keras_center_loss_with_self_defined_layer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import mnist\n",
    "from keras.layers import Input,Conv2D,Activation,Dense,Lambda,Flatten,Embedding,PReLU,BatchNormalization\n",
    "from keras.models import Model\n",
    "import keras.backend as K\n",
    "from keras.utils import to_categorical\n",
    "import numpy as np\n",
    "from keras.callbacks import EarlyStopping\n",
    "from keras.engine.topology import Layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 28, 28, 1)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 26, 26, 32)        320       \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 26, 26, 32)        128       \n",
      "_________________________________________________________________\n",
      "p_re_lu_1 (PReLU)            (None, 26, 26, 32)        21632     \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 24, 24, 32)        9248      \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 24, 24, 32)        128       \n",
      "_________________________________________________________________\n",
      "p_re_lu_2 (PReLU)            (None, 24, 24, 32)        18432     \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 20, 20, 64)        51264     \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 20, 20, 64)        256       \n",
      "_________________________________________________________________\n",
      "p_re_lu_3 (PReLU)            (None, 20, 20, 64)        25600     \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 16, 16, 64)        102464    \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 16, 16, 64)        256       \n",
      "_________________________________________________________________\n",
      "p_re_lu_4 (PReLU)            (None, 16, 16, 64)        16384     \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 10, 10, 128)       401536    \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 10, 10, 128)       512       \n",
      "_________________________________________________________________\n",
      "p_re_lu_5 (PReLU)            (None, 10, 10, 128)       12800     \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 4, 4, 128)         802944    \n",
      "_________________________________________________________________\n",
      "batch_normalization_6 (Batch (None, 4, 4, 128)         512       \n",
      "_________________________________________________________________\n",
      "p_re_lu_6 (PReLU)            (None, 4, 4, 128)         2048      \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 2048)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 2)                 4098      \n",
      "_________________________________________________________________\n",
      "out1 (PReLU)                 (None, 2)                 2         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                30        \n",
      "=================================================================\n",
      "Total params: 1,470,594\n",
      "Trainable params: 1,469,698\n",
      "Non-trainable params: 896\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "inputs=Input((28,28,1))\n",
    "x=Conv2D(32,(3,3))(inputs)\n",
    "x=BatchNormalization()(x)\n",
    "x=PReLU()(x)\n",
    "x=Conv2D(32,(3,3))(x)\n",
    "x=BatchNormalization()(x)\n",
    "x=PReLU()(x)\n",
    "x=Conv2D(64,(5,5))(x)\n",
    "x=BatchNormalization()(x)\n",
    "x=PReLU()(x)\n",
    "x=Conv2D(64,(5,5))(x)\n",
    "x=BatchNormalization()(x)\n",
    "x=PReLU()(x)\n",
    "x=Conv2D(128,(7,7))(x)\n",
    "x=BatchNormalization()(x)\n",
    "x=PReLU()(x)\n",
    "x=Conv2D(128,(7,7))(x)\n",
    "x=BatchNormalization()(x)\n",
    "x=PReLU()(x)\n",
    "x=Flatten()(x)\n",
    "x=Dense(2)(x)\n",
    "out1=PReLU(name=\"out1\")(x)\n",
    "out2=Dense(10,activation=\"softmax\")(out1)\n",
    "\n",
    "model=Model(inputs,out2)\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 28, 28, 1)    0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 26, 26, 32)   320         input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 26, 26, 32)   128         conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "p_re_lu_1 (PReLU)               (None, 26, 26, 32)   21632       batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 24, 24, 32)   9248        p_re_lu_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 24, 24, 32)   128         conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "p_re_lu_2 (PReLU)               (None, 24, 24, 32)   18432       batch_normalization_2[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_3 (Conv2D)               (None, 20, 20, 64)   51264       p_re_lu_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 20, 20, 64)   256         conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "p_re_lu_3 (PReLU)               (None, 20, 20, 64)   25600       batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_4 (Conv2D)               (None, 16, 16, 64)   102464      p_re_lu_3[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_4 (BatchNor (None, 16, 16, 64)   256         conv2d_4[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "p_re_lu_4 (PReLU)               (None, 16, 16, 64)   16384       batch_normalization_4[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_5 (Conv2D)               (None, 10, 10, 128)  401536      p_re_lu_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_5 (BatchNor (None, 10, 10, 128)  512         conv2d_5[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "p_re_lu_5 (PReLU)               (None, 10, 10, 128)  12800       batch_normalization_5[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_6 (Conv2D)               (None, 4, 4, 128)    802944      p_re_lu_5[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_6 (BatchNor (None, 4, 4, 128)    512         conv2d_6[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "p_re_lu_6 (PReLU)               (None, 4, 4, 128)    2048        batch_normalization_6[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "flatten_1 (Flatten)             (None, 2048)         0           p_re_lu_6[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 2)            4098        flatten_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "out1 (PReLU)                    (None, 2)            2           dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "input_2 (InputLayer)            (None, 10)           0                                            \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 10)           30          out1[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "center_loss_layer_1 (CenterLoss (None, 1)            20          out1[0][0]                       \n",
      "                                                                 input_2[0][0]                    \n",
      "==================================================================================================\n",
      "Total params: 1,470,614\n",
      "Trainable params: 1,469,698\n",
      "Non-trainable params: 916\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "-\n",
    "model_center_loss.summary()        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(30000, 28, 28, 1)\n",
      "(30000,)\n",
      "(30000, 10)\n"
     ]
    }
   ],
   "source": [
    "path=\"mnist.npz\"\n",
    "f = np.load(path)\n",
    "x_train, y_train = f['x_train'][:30000], f['y_train'][:30000]\n",
    "x_test, y_test = f['x_test'], f['y_test']\n",
    "f.close()\n",
    "\n",
    "x_train=x_train.astype(\"float32\")\n",
    "x_test=x_test.astype(\"float32\")\n",
    "x_train/=255.\n",
    "x_test/=255.\n",
    "x_train=np.expand_dims(x_train,axis=-1)\n",
    "x_test=np.expand_dims(x_test,axis=-1)\n",
    "\n",
    "y_train_one_hot=to_categorical(y_train)\n",
    "y_test_one_hot=to_categorical(y_test)\n",
    "\n",
    "print(x_train.shape)\n",
    "print(y_train.shape)\n",
    "print(y_train_one_hot.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 30000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "30000/30000 [==============================] - 7s 235us/step - loss: 2.3103 - dense_2_loss: 2.2998 - center_loss_layer_1_loss: 0.0210 - dense_2_acc: 0.1297 - center_loss_layer_1_acc: 0.9905 - val_loss: 2.2985 - val_dense_2_loss: 2.2945 - val_center_loss_layer_1_loss: 0.0079 - val_dense_2_acc: 0.1481 - val_center_loss_layer_1_acc: 0.9973\n",
      "Epoch 2/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 2.2850 - dense_2_loss: 2.2643 - center_loss_layer_1_loss: 0.0413 - dense_2_acc: 0.1505 - center_loss_layer_1_acc: 0.9771 - val_loss: 2.2612 - val_dense_2_loss: 2.2033 - val_center_loss_layer_1_loss: 0.1158 - val_dense_2_acc: 0.1647 - val_center_loss_layer_1_acc: 0.9331\n",
      "Epoch 3/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 2.1977 - dense_2_loss: 2.1397 - center_loss_layer_1_loss: 0.1160 - dense_2_acc: 0.1865 - center_loss_layer_1_acc: 0.9464 - val_loss: 2.1450 - val_dense_2_loss: 2.0420 - val_center_loss_layer_1_loss: 0.2059 - val_dense_2_acc: 0.2420 - val_center_loss_layer_1_acc: 0.8951\n",
      "Epoch 4/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 2.0788 - dense_2_loss: 1.9799 - center_loss_layer_1_loss: 0.1977 - dense_2_acc: 0.3046 - center_loss_layer_1_acc: 0.9093 - val_loss: 2.0246 - val_dense_2_loss: 1.8946 - val_center_loss_layer_1_loss: 0.2601 - val_dense_2_acc: 0.3069 - val_center_loss_layer_1_acc: 0.8686\n",
      "Epoch 5/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 1.9721 - dense_2_loss: 1.8305 - center_loss_layer_1_loss: 0.2832 - dense_2_acc: 0.3081 - center_loss_layer_1_acc: 0.8632 - val_loss: 1.9438 - val_dense_2_loss: 1.7449 - val_center_loss_layer_1_loss: 0.3977 - val_dense_2_acc: 0.3089 - val_center_loss_layer_1_acc: 0.7672\n",
      "Epoch 6/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 1.8622 - dense_2_loss: 1.7120 - center_loss_layer_1_loss: 0.3004 - dense_2_acc: 0.3091 - center_loss_layer_1_acc: 0.8502 - val_loss: 1.8491 - val_dense_2_loss: 1.6747 - val_center_loss_layer_1_loss: 0.3487 - val_dense_2_acc: 0.3107 - val_center_loss_layer_1_acc: 0.8204\n",
      "Epoch 7/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 1.7981 - dense_2_loss: 1.6373 - center_loss_layer_1_loss: 0.3215 - dense_2_acc: 0.3110 - center_loss_layer_1_acc: 0.8342 - val_loss: 1.7953 - val_dense_2_loss: 1.6185 - val_center_loss_layer_1_loss: 0.3536 - val_dense_2_acc: 0.3147 - val_center_loss_layer_1_acc: 0.8262\n",
      "Epoch 8/100\n",
      "30000/30000 [==============================] - 5s 163us/step - loss: 1.7498 - dense_2_loss: 1.5824 - center_loss_layer_1_loss: 0.3348 - dense_2_acc: 0.3140 - center_loss_layer_1_acc: 0.8305 - val_loss: 1.8063 - val_dense_2_loss: 1.5865 - val_center_loss_layer_1_loss: 0.4396 - val_dense_2_acc: 0.3151 - val_center_loss_layer_1_acc: 0.7668\n",
      "Epoch 9/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 1.6866 - dense_2_loss: 1.5380 - center_loss_layer_1_loss: 0.2971 - dense_2_acc: 0.3179 - center_loss_layer_1_acc: 0.8556 - val_loss: 1.7723 - val_dense_2_loss: 1.5495 - val_center_loss_layer_1_loss: 0.4456 - val_dense_2_acc: 0.3148 - val_center_loss_layer_1_acc: 0.7692\n",
      "Epoch 10/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 1.6760 - dense_2_loss: 1.5011 - center_loss_layer_1_loss: 0.3498 - dense_2_acc: 0.3228 - center_loss_layer_1_acc: 0.8206 - val_loss: 1.7106 - val_dense_2_loss: 1.5195 - val_center_loss_layer_1_loss: 0.3822 - val_dense_2_acc: 0.3214 - val_center_loss_layer_1_acc: 0.8152\n",
      "Epoch 11/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 1.6292 - dense_2_loss: 1.4631 - center_loss_layer_1_loss: 0.3322 - dense_2_acc: 0.3275 - center_loss_layer_1_acc: 0.8302 - val_loss: 1.6595 - val_dense_2_loss: 1.4701 - val_center_loss_layer_1_loss: 0.3788 - val_dense_2_acc: 0.3404 - val_center_loss_layer_1_acc: 0.8318\n",
      "Epoch 12/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 1.5840 - dense_2_loss: 1.4259 - center_loss_layer_1_loss: 0.3161 - dense_2_acc: 0.3320 - center_loss_layer_1_acc: 0.8392 - val_loss: 1.7075 - val_dense_2_loss: 1.4461 - val_center_loss_layer_1_loss: 0.5228 - val_dense_2_acc: 0.3462 - val_center_loss_layer_1_acc: 0.7184\n",
      "Epoch 13/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 1.5797 - dense_2_loss: 1.3920 - center_loss_layer_1_loss: 0.3755 - dense_2_acc: 0.3414 - center_loss_layer_1_acc: 0.7911 - val_loss: 1.5880 - val_dense_2_loss: 1.3982 - val_center_loss_layer_1_loss: 0.3796 - val_dense_2_acc: 0.3508 - val_center_loss_layer_1_acc: 0.8247\n",
      "Epoch 14/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 1.5300 - dense_2_loss: 1.3535 - center_loss_layer_1_loss: 0.3529 - dense_2_acc: 0.3539 - center_loss_layer_1_acc: 0.8108 - val_loss: 1.5897 - val_dense_2_loss: 1.3371 - val_center_loss_layer_1_loss: 0.5051 - val_dense_2_acc: 0.3870 - val_center_loss_layer_1_acc: 0.7439\n",
      "Epoch 15/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 1.5264 - dense_2_loss: 1.3054 - center_loss_layer_1_loss: 0.4420 - dense_2_acc: 0.3923 - center_loss_layer_1_acc: 0.7512 - val_loss: 1.5777 - val_dense_2_loss: 1.2867 - val_center_loss_layer_1_loss: 0.5821 - val_dense_2_acc: 0.4727 - val_center_loss_layer_1_acc: 0.7101\n",
      "Epoch 16/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 1.4582 - dense_2_loss: 1.2430 - center_loss_layer_1_loss: 0.4305 - dense_2_acc: 0.4737 - center_loss_layer_1_acc: 0.7792 - val_loss: 1.7742 - val_dense_2_loss: 1.2809 - val_center_loss_layer_1_loss: 0.9865 - val_dense_2_acc: 0.4258 - val_center_loss_layer_1_acc: 0.4355\n",
      "Epoch 17/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 1.4149 - dense_2_loss: 1.1867 - center_loss_layer_1_loss: 0.4563 - dense_2_acc: 0.5242 - center_loss_layer_1_acc: 0.7560 - val_loss: 1.4867 - val_dense_2_loss: 1.1847 - val_center_loss_layer_1_loss: 0.6040 - val_dense_2_acc: 0.5388 - val_center_loss_layer_1_acc: 0.7400\n",
      "Epoch 18/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 1.3585 - dense_2_loss: 1.1256 - center_loss_layer_1_loss: 0.4657 - dense_2_acc: 0.5607 - center_loss_layer_1_acc: 0.7531 - val_loss: 1.5809 - val_dense_2_loss: 1.1606 - val_center_loss_layer_1_loss: 0.8405 - val_dense_2_acc: 0.5362 - val_center_loss_layer_1_acc: 0.5219\n",
      "Epoch 19/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 1.2931 - dense_2_loss: 1.0709 - center_loss_layer_1_loss: 0.4442 - dense_2_acc: 0.5868 - center_loss_layer_1_acc: 0.7744 - val_loss: 1.4372 - val_dense_2_loss: 1.0947 - val_center_loss_layer_1_loss: 0.6849 - val_dense_2_acc: 0.6035 - val_center_loss_layer_1_acc: 0.6821\n",
      "Epoch 20/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 1.2550 - dense_2_loss: 1.0179 - center_loss_layer_1_loss: 0.4742 - dense_2_acc: 0.6009 - center_loss_layer_1_acc: 0.7508 - val_loss: 1.3797 - val_dense_2_loss: 1.0619 - val_center_loss_layer_1_loss: 0.6357 - val_dense_2_acc: 0.6143 - val_center_loss_layer_1_acc: 0.7502\n",
      "Epoch 21/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 1.2165 - dense_2_loss: 0.9749 - center_loss_layer_1_loss: 0.4830 - dense_2_acc: 0.6144 - center_loss_layer_1_acc: 0.7392 - val_loss: 1.4626 - val_dense_2_loss: 1.0503 - val_center_loss_layer_1_loss: 0.8247 - val_dense_2_acc: 0.6373 - val_center_loss_layer_1_acc: 0.6366\n",
      "Epoch 22/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 1.1692 - dense_2_loss: 0.9370 - center_loss_layer_1_loss: 0.4644 - dense_2_acc: 0.6222 - center_loss_layer_1_acc: 0.7550 - val_loss: 1.3056 - val_dense_2_loss: 0.9789 - val_center_loss_layer_1_loss: 0.6534 - val_dense_2_acc: 0.6233 - val_center_loss_layer_1_acc: 0.7214\n",
      "Epoch 23/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 1.1429 - dense_2_loss: 0.9042 - center_loss_layer_1_loss: 0.4776 - dense_2_acc: 0.6283 - center_loss_layer_1_acc: 0.7474 - val_loss: 1.2400 - val_dense_2_loss: 0.9329 - val_center_loss_layer_1_loss: 0.6142 - val_dense_2_acc: 0.6362 - val_center_loss_layer_1_acc: 0.7648\n",
      "Epoch 24/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 1.1267 - dense_2_loss: 0.8780 - center_loss_layer_1_loss: 0.4973 - dense_2_acc: 0.6381 - center_loss_layer_1_acc: 0.7256 - val_loss: 1.2182 - val_dense_2_loss: 0.9206 - val_center_loss_layer_1_loss: 0.5951 - val_dense_2_acc: 0.6343 - val_center_loss_layer_1_acc: 0.7966\n",
      "Epoch 25/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 1.0530 - dense_2_loss: 0.8449 - center_loss_layer_1_loss: 0.4162 - dense_2_acc: 0.6445 - center_loss_layer_1_acc: 0.7759 - val_loss: 1.4412 - val_dense_2_loss: 0.9341 - val_center_loss_layer_1_loss: 1.0142 - val_dense_2_acc: 0.6464 - val_center_loss_layer_1_acc: 0.5690\n",
      "Epoch 26/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 1.0386 - dense_2_loss: 0.8221 - center_loss_layer_1_loss: 0.4329 - dense_2_acc: 0.6503 - center_loss_layer_1_acc: 0.7683 - val_loss: 1.2627 - val_dense_2_loss: 0.8772 - val_center_loss_layer_1_loss: 0.7710 - val_dense_2_acc: 0.6540 - val_center_loss_layer_1_acc: 0.6899\n",
      "Epoch 27/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 1.0453 - dense_2_loss: 0.8094 - center_loss_layer_1_loss: 0.4718 - dense_2_acc: 0.6589 - center_loss_layer_1_acc: 0.7410 - val_loss: 1.1631 - val_dense_2_loss: 0.8486 - val_center_loss_layer_1_loss: 0.6290 - val_dense_2_acc: 0.6580 - val_center_loss_layer_1_acc: 0.7829\n",
      "Epoch 28/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 1.0374 - dense_2_loss: 0.7899 - center_loss_layer_1_loss: 0.4950 - dense_2_acc: 0.6648 - center_loss_layer_1_acc: 0.7436 - val_loss: 1.2297 - val_dense_2_loss: 0.8576 - val_center_loss_layer_1_loss: 0.7442 - val_dense_2_acc: 0.7013 - val_center_loss_layer_1_acc: 0.7328\n",
      "Epoch 29/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 1.0012 - dense_2_loss: 0.7680 - center_loss_layer_1_loss: 0.4665 - dense_2_acc: 0.6750 - center_loss_layer_1_acc: 0.7411 - val_loss: 1.1933 - val_dense_2_loss: 0.8237 - val_center_loss_layer_1_loss: 0.7392 - val_dense_2_acc: 0.6731 - val_center_loss_layer_1_acc: 0.7208\n",
      "Epoch 30/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.9678 - dense_2_loss: 0.7496 - center_loss_layer_1_loss: 0.4364 - dense_2_acc: 0.6819 - center_loss_layer_1_acc: 0.7539 - val_loss: 1.1954 - val_dense_2_loss: 0.7929 - val_center_loss_layer_1_loss: 0.8049 - val_dense_2_acc: 0.6944 - val_center_loss_layer_1_acc: 0.6639\n",
      "Epoch 31/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 0.9468 - dense_2_loss: 0.7273 - center_loss_layer_1_loss: 0.4391 - dense_2_acc: 0.6904 - center_loss_layer_1_acc: 0.7568 - val_loss: 1.2283 - val_dense_2_loss: 0.7963 - val_center_loss_layer_1_loss: 0.8640 - val_dense_2_acc: 0.7652 - val_center_loss_layer_1_acc: 0.6428\n",
      "Epoch 32/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.9622 - dense_2_loss: 0.7168 - center_loss_layer_1_loss: 0.4907 - dense_2_acc: 0.7042 - center_loss_layer_1_acc: 0.7215 - val_loss: 1.3220 - val_dense_2_loss: 0.7986 - val_center_loss_layer_1_loss: 1.0468 - val_dense_2_acc: 0.7870 - val_center_loss_layer_1_acc: 0.5676\n",
      "Epoch 33/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 0.9274 - dense_2_loss: 0.7026 - center_loss_layer_1_loss: 0.4494 - dense_2_acc: 0.7222 - center_loss_layer_1_acc: 0.7420 - val_loss: 1.3969 - val_dense_2_loss: 0.7867 - val_center_loss_layer_1_loss: 1.2203 - val_dense_2_acc: 0.6909 - val_center_loss_layer_1_acc: 0.5306\n",
      "Epoch 34/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.9185 - dense_2_loss: 0.6850 - center_loss_layer_1_loss: 0.4669 - dense_2_acc: 0.7319 - center_loss_layer_1_acc: 0.7321 - val_loss: 1.4856 - val_dense_2_loss: 0.7735 - val_center_loss_layer_1_loss: 1.4243 - val_dense_2_acc: 0.7274 - val_center_loss_layer_1_acc: 0.3508\n",
      "Epoch 35/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.9198 - dense_2_loss: 0.6745 - center_loss_layer_1_loss: 0.4906 - dense_2_acc: 0.7412 - center_loss_layer_1_acc: 0.7296 - val_loss: 1.1101 - val_dense_2_loss: 0.7238 - val_center_loss_layer_1_loss: 0.7728 - val_dense_2_acc: 0.7804 - val_center_loss_layer_1_acc: 0.7392\n",
      "Epoch 36/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.8475 - dense_2_loss: 0.6497 - center_loss_layer_1_loss: 0.3955 - dense_2_acc: 0.7646 - center_loss_layer_1_acc: 0.7881 - val_loss: 1.3998 - val_dense_2_loss: 0.7306 - val_center_loss_layer_1_loss: 1.3384 - val_dense_2_acc: 0.7320 - val_center_loss_layer_1_acc: 0.4898\n",
      "Epoch 37/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.8417 - dense_2_loss: 0.6396 - center_loss_layer_1_loss: 0.4043 - dense_2_acc: 0.7853 - center_loss_layer_1_acc: 0.7746 - val_loss: 1.4176 - val_dense_2_loss: 0.7365 - val_center_loss_layer_1_loss: 1.3623 - val_dense_2_acc: 0.8242 - val_center_loss_layer_1_acc: 0.4754\n",
      "Epoch 38/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.8967 - dense_2_loss: 0.6391 - center_loss_layer_1_loss: 0.5153 - dense_2_acc: 0.8038 - center_loss_layer_1_acc: 0.7255 - val_loss: 1.2299 - val_dense_2_loss: 0.6734 - val_center_loss_layer_1_loss: 1.1130 - val_dense_2_acc: 0.7439 - val_center_loss_layer_1_acc: 0.5778\n",
      "Epoch 39/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 0.8189 - dense_2_loss: 0.6136 - center_loss_layer_1_loss: 0.4105 - dense_2_acc: 0.8266 - center_loss_layer_1_acc: 0.7742 - val_loss: 1.1201 - val_dense_2_loss: 0.6816 - val_center_loss_layer_1_loss: 0.8770 - val_dense_2_acc: 0.7826 - val_center_loss_layer_1_acc: 0.7057\n",
      "Epoch 40/100\n",
      "30000/30000 [==============================] - 5s 164us/step - loss: 0.8302 - dense_2_loss: 0.6032 - center_loss_layer_1_loss: 0.4539 - dense_2_acc: 0.8450 - center_loss_layer_1_acc: 0.7389 - val_loss: 1.0335 - val_dense_2_loss: 0.6494 - val_center_loss_layer_1_loss: 0.7682 - val_dense_2_acc: 0.8312 - val_center_loss_layer_1_acc: 0.7545\n",
      "Epoch 41/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.8651 - dense_2_loss: 0.5934 - center_loss_layer_1_loss: 0.5434 - dense_2_acc: 0.8626 - center_loss_layer_1_acc: 0.7094 - val_loss: 1.2311 - val_dense_2_loss: 0.6618 - val_center_loss_layer_1_loss: 1.1386 - val_dense_2_acc: 0.8394 - val_center_loss_layer_1_acc: 0.5891\n",
      "Epoch 42/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.8122 - dense_2_loss: 0.5816 - center_loss_layer_1_loss: 0.4611 - dense_2_acc: 0.8870 - center_loss_layer_1_acc: 0.7314 - val_loss: 1.0690 - val_dense_2_loss: 0.6409 - val_center_loss_layer_1_loss: 0.8562 - val_dense_2_acc: 0.8846 - val_center_loss_layer_1_acc: 0.7601\n",
      "Epoch 43/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7685 - dense_2_loss: 0.5671 - center_loss_layer_1_loss: 0.4028 - dense_2_acc: 0.9081 - center_loss_layer_1_acc: 0.7683 - val_loss: 1.1751 - val_dense_2_loss: 0.6203 - val_center_loss_layer_1_loss: 1.1095 - val_dense_2_acc: 0.8482 - val_center_loss_layer_1_acc: 0.6200\n",
      "Epoch 44/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7852 - dense_2_loss: 0.5550 - center_loss_layer_1_loss: 0.4604 - dense_2_acc: 0.9217 - center_loss_layer_1_acc: 0.7351 - val_loss: 1.2272 - val_dense_2_loss: 0.6289 - val_center_loss_layer_1_loss: 1.1966 - val_dense_2_acc: 0.8678 - val_center_loss_layer_1_acc: 0.5240\n",
      "Epoch 45/100\n",
      "30000/30000 [==============================] - 5s 158us/step - loss: 0.8092 - dense_2_loss: 0.5539 - center_loss_layer_1_loss: 0.5106 - dense_2_acc: 0.9317 - center_loss_layer_1_acc: 0.7157 - val_loss: 1.6211 - val_dense_2_loss: 0.6774 - val_center_loss_layer_1_loss: 1.8873 - val_dense_2_acc: 0.8473 - val_center_loss_layer_1_acc: 0.2601\n",
      "Epoch 46/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.7499 - dense_2_loss: 0.5356 - center_loss_layer_1_loss: 0.4287 - dense_2_acc: 0.9531 - center_loss_layer_1_acc: 0.7638 - val_loss: 1.5690 - val_dense_2_loss: 0.6489 - val_center_loss_layer_1_loss: 1.8402 - val_dense_2_acc: 0.9148 - val_center_loss_layer_1_acc: 0.4289\n",
      "Epoch 47/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.7637 - dense_2_loss: 0.5287 - center_loss_layer_1_loss: 0.4701 - dense_2_acc: 0.9595 - center_loss_layer_1_acc: 0.7405 - val_loss: 1.1300 - val_dense_2_loss: 0.6014 - val_center_loss_layer_1_loss: 1.0573 - val_dense_2_acc: 0.9269 - val_center_loss_layer_1_acc: 0.6006\n",
      "Epoch 48/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7209 - dense_2_loss: 0.5122 - center_loss_layer_1_loss: 0.4173 - dense_2_acc: 0.9657 - center_loss_layer_1_acc: 0.7610 - val_loss: 1.4633 - val_dense_2_loss: 0.6415 - val_center_loss_layer_1_loss: 1.6436 - val_dense_2_acc: 0.8677 - val_center_loss_layer_1_acc: 0.4059\n",
      "Epoch 49/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.7638 - dense_2_loss: 0.5059 - center_loss_layer_1_loss: 0.5157 - dense_2_acc: 0.9715 - center_loss_layer_1_acc: 0.7455 - val_loss: 1.8432 - val_dense_2_loss: 0.6463 - val_center_loss_layer_1_loss: 2.3939 - val_dense_2_acc: 0.8255 - val_center_loss_layer_1_acc: 0.2837\n",
      "Epoch 50/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7895 - dense_2_loss: 0.5105 - center_loss_layer_1_loss: 0.5579 - dense_2_acc: 0.9682 - center_loss_layer_1_acc: 0.6842 - val_loss: 1.3800 - val_dense_2_loss: 0.6018 - val_center_loss_layer_1_loss: 1.5563 - val_dense_2_acc: 0.9266 - val_center_loss_layer_1_acc: 0.4375\n",
      "Epoch 51/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.6614 - dense_2_loss: 0.4825 - center_loss_layer_1_loss: 0.3578 - dense_2_acc: 0.9825 - center_loss_layer_1_acc: 0.8080 - val_loss: 1.1322 - val_dense_2_loss: 0.5576 - val_center_loss_layer_1_loss: 1.1493 - val_dense_2_acc: 0.9116 - val_center_loss_layer_1_acc: 0.6030\n",
      "Epoch 52/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7342 - dense_2_loss: 0.4833 - center_loss_layer_1_loss: 0.5018 - dense_2_acc: 0.9775 - center_loss_layer_1_acc: 0.7077 - val_loss: 1.1553 - val_dense_2_loss: 0.5428 - val_center_loss_layer_1_loss: 1.2250 - val_dense_2_acc: 0.9441 - val_center_loss_layer_1_acc: 0.5350\n",
      "Epoch 53/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.7179 - dense_2_loss: 0.4727 - center_loss_layer_1_loss: 0.4904 - dense_2_acc: 0.9835 - center_loss_layer_1_acc: 0.7427 - val_loss: 1.0054 - val_dense_2_loss: 0.5431 - val_center_loss_layer_1_loss: 0.9244 - val_dense_2_acc: 0.9256 - val_center_loss_layer_1_acc: 0.7298\n",
      "Epoch 54/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.6781 - dense_2_loss: 0.4644 - center_loss_layer_1_loss: 0.4273 - dense_2_acc: 0.9857 - center_loss_layer_1_acc: 0.7884 - val_loss: 1.6337 - val_dense_2_loss: 0.6131 - val_center_loss_layer_1_loss: 2.0412 - val_dense_2_acc: 0.9051 - val_center_loss_layer_1_acc: 0.3442\n",
      "Epoch 55/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7045 - dense_2_loss: 0.4585 - center_loss_layer_1_loss: 0.4920 - dense_2_acc: 0.9864 - center_loss_layer_1_acc: 0.7197 - val_loss: 2.7111 - val_dense_2_loss: 0.6595 - val_center_loss_layer_1_loss: 4.1033 - val_dense_2_acc: 0.8913 - val_center_loss_layer_1_acc: 0.2509\n",
      "Epoch 56/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.7240 - dense_2_loss: 0.4558 - center_loss_layer_1_loss: 0.5362 - dense_2_acc: 0.9863 - center_loss_layer_1_acc: 0.7099 - val_loss: 1.3628 - val_dense_2_loss: 0.5613 - val_center_loss_layer_1_loss: 1.6030 - val_dense_2_acc: 0.9389 - val_center_loss_layer_1_acc: 0.4069\n",
      "Epoch 57/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.6695 - dense_2_loss: 0.4419 - center_loss_layer_1_loss: 0.4553 - dense_2_acc: 0.9882 - center_loss_layer_1_acc: 0.7366 - val_loss: 1.0675 - val_dense_2_loss: 0.4916 - val_center_loss_layer_1_loss: 1.1517 - val_dense_2_acc: 0.9486 - val_center_loss_layer_1_acc: 0.5935\n",
      "Epoch 58/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.6456 - dense_2_loss: 0.4312 - center_loss_layer_1_loss: 0.4288 - dense_2_acc: 0.9918 - center_loss_layer_1_acc: 0.7410 - val_loss: 1.3444 - val_dense_2_loss: 0.5172 - val_center_loss_layer_1_loss: 1.6544 - val_dense_2_acc: 0.9338 - val_center_loss_layer_1_acc: 0.4532\n",
      "Epoch 59/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.6571 - dense_2_loss: 0.4227 - center_loss_layer_1_loss: 0.4688 - dense_2_acc: 0.9917 - center_loss_layer_1_acc: 0.7353 - val_loss: 1.1803 - val_dense_2_loss: 0.5187 - val_center_loss_layer_1_loss: 1.3232 - val_dense_2_acc: 0.9499 - val_center_loss_layer_1_acc: 0.5300\n",
      "Epoch 60/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.6437 - dense_2_loss: 0.4195 - center_loss_layer_1_loss: 0.4483 - dense_2_acc: 0.9918 - center_loss_layer_1_acc: 0.7408 - val_loss: 1.0701 - val_dense_2_loss: 0.4961 - val_center_loss_layer_1_loss: 1.1480 - val_dense_2_acc: 0.9557 - val_center_loss_layer_1_acc: 0.6307\n",
      "Epoch 61/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.8432 - dense_2_loss: 0.4510 - center_loss_layer_1_loss: 0.7843 - dense_2_acc: 0.9740 - center_loss_layer_1_acc: 0.6581 - val_loss: 1.0511 - val_dense_2_loss: 0.4932 - val_center_loss_layer_1_loss: 1.1160 - val_dense_2_acc: 0.9494 - val_center_loss_layer_1_acc: 0.6174\n",
      "Epoch 62/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.5914 - dense_2_loss: 0.4047 - center_loss_layer_1_loss: 0.3733 - dense_2_acc: 0.9954 - center_loss_layer_1_acc: 0.7998 - val_loss: 1.0760 - val_dense_2_loss: 0.4741 - val_center_loss_layer_1_loss: 1.2037 - val_dense_2_acc: 0.9580 - val_center_loss_layer_1_acc: 0.6094\n",
      "Epoch 63/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 0.6733 - dense_2_loss: 0.4026 - center_loss_layer_1_loss: 0.5414 - dense_2_acc: 0.9905 - center_loss_layer_1_acc: 0.7092 - val_loss: 1.0172 - val_dense_2_loss: 0.4775 - val_center_loss_layer_1_loss: 1.0794 - val_dense_2_acc: 0.9558 - val_center_loss_layer_1_acc: 0.6691\n",
      "Epoch 64/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.7077 - dense_2_loss: 0.4043 - center_loss_layer_1_loss: 0.6068 - dense_2_acc: 0.9935 - center_loss_layer_1_acc: 0.6488 - val_loss: 1.6015 - val_dense_2_loss: 0.4834 - val_center_loss_layer_1_loss: 2.2361 - val_dense_2_acc: 0.9525 - val_center_loss_layer_1_acc: 0.5306\n",
      "Epoch 65/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.5699 - dense_2_loss: 0.3829 - center_loss_layer_1_loss: 0.3740 - dense_2_acc: 0.9964 - center_loss_layer_1_acc: 0.7933 - val_loss: 1.0679 - val_dense_2_loss: 0.4783 - val_center_loss_layer_1_loss: 1.1792 - val_dense_2_acc: 0.9482 - val_center_loss_layer_1_acc: 0.6373\n",
      "Epoch 66/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.6751 - dense_2_loss: 0.3849 - center_loss_layer_1_loss: 0.5805 - dense_2_acc: 0.9930 - center_loss_layer_1_acc: 0.6869 - val_loss: 1.2439 - val_dense_2_loss: 0.4838 - val_center_loss_layer_1_loss: 1.5202 - val_dense_2_acc: 0.9527 - val_center_loss_layer_1_acc: 0.5186\n",
      "Epoch 67/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.6646 - dense_2_loss: 0.3776 - center_loss_layer_1_loss: 0.5741 - dense_2_acc: 0.9921 - center_loss_layer_1_acc: 0.7118 - val_loss: 1.1312 - val_dense_2_loss: 0.4746 - val_center_loss_layer_1_loss: 1.3130 - val_dense_2_acc: 0.9510 - val_center_loss_layer_1_acc: 0.5653\n",
      "Epoch 68/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.9803 - dense_2_loss: 0.4474 - center_loss_layer_1_loss: 1.0657 - dense_2_acc: 0.9706 - center_loss_layer_1_acc: 0.6589 - val_loss: 1.6361 - val_dense_2_loss: 0.4793 - val_center_loss_layer_1_loss: 2.3136 - val_dense_2_acc: 0.9408 - val_center_loss_layer_1_acc: 0.4529\n",
      "Epoch 69/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.5806 - dense_2_loss: 0.4012 - center_loss_layer_1_loss: 0.3588 - dense_2_acc: 0.9958 - center_loss_layer_1_acc: 0.8002 - val_loss: 0.9202 - val_dense_2_loss: 0.4626 - val_center_loss_layer_1_loss: 0.9152 - val_dense_2_acc: 0.9637 - val_center_loss_layer_1_acc: 0.7149\n",
      "Epoch 70/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 0.5467 - dense_2_loss: 0.3705 - center_loss_layer_1_loss: 0.3524 - dense_2_acc: 0.9974 - center_loss_layer_1_acc: 0.8024 - val_loss: 1.0752 - val_dense_2_loss: 0.4666 - val_center_loss_layer_1_loss: 1.2172 - val_dense_2_acc: 0.9537 - val_center_loss_layer_1_acc: 0.5425\n",
      "Epoch 71/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.5510 - dense_2_loss: 0.3569 - center_loss_layer_1_loss: 0.3882 - dense_2_acc: 0.9983 - center_loss_layer_1_acc: 0.7641 - val_loss: 1.2115 - val_dense_2_loss: 0.4650 - val_center_loss_layer_1_loss: 1.4931 - val_dense_2_acc: 0.9533 - val_center_loss_layer_1_acc: 0.4893\n",
      "Epoch 72/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.5647 - dense_2_loss: 0.3490 - center_loss_layer_1_loss: 0.4314 - dense_2_acc: 0.9977 - center_loss_layer_1_acc: 0.7504 - val_loss: 1.3484 - val_dense_2_loss: 0.4591 - val_center_loss_layer_1_loss: 1.7786 - val_dense_2_acc: 0.9466 - val_center_loss_layer_1_acc: 0.3229\n",
      "Epoch 73/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.5937 - dense_2_loss: 0.3455 - center_loss_layer_1_loss: 0.4964 - dense_2_acc: 0.9965 - center_loss_layer_1_acc: 0.6968 - val_loss: 0.9860 - val_dense_2_loss: 0.4339 - val_center_loss_layer_1_loss: 1.1042 - val_dense_2_acc: 0.9562 - val_center_loss_layer_1_acc: 0.6770\n",
      "Epoch 74/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.6110 - dense_2_loss: 0.3450 - center_loss_layer_1_loss: 0.5319 - dense_2_acc: 0.9950 - center_loss_layer_1_acc: 0.7236 - val_loss: 1.1578 - val_dense_2_loss: 0.4487 - val_center_loss_layer_1_loss: 1.4183 - val_dense_2_acc: 0.9527 - val_center_loss_layer_1_acc: 0.5218\n",
      "Epoch 75/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.5956 - dense_2_loss: 0.3339 - center_loss_layer_1_loss: 0.5234 - dense_2_acc: 0.9964 - center_loss_layer_1_acc: 0.6993 - val_loss: 1.1643 - val_dense_2_loss: 0.4196 - val_center_loss_layer_1_loss: 1.4893 - val_dense_2_acc: 0.9556 - val_center_loss_layer_1_acc: 0.5243\n",
      "Epoch 76/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.5426 - dense_2_loss: 0.3232 - center_loss_layer_1_loss: 0.4389 - dense_2_acc: 0.9978 - center_loss_layer_1_acc: 0.7394 - val_loss: 1.0859 - val_dense_2_loss: 0.4145 - val_center_loss_layer_1_loss: 1.3427 - val_dense_2_acc: 0.9619 - val_center_loss_layer_1_acc: 0.5860\n",
      "Epoch 77/100\n",
      "30000/30000 [==============================] - 5s 164us/step - loss: 0.5812 - dense_2_loss: 0.3197 - center_loss_layer_1_loss: 0.5230 - dense_2_acc: 0.9976 - center_loss_layer_1_acc: 0.6848 - val_loss: 1.3034 - val_dense_2_loss: 0.4383 - val_center_loss_layer_1_loss: 1.7302 - val_dense_2_acc: 0.9462 - val_center_loss_layer_1_acc: 0.3623\n",
      "Epoch 78/100\n",
      "30000/30000 [==============================] - 5s 163us/step - loss: 0.6025 - dense_2_loss: 0.3184 - center_loss_layer_1_loss: 0.5682 - dense_2_acc: 0.9926 - center_loss_layer_1_acc: 0.6977 - val_loss: 1.9520 - val_dense_2_loss: 0.4838 - val_center_loss_layer_1_loss: 2.9364 - val_dense_2_acc: 0.8967 - val_center_loss_layer_1_acc: 0.3509\n",
      "Epoch 79/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.6418 - dense_2_loss: 0.3244 - center_loss_layer_1_loss: 0.6349 - dense_2_acc: 0.9957 - center_loss_layer_1_acc: 0.6594 - val_loss: 1.1709 - val_dense_2_loss: 0.4101 - val_center_loss_layer_1_loss: 1.5216 - val_dense_2_acc: 0.9549 - val_center_loss_layer_1_acc: 0.5639\n",
      "Epoch 80/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.5711 - dense_2_loss: 0.3096 - center_loss_layer_1_loss: 0.5229 - dense_2_acc: 0.9965 - center_loss_layer_1_acc: 0.7331 - val_loss: 1.0377 - val_dense_2_loss: 0.3915 - val_center_loss_layer_1_loss: 1.2924 - val_dense_2_acc: 0.9580 - val_center_loss_layer_1_acc: 0.5777\n",
      "Epoch 81/100\n",
      "30000/30000 [==============================] - 5s 163us/step - loss: 0.6162 - dense_2_loss: 0.3057 - center_loss_layer_1_loss: 0.6209 - dense_2_acc: 0.9978 - center_loss_layer_1_acc: 0.6508 - val_loss: 1.0640 - val_dense_2_loss: 0.4031 - val_center_loss_layer_1_loss: 1.3218 - val_dense_2_acc: 0.9556 - val_center_loss_layer_1_acc: 0.5970\n",
      "Epoch 82/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.5927 - dense_2_loss: 0.3017 - center_loss_layer_1_loss: 0.5819 - dense_2_acc: 0.9955 - center_loss_layer_1_acc: 0.6927 - val_loss: 1.2017 - val_dense_2_loss: 0.3994 - val_center_loss_layer_1_loss: 1.6045 - val_dense_2_acc: 0.9623 - val_center_loss_layer_1_acc: 0.5230\n",
      "Epoch 83/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.5469 - dense_2_loss: 0.2926 - center_loss_layer_1_loss: 0.5087 - dense_2_acc: 0.9976 - center_loss_layer_1_acc: 0.7067 - val_loss: 1.0675 - val_dense_2_loss: 0.3945 - val_center_loss_layer_1_loss: 1.3460 - val_dense_2_acc: 0.9507 - val_center_loss_layer_1_acc: 0.6193\n",
      "Epoch 84/100\n",
      "30000/30000 [==============================] - 5s 159us/step - loss: 0.6437 - dense_2_loss: 0.2944 - center_loss_layer_1_loss: 0.6986 - dense_2_acc: 0.9959 - center_loss_layer_1_acc: 0.6101 - val_loss: 1.3942 - val_dense_2_loss: 0.4027 - val_center_loss_layer_1_loss: 1.9830 - val_dense_2_acc: 0.9569 - val_center_loss_layer_1_acc: 0.4108\n",
      "Epoch 85/100\n",
      "30000/30000 [==============================] - 5s 161us/step - loss: 0.5895 - dense_2_loss: 0.2883 - center_loss_layer_1_loss: 0.6023 - dense_2_acc: 0.9970 - center_loss_layer_1_acc: 0.6671 - val_loss: 1.4516 - val_dense_2_loss: 0.4047 - val_center_loss_layer_1_loss: 2.0938 - val_dense_2_acc: 0.9518 - val_center_loss_layer_1_acc: 0.3178\n",
      "Epoch 86/100\n",
      "30000/30000 [==============================] - 5s 162us/step - loss: 0.5466 - dense_2_loss: 0.2780 - center_loss_layer_1_loss: 0.5372 - dense_2_acc: 0.9980 - center_loss_layer_1_acc: 0.7021 - val_loss: 0.9931 - val_dense_2_loss: 0.3701 - val_center_loss_layer_1_loss: 1.2461 - val_dense_2_acc: 0.9630 - val_center_loss_layer_1_acc: 0.6754\n",
      "Epoch 87/100\n",
      "30000/30000 [==============================] - 5s 163us/step - loss: 0.5206 - dense_2_loss: 0.2633 - center_loss_layer_1_loss: 0.5145 - dense_2_acc: 0.9988 - center_loss_layer_1_acc: 0.7079 - val_loss: 1.4329 - val_dense_2_loss: 0.3699 - val_center_loss_layer_1_loss: 2.1260 - val_dense_2_acc: 0.9500 - val_center_loss_layer_1_acc: 0.4784\n",
      "Epoch 88/100\n",
      "30000/30000 [==============================] - 5s 160us/step - loss: 0.5575 - dense_2_loss: 0.2611 - center_loss_layer_1_loss: 0.5927 - dense_2_acc: 0.9977 - center_loss_layer_1_acc: 0.6793 - val_loss: 1.3056 - val_dense_2_loss: 0.3664 - val_center_loss_layer_1_loss: 1.8783 - val_dense_2_acc: 0.9607 - val_center_loss_layer_1_acc: 0.4720\n",
      "Epoch 89/100\n",
      "30000/30000 [==============================] - 5s 163us/step - loss: 0.5554 - dense_2_loss: 0.2586 - center_loss_layer_1_loss: 0.5935 - dense_2_acc: 0.9980 - center_loss_layer_1_acc: 0.6441 - val_loss: 1.4146 - val_dense_2_loss: 0.4038 - val_center_loss_layer_1_loss: 2.0217 - val_dense_2_acc: 0.9160 - val_center_loss_layer_1_acc: 0.3472\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1a6173f81d0>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dummy_matrix1=np.zeros((x_train.shape[0],1))\n",
    "dummy_matrix2=np.zeros((x_test.shape[0],1))\n",
    "\n",
    "model_center_loss.fit(x=[x_train,y_train_one_hot],y=[y_train_one_hot,dummy_matrix1],\n",
    "                     batch_size=512,epochs=100,verbose=1,validation_data=([x_test,y_test_one_hot],[y_test_one_hot,dummy_matrix2]),\n",
    "                     callbacks=[EarlyStopping(monitor=\"val_loss\",patience=20)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 2)\n",
      "(10, 2)\n"
     ]
    }
   ],
   "source": [
    "func=K.function([model.input],[model.get_layer('out1').output])\n",
    "test_features= func([x_test])[0]\n",
    "print(test_features.shape)\n",
    "\n",
    "test_centers=np.dot(np.transpose(y_test_one_hot),test_features)\n",
    "test_centers_count=np.sum(np.transpose(y_test_one_hot),axis=1,keepdims=True)\n",
    "test_centers/=test_centers_count\n",
    "print(test_centers.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhAAAAFkCAYAAABxWwLDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a8963f0a58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.scatter(test_features[:,0],test_features[:,1],c=y_test,edgecolor=\"none\",s=5)\n",
    "plt.scatter(test_centers[:,0],test_centers[:,1],c=\"black\",marker=\"*\",edgecolor=\"none\",s=50)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}