSlices to a dense Tensor of unknown shape. \"\n"],"name":"stderr"}]},{"cell_type":"code","metadata":{"id":"rD1q87PJqzCh","colab_type":"code","outputId":"44811557-0fb9-40f7-9a7f-7d937d0d6b73","executionInfo":{"status":"ok","timestamp":1559892576365,"user_tz":-120,"elapsed":389598,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":235}},"source":["num_steps = 5000\n","model.train(\n","    num_steps=num_steps, \n","    iter_to_switch_to_batch=10000,\n","    iter_to_switch_to_sgd=10000)\n","iter_to_load = num_steps - 1\n"],"execution_count":23,"outputs":[{"output_type":"stream","text":["Training for 5000 steps\n","Step 0: loss = 2496810496.00000000 (2.900 sec)\n","Step 1000: loss = 5504.71240234 (0.073 sec)\n","Step 2000: loss = 2858.06127930 (0.072 sec)\n","Step 3000: loss = 5210.56542969 (0.073 sec)\n","Step 4000: loss = 33484.22265625 (0.073 sec)\n","Train loss (w reg) on all data: 1686422.1\n","Train loss (w/o reg) on all data: 1686422.1\n","Test loss (w/o reg) on all data: 987032.94\n","Train acc on all data:  0.7107438016528925\n","Test acc on all data:   0.425\n","Norm of the mean of gradients: 60217984.0\n","Norm of the params: 188.62766\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"mR4M2LnKQOeL","colab_type":"text"},"source":["calculating the influence"]},{"cell_type":"code","metadata":{"id":"vJKVm1ZgQTgT","colab_type":"code","outputId":"07ba4c89-69e9-4372-dc24-9bb51942564f","executionInfo":{"status":"ok","timestamp":1559892841424,"user_tz":-120,"elapsed":636659,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}},"colab":{"base_uri":"https://localhost:8080/","height":487}},"source":["test_idx = 50\n","\n","CNN_predicted_loss_diffs = model.get_influence_on_test_loss(\n","    [test_idx], \n","    np.arange(len(model.data_sets.train.labels)),\n","    force_refresh=True)"],"execution_count":24,"outputs":[{"output_type":"stream","text":["Norm of test gradient: 729.28955\n","Function value: -324261942624.8125\n","Split function value: -324261412864.0, -529760.8\n","Predicted loss diff on train_idx 50: 0.0\n","Function value: -3125439112177.5\n","Split function value: -3125438578688.0, -533489.5\n","Predicted loss diff on train_idx 50: 0.0\n","Function value: -1.0545257967262898e+18\n","Split function value: -1.0545257967203123e+18, -5977413.5\n","Predicted loss diff on train_idx 50: 0.0\n","Function value: -2.828248746245584e+18\n","Split function value: -2.828248746238673e+18, -6910767.0\n","Predicted loss diff on train_idx 50: 0.0\n","Function value: -1.0717206665766371e+24\n","Split function value: -1.0717206665766344e+24, -2735750000.0\n","Predicted loss diff on train_idx 50: 0.0\n","Function value: -1.0717206665766371e+24\n","Split function value: -1.0717206665766344e+24, -2735750000.0\n","Predicted loss diff on train_idx 50: 0.0\n","Optimization terminated successfully.\n","         Current function value: -1071720666576637124411392.000000\n","         Iterations: 6\n","         Function evaluations: 38\n","         Gradient evaluations: 41\n","         Hessian evaluations: 9\n","Saved inverse HVP to output/mnist_small_all_cnn_c-cg-normal_loss-test-[50].npz\n","Inverse HVP took 158.165473223 sec\n","Multiplying by 484 train examples took 106.477807045 sec\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"XXrAiyFBQXle","colab_type":"text"},"source":["saving the influence"]},{"cell_type":"code","metadata":{"id":"jdhXLNryQbRs","colab_type":"code","colab":{}},"source":["np.savez(\n","    'output/CNN_results', \n","    test_idx=test_idx,\n","    CNN_predicted_loss_diffs=CNN_predicted_loss_diffs\n","    \n",")"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"dNKZfnfnII1I","colab_type":"text"},"source":["# **Loading influences and plotting the most influential pictures**"]},{"cell_type":"markdown","metadata":{"id":"W3VX16-wJO7x","colab_type":"text"},"source":["getting Xtrain of the Training Dataset"]},{"cell_type":"code","metadata":{"id":"EGk1yT69JSwh","colab_type":"code","colab":{}},"source":["X_train = data_sets.train.x"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"ZZkwD5jtIrxr","colab_type":"text"},"source":["loading the influences"]},{"cell_type":"code","metadata":{"id":"wNC8JCIgIWNJ","colab_type":"code","colab":{}},"source":["f = np.load('output/CNN_results.npz')\n","    \n","test_idx = f['test_idx']\n","CNN_predicted_loss_diffs = f['CNN_predicted_loss_diffs']\n"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"USVp-mUN5TLK","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":307},"outputId":"4e3d5634-033e-4d48-8657-70e85f539a71","executionInfo":{"status":"ok","timestamp":1559893460247,"user_tz":-120,"elapsed":1136,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["print(\"Test Image:\")\n","plt.spy(np.reshape(X_train[test_idx, :], [128, 128]))\n","print(\"Top 5 most influenatial matrices\")"],"execution_count":35,"outputs":[{"output_type":"stream","text":["Test Image:\n","Top 5 most influenatial matrices\n"],"name":"stdout"},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAQwAAAEBCAYAAACNEptNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"QILFfsy24om6","colab_type":"code","colab":{}},"source":["x_train = []\n","for counter, train_idx in enumerate(np.argsort(CNN_predicted_loss_diffs)[-5:]):\n","    x_train.append(X_train[train_idx, :])\n","    \n","    "],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"DegmGNeN6Ul7","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":290},"outputId":"ffd73941-43fa-42ae-a600-093cc111bde1","executionInfo":{"status":"ok","timestamp":1559893864320,"user_tz":-120,"elapsed":1122,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["plt.spy(np.reshape(x_train[0], [128, 128]))"],"execution_count":44,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7fde6c37af10>"]},"metadata":{"tags":[]},"execution_count":44},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAQwAAAEBCAYAAACNEptNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"iqEtZQ85637H","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":290},"outputId":"a4893620-c46d-4d6d-d223-00e601d87afa","executionInfo":{"status":"ok","timestamp":1559893928846,"user_tz":-120,"elapsed":1330,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["plt.spy(np.reshape(x_train[1], [128, 128]))"],"execution_count":49,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7fde6c194d90>"]},"metadata":{"tags":[]},"execution_count":49},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAQwAAAEBCAYAAACNEptNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi40LCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcv7US4rQAAEnVJREFUeJzt3V1ok/f7x/FPktLUCpJGqaatWDxQ\nKoKFBnsoq2KFZVUYrK4oG+o8kPlwoNIJa8W6sVgRx9rNKTIYiJ5s+NCB6YGebGxq1Q6yOpX6VGy0\na9pSH7Bicv8ORvu3/mv31Ty2eb+gB01ovpfk3mfX97rv3LFZlmUJAAzYU10AgImDwABgjMAAYIzA\nAGCMwABgjMAAYCwtA+P27duqrq5WZWWlqqurdefOnYSv2d/fr08++USVlZV677339Omnn6qvr0+S\n1N7erqqqKlVWVmrdunUKh8MJr0eSmpqaNH/+fN24cSMldQwNDam+vl7Lly/Xe++9p88//1xS8t+f\n8+fPa9WqVVq5cqWqqqrU2tqalDr8fr8qKipGvQf/tW4iahqrjvGOVymBx4qVhtauXWudPHnSsizL\nOnnypLV27dqEr9nf32/98ccfI79/9dVX1meffWZFIhFr2bJl1qVLlyzLsqzm5martrY24fUEg0Fr\n/fr11jvvvGNdv349JXU0NDRYX3zxhRWNRi3Lsqx//vnHsqzkvj/RaNTyer3W9evXLcuyrGvXrlml\npaVWJBJJeB2XLl2yuru7R96DYeOtm4iaxqrjdcerZVkJPVbSLjB6e3utsrIy68WLF5ZlWdaLFy+s\nsrIyKxwOJ7WOs2fPWh999JH1559/Wu++++7I4+Fw2CotLU3o2kNDQ9YHH3xgdXV1jRwkya7j8ePH\nVllZmfX48eNRjyf7/YlGo9bixYuttrY2y7Is6+LFi9by5cuTWsfL/6GOt26ia3o1uF42fLxalpXQ\nYyUrPn1K/IRCIc2cOVMOh0OS5HA4lJ+fr1AoJLfbnZQaotGojh8/roqKCoVCIRUUFIw853a7FY1G\nNTAwIJfLlZD1v/76a1VVVamoqGjksWTX0dXVJZfLpaamJl24cEFTp07V1q1blZOTk9T3x2az6eDB\ng9q0aZNyc3P15MkTHT58OGXHyXjrWpaVkppePl6Ha0zUsZKWM4xUa2hoUG5urtasWZP0ta9evapg\nMKiampqkr/2ySCSirq4uLViwQD///LO2b9+uzZs36+nTp0mt48WLF/r+++/17bff6vz58/ruu++0\nbdu2pNeRzpJ5vKZdh+HxePTw4UNFIhE5HA5FIhH19PTI4/EkZX2/36+7d+/q0KFDstvt8ng86u7u\nHnm+r69Pdrs9Yd3FpUuX1NnZqaVLl0qSHjx4oPXr12vt2rVJrcPj8SgrK0s+n0+StGjRIuXl5Skn\nJyep78+1a9fU09OjsrIySVJZWZmmTJkip9OZkuNkvOPTsqyk1/Tq8TpcY6KOlbTrMKZPn66SkhK1\ntLRIklpaWlRSUpKU7ciBAwcUDAbV3Nys7OxsSdLChQv17NkztbW1SZJOnDihFStWJKyGjRs36tdf\nf9W5c+d07tw5zZo1S0ePHtWGDRuSWofb7VZ5ebl+++03Sf9O/8PhsIqLi5P6/syaNUsPHjzQrVu3\nJEmdnZ0Kh8OaM2dOSo6T8Y7PZB+7Yx2vUmKPWZtlpd+nVTs7O1VbW6vBwUFNmzZNfr9fc+fOTeia\nN2/elM/nU3FxsXJyciRJRUVFam5u1pUrV1RfX6+hoSEVFhaqsbFRM2bMSGg9wyoqKnTo0CHNmzcv\n6XV0dXVp165dGhgYUFZWlrZt26YlS5Yk/f05ffq0jhw5IpvNJknasmWLli1blvA69u7dq9bWVvX2\n9iovL08ul0u//PLLuOsmoqax6jh48OBrj1dJCTtW0jIwAKSntNuSAEhfBAYAYwQGAGMEBgBjaRsY\ng4OD+uabbzQ4OJjxdaRDDdRBHVKaB0ZTU1NavBmpriMdaqAO6pDSODAApB8CA4CxhAZGKm6EAyBx\nEhoY9fX1qqmpUSAQUE1Njerq6oz/Njs7W7Nnzx75qHCqOBwOFRYWprSOdKiBOqhDSuCl4eFwWJWV\nlbpw4cLIJ/fKy8vV2tqatPtaAIivhHUY491oBMDElPZDz5fvHAQgtRIWGC/faETSW99MpLu7WwUF\nBQQHkAYSFhipvBEOgMRI6P0w4n0zkeEu4+XbjwFIngl1Ax0CA0itCRUYwwgOIDXS/iwJgPQxITuM\nYQUFBXQZQBJN6MCQ2J4AycSWBICxCd9hDKPTABKPDgOAsUnTYQxjEAokzqQLDIntCZAobEkAGJuU\nHcYwOg0gvugwABib1B3GMAahQHxkRGBIbE+AeGBLAsBYxnQYw+g0gLdHhwHAWMZ1GMPoNIA3R4cB\nwFjGdhjDOOUKmMv4wJDYngCm2JIAMEaH8RI6DWB8dBgAjNFhjIFBKDC2mAKjv79fO3fu1L1795Sd\nna05c+Zoz549crvdam9vV11dnYaGhlRYWKjGxkZNnz49nrUn1Mtf/jz8hdCECDJdTFsSm82mDRs2\nKBAI6MyZM5o9e7b279+vaDSqHTt2qK6uToFAQF6vV/v3749XzQBSJKbAcLlcKi8vH/m9tLRU3d3d\nCgaDcjqd8nq9kqTVq1fr7NmzsVUKIOWy4vVC0WhUx48fV0VFhUKh0KiW3u12KxqNamBgQC6XK15L\nJtSr2w+2I0Acz5I0NDQoNzdXa9asiddLplRBQcHIz/DvQKaLS4fh9/t19+5dHTp0SHa7XR6PZ9T/\nkfv6+mS32ydMdyGN3WG8LjToPpApYu4wDhw4oGAwqObmZmVnZ0uSFi5cqGfPnqmtrU2SdOLECa1Y\nsSLWpQCkWEynVW/evCmfz6fi4mLl5ORIkoqKitTc3KwrV66ovr5+1GnVGTNmxK3wVKDDQKbjwq03\nQGAg03FpOABjBAYAY2xJ4mCsrQrbFExGdBgAjBEYAIwRGACMERgAjDH0TBAGoZiM6DAAGCMwABgj\nMAAYIzAAGGPomUQMQjHR0WEAMEZgADBGYAAwRmAAMMbQM8UYhGIiocMAYIzAAGCMwABgjMAAYIyh\nZxpiEIp0RYcBwFjcAqOpqUnz58/XjRs3JEnt7e2qqqpSZWWl1q1bp3A4HK+lAKRIXALjr7/+Unt7\nuwoLCyVJ0WhUO3bsUF1dnQKBgLxer/bv3x+PpQCkUMyB8fz5c+3Zs0e7d+8eeSwYDMrpdMrr9UqS\nVq9erbNnz8a6FIAUy4r1Bb7++mtVVVWpqKho5LFQKDRqcOd2uxWNRjUwMCCXyxXrkpPe6wacrw5D\nGYQi2WIKjKtXryoYDGr79u3xqgfjICCQajEFxqVLl9TZ2amlS5dKkh48eKD169dr7dq1ow7uvr4+\n2e12uosY0WEg1WKaYWzcuFG//vqrzp07p3PnzmnWrFk6evSoNmzYoGfPnqmtrU2SdOLECa1YsSIu\nBQNInZhnGGOx2+3at2+f6uvrNTQ0pMLCQjU2NiZiKQBJxJWeExhXhCLZuNITgDECA4AxAgOAMQID\ngDGGnpMMg1AkEh0GAGMEBgBjBAYAYwQGAGMMPTPAWINQiWEo3hwdBgBjBAYAYwQGAGMEBgBjBAYA\nY5wlyWBcRo43RYcBwBiBAcAYgQHAGIEBwBhDT4zCIBTjocMAYIzAAGCMwABgjMAAYCzmoefQ0JC+\n/PJL/f7773I6nSotLVVDQ4Nu376t2tpaDQwMyOVyye/3q7i4OE5lI5leHoR2d3eroKCAQWiGivm7\nVRsbG+V0OhUIBGSz2dTb2ytJqq+vV01NjVauXKlTp06prq5OP/74Y8wFA0idmLYkT5480cmTJ7V1\n61bZbDZJ0owZMxQOh9XR0SGfzydJ8vl86ujoUF9fX+wVA0iZmDqMrq4uuVwuNTU16cKFC5o6daq2\nbt2qnJwczZw5Uw6HQ5LkcDiUn5+vUCgkt9sdl8IBJF9MHUYkElFXV5cWLFign3/+Wdu3b9fmzZv1\n9OnTeNWHNNDd3T3yM/w7MlNMHYbH41FWVtbI1mPRokXKy8tTTk6OHj58qEgkIofDoUgkop6eHnk8\nnrgUjdTjitDMFFOH4Xa7VV5ert9++02SdPv2bYXDYRUXF6ukpEQtLS2SpJaWFpWUlLAdASa4mE+r\ndnV1adeuXRoYGFBWVpa2bdumJUuWqLOzU7W1tRocHNS0adPk9/s1d+7ceNWNFKPDyEx8+AxvhcDI\nTFzpCcAYHQbihq5j8qPDAGCMwABgjMAAYIzAAGCMoScSikHo5EKHAcAYgQHAGIEBwBiBAcAYQ08k\nHYPQiYsOA4AxAgOAMQIDgDECA4Axhp5ICwxCJwY6DADGCAwAxggMAMYIDADGGHoibY01CJUYhqYS\nHQYAYwQGAGMEBgBjMQfG+fPntWrVKq1cuVJVVVVqbW2V9O/3rFZXV6uyslLV1dW6c+dOrEsBSLGY\nhp6WZWnx4sU6duyY5s2bp7///lsffvihLl++rI8//ljvv/++Vq5cqVOnTumnn37Sjz/+GM/aMckx\n9Ew/WbG+gN1u16NHjyRJjx49Un5+vvr7+9XR0aEffvhBkuTz+dTQ0KC+vj6+wR3GXhcMBQUF6u7u\nHhUohEhyxBQYNptNBw8e1KZNm5Sbm6snT57o8OHDCoVCmjlzphwOhyTJ4XAoPz9foVCIwAAmsJgC\n48WLF/r+++/17bffqqysTJcvX9a2bdu0b9++eNUH/D/D3QRdRfLFNPS8du2aenp6VFZWJkkqKyvT\nlClT5HQ69fDhQ0UiEUlSJBJRT0+PPB5P7BUj4w1vRQoKCkZ+kBwxBcasWbP04MED3bp1S5LU2dmp\ncDisOXPmqKSkRC0tLZKklpYWlZSUsB0BJriYLw0/ffq0jhw5IpvNJknasmWLli1bps7OTtXW1mpw\ncFDTpk2T3+/X3Llz41I08Crup5EcfJYEkwKBkRxc6QnAGIEBwBiBAcAYMwxMWsw14o8OA4AxAgOA\nMQIDgDECA4Axhp7IKAxCY0OHAcAYgQHAGIEBwBiBAcAYQ09kPAah5ugwABgjMAAYIzAAGCMwABhj\n6AmMgUHo2OgwABgjMAAYIzAAGCMwABhj6AkYYhBKhwHgDfxnYPj9flVUVGj+/Pm6cePGyOO3b99W\ndXW1KisrVV1drTt37hg9B2Di+s/AWLp0qY4dO6bCwsJRj9fX16umpkaBQEA1NTWqq6szeg7AxPWf\ngeH1euXxeEY9Fg6H1dHRIZ/PJ0ny+Xzq6OhQX1/fuM8BmNiy3uaPQqGQZs6cKYfDIUlyOBzKz89X\nKBSSZVmvfc7tdsevciDJXjfgfHUYOpkHoW8VGAD+z2QOiFe9VWB4PB49fPhQkUhEDodDkUhEPT09\n8ng8sizrtc8Bk1EmdRhvdVp1+vTpKikpUUtLiySppaVFJSUlcrvd4z4HYGL7zwu39u7dq9bWVvX2\n9iovL08ul0u//PKLOjs7VVtbq8HBQU2bNk1+v19z586VpHGfAyabTOowuNITiLOxrgiVJkeQcKUn\nAGMEBgBjBAYAYwQGAGMEBgBjnCUBkmQy3E+DDgOAMQIDgDECA4AxAgOAMYaeQApNtEEoHQYAYwQG\nAGMEBgBjBAYAYww9gTSTzoNQOgwAxggMAMYIDADGCAwAxhh6AhNAugxC6TAAGCMwABgjMAAYIzAA\nGDMaevr9fgUCAd2/f19nzpzRvHnz1N/fr507d+revXvKzs7WnDlztGfPnpHvUG1vb1ddXZ2GhoZU\nWFioxsZGTZ8+PeH/ICBTvDwI7e7uVkFBQcIHoUYdxtKlS3Xs2DEVFhaOPGaz2bRhwwYFAgGdOXNG\ns2fP1v79+yVJ0WhUO3bsUF1dnQKBgLxe78hzACYuo8Dwer3yeDyjHnO5XCovLx/5vbS0dCTdgsGg\nnE6nvF6vJGn16tU6e/ZsvGoGkCJxmWFEo1EdP35cFRUVkqRQKDSqXXK73YpGoxoYGIjHcgBSJCse\nL9LQ0KDc3FytWbMmHi8HwMCr84pkXMgVc2D4/X7dvXtXhw4dkt3+b8Pi8XhGFd/X1ye73S6XyxXr\ncgDGkegrQmPakhw4cEDBYFDNzc3Kzs4eeXzhwoV69uyZ2traJEknTpzQihUrYqsUQMoZnVbdu3ev\nWltb1dvbq7y8PLlcLh08eFA+n0/FxcXKycmRJBUVFam5uVmSdOXKFdXX1486rTpjxozE/muADJfo\nDoMPnwGTSFpvSQBkFjoMYJKLZ9dBhwHAGIEBwBiBAcAYgQHAGENPIAMtXrxYFy9efOO/o8MAMtCr\nnz43RYcBwFjadhg9PT1atmyZQqFQSusIhUKqqKhIaR3pUAN1UIeUxoHx/PlzdXV1KRKJpLSOSCSi\n+/fvp7SOdKiBOqhDSuPAAJB+CAwAxggMAMYcu3fv3p3qIl7H6XSqvLxcTqcz4+tIhxqogzo4rQrA\nGFsSAMYIDADGCAwAxggMAMYIDADG/gd/ZPIEvrSbmQAAAABJRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"o8Qb-6fN64kd","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":290},"outputId":"e42aa556-130f-4535-e694-c35b4a95f819","executionInfo":{"status":"ok","timestamp":1559893933002,"user_tz":-120,"elapsed":1045,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["plt.spy(np.reshape(x_train[2], [128, 128]))"],"execution_count":50,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7fde6c1123d0>"]},"metadata":{"tags":[]},"execution_count":50},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAQwAAAEBCAYAAACNEptNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAMtoGAAso2EAsIyGAcCy/wPavWfGErvZGQAAAABJRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"ToCk8IOS64y2","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":290},"outputId":"a609d24a-49a3-4380-ea00-e3698c0452f7","executionInfo":{"status":"ok","timestamp":1559893937007,"user_tz":-120,"elapsed":1014,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["plt.spy(np.reshape(x_train[3], [128, 128]))"],"execution_count":51,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7fde6c1079d0>"]},"metadata":{"tags":[]},"execution_count":51},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAQwAAAEBCAYAAACNEptNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ng0FEwthgEJEwNhhEJOz/AJKZmY4ieODfAAAAAElFTkSuQmCC\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"L0RJCjqx64-L","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":290},"outputId":"f29aaf89-0200-4b13-a745-cd33758436e4","executionInfo":{"status":"ok","timestamp":1559893939808,"user_tz":-120,"elapsed":1038,"user":{"displayName":"Holdthedoor","photoUrl":"","userId":"05144608278972732019"}}},"source":["plt.spy(np.reshape(x_train[4], [128, 128]))"],"execution_count":52,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.image.AxesImage at 0x7fde6c079fd0>"]},"metadata":{"tags":[]},"execution_count":52},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAQwAAAEBCAYAAACNEptNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ng0FEwthgEJEwNhhEJOz/AJKZmY4ieODfAAAAAElFTkSuQmCC\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]}]}