深度殘差網路+自適應引數化ReLU啟用函式(調參記錄22)Cifar10~95.25%
本文在調參記錄21的基礎上,將殘差模組的個數,從60個增加到120個,測試深度殘差網路+自適應引數化ReLU啟用函式在Cifar10資料集上的效果。
自適應引數化ReLU啟用函式的基本原理如下:
Keras程式:
#!/usr/bin/env python3 # -*- coding: utf-8 -*- """ Created on Tue Apr 14 04:17:45 2020 Implemented using TensorFlow 1.10.0 and Keras 2.2.1 Minghang Zhao, Shisheng Zhong, Xuyun Fu, Baoping Tang, Shaojiang Dong, Michael Pecht, Deep Residual Networks with Adaptively Parametric Rectifier Linear Units for Fault Diagnosis, IEEE Transactions on Industrial Electronics, 2020, DOI: 10.1109/TIE.2020.2972458 @author: Minghang Zhao """ from __future__ import print_function import keras import numpy as np from keras.datasets import cifar10 from keras.layers import Dense, Conv2D, BatchNormalization, Activation, Minimum from keras.layers import AveragePooling2D, Input, GlobalAveragePooling2D, Concatenate, Reshape from keras.regularizers import l2 from keras import backend as K from keras.models import Model from keras import optimizers from keras.preprocessing.image import ImageDataGenerator from keras.callbacks import LearningRateScheduler K.set_learning_phase(1) # The data, split between train and test sets (x_train, y_train), (x_test, y_test) = cifar10.load_data() # Noised data x_train = x_train.astype('float32') / 255. x_test = x_test.astype('float32') / 255. x_test = x_test-np.mean(x_train) x_train = x_train-np.mean(x_train) print('x_train shape:', x_train.shape) print(x_train.shape[0], 'train samples') print(x_test.shape[0], 'test samples') # convert class vectors to binary class matrices y_train = keras.utils.to_categorical(y_train, 10) y_test = keras.utils.to_categorical(y_test, 10) # Schedule the learning rate, multiply 0.1 every 150 epoches def scheduler(epoch): if epoch % 150 == 0 and epoch != 0: lr = K.get_value(model.optimizer.lr) K.set_value(model.optimizer.lr, lr * 0.1) print("lr changed to {}".format(lr * 0.1)) return K.get_value(model.optimizer.lr) # An adaptively parametric rectifier linear unit (APReLU) def aprelu(inputs): # get the number of channels channels = inputs.get_shape().as_list()[-1] # get a zero feature map zeros_input = keras.layers.subtract([inputs, inputs]) # get a feature map with only positive features pos_input = Activation('relu')(inputs) # get a feature map with only negative features neg_input = Minimum()([inputs,zeros_input]) # define a network to obtain the scaling coefficients scales_p = GlobalAveragePooling2D()(pos_input) scales_n = GlobalAveragePooling2D()(neg_input) scales = Concatenate()([scales_n, scales_p]) scales = Dense(channels//16, activation='linear', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(scales) scales = BatchNormalization(momentum=0.9, gamma_regularizer=l2(1e-4))(scales) scales = Activation('relu')(scales) scales = Dense(channels, activation='linear', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(scales) scales = BatchNormalization(momentum=0.9, gamma_regularizer=l2(1e-4))(scales) scales = Activation('sigmoid')(scales) scales = Reshape((1,1,channels))(scales) # apply a paramtetric relu neg_part = keras.layers.multiply([scales, neg_input]) return keras.layers.add([pos_input, neg_part]) # Residual Block def residual_block(incoming, nb_blocks, out_channels, downsample=False, downsample_strides=2): residual = incoming in_channels = incoming.get_shape().as_list()[-1] for i in range(nb_blocks): identity = residual if not downsample: downsample_strides = 1 residual = BatchNormalization(momentum=0.9, gamma_regularizer=l2(1e-4))(residual) residual = Activation('relu')(residual) residual = Conv2D(out_channels, 3, strides=(downsample_strides, downsample_strides), padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(residual) residual = BatchNormalization(momentum=0.9, gamma_regularizer=l2(1e-4))(residual) residual = Activation('relu')(residual) residual = Conv2D(out_channels, 3, padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(residual) residual = aprelu(residual) # Downsampling if downsample_strides > 1: identity = AveragePooling2D(pool_size=(1,1), strides=(2,2))(identity) # Zero_padding to match channels if in_channels != out_channels: zeros_identity = keras.layers.subtract([identity, identity]) identity = keras.layers.concatenate([identity, zeros_identity]) in_channels = out_channels residual = keras.layers.add([residual, identity]) return residual # define and train a model inputs = Input(shape=(32, 32, 3)) net = Conv2D(16, 3, padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(inputs) net = residual_block(net, 40, 32, downsample=False) net = residual_block(net, 1, 32, downsample=True) net = residual_block(net, 39, 32, downsample=False) net = residual_block(net, 1, 64, downsample=True) net = residual_block(net, 39, 64, downsample=False) net = BatchNormalization(momentum=0.9, gamma_regularizer=l2(1e-4))(net) net = Activation('relu')(net) net = GlobalAveragePooling2D()(net) outputs = Dense(10, activation='softmax', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(net) model = Model(inputs=inputs, outputs=outputs) sgd = optimizers.SGD(lr=0.1, decay=0., momentum=0.9, nesterov=True) model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy']) # data augmentation datagen = ImageDataGenerator( # randomly rotate images in the range (deg 0 to 180) rotation_range=30, # Range for random zoom zoom_range = 0.2, # shear angle in counter-clockwise direction in degrees shear_range = 30, # randomly flip images horizontal_flip=True, # randomly shift images horizontally width_shift_range=0.125, # randomly shift images vertically height_shift_range=0.125) reduce_lr = LearningRateScheduler(scheduler) # fit the model on the batches generated by datagen.flow(). model.fit_generator(datagen.flow(x_train, y_train, batch_size=100), validation_data=(x_test, y_test), epochs=500, verbose=1, callbacks=[reduce_lr], workers=4) # get results K.set_learning_phase(0) DRSN_train_score = model.evaluate(x_train, y_train, batch_size=100, verbose=0) print('Train loss:', DRSN_train_score[0]) print('Train accuracy:', DRSN_train_score[1]) DRSN_test_score = model.evaluate(x_test, y_test, batch_size=100, verbose=0) print('Test loss:', DRSN_test_score[0]) print('Test accuracy:', DRSN_test_score[1])
實驗結果:
Using TensorFlow backend. x_train shape: (50000, 32, 32, 3) 50000 train samples 10000 test samples Epoch 1/500 318s 637ms/step - loss: 6.0165 - acc: 0.3791 - val_loss: 5.2157 - val_acc: 0.5307 Epoch 2/500 221s 443ms/step - loss: 4.8584 - acc: 0.5361 - val_loss: 4.2761 - val_acc: 0.6383 Epoch 3/500 221s 442ms/step - loss: 4.0487 - acc: 0.6159 - val_loss: 3.5837 - val_acc: 0.6913 Epoch 4/500 221s 442ms/step - loss: 3.4323 - acc: 0.6610 - val_loss: 3.0189 - val_acc: 0.7396 Epoch 5/500 221s 442ms/step - loss: 2.9384 - acc: 0.6943 - val_loss: 2.5795 - val_acc: 0.7697 Epoch 6/500 221s 442ms/step - loss: 2.5470 - acc: 0.7181 - val_loss: 2.2296 - val_acc: 0.7848 Epoch 7/500 221s 442ms/step - loss: 2.2227 - acc: 0.7400 - val_loss: 1.9631 - val_acc: 0.7931 Epoch 8/500 222s 444ms/step - loss: 1.9632 - acc: 0.7546 - val_loss: 1.7318 - val_acc: 0.8098 Epoch 9/500 221s 443ms/step - loss: 1.7535 - acc: 0.7685 - val_loss: 1.5313 - val_acc: 0.8197 Epoch 10/500 221s 442ms/step - loss: 1.5759 - acc: 0.7798 - val_loss: 1.4001 - val_acc: 0.8214 Epoch 11/500 221s 442ms/step - loss: 1.4432 - acc: 0.7859 - val_loss: 1.2776 - val_acc: 0.8309 Epoch 12/500 221s 442ms/step - loss: 1.3201 - acc: 0.7977 - val_loss: 1.1707 - val_acc: 0.8349 Epoch 13/500 222s 443ms/step - loss: 1.2295 - acc: 0.8028 - val_loss: 1.0760 - val_acc: 0.8454 Epoch 14/500 222s 443ms/step - loss: 1.1552 - acc: 0.8069 - val_loss: 1.0225 - val_acc: 0.8432 Epoch 15/500 221s 441ms/step - loss: 1.0964 - acc: 0.8119 - val_loss: 0.9549 - val_acc: 0.8514 Epoch 16/500 221s 442ms/step - loss: 1.0386 - acc: 0.8174 - val_loss: 0.9072 - val_acc: 0.8614 Epoch 17/500 221s 442ms/step - loss: 0.9979 - acc: 0.8204 - val_loss: 0.8765 - val_acc: 0.8566 Epoch 18/500 221s 441ms/step - loss: 0.9611 - acc: 0.8260 - val_loss: 0.8820 - val_acc: 0.8502 Epoch 19/500 220s 441ms/step - loss: 0.9351 - acc: 0.8290 - val_loss: 0.8319 - val_acc: 0.8601 Epoch 20/500 221s 441ms/step - loss: 0.9130 - acc: 0.8295 - val_loss: 0.8077 - val_acc: 0.8643 Epoch 21/500 221s 441ms/step - loss: 0.8837 - acc: 0.8347 - val_loss: 0.7924 - val_acc: 0.8683 Epoch 22/500 221s 441ms/step - loss: 0.8741 - acc: 0.8349 - val_loss: 0.7675 - val_acc: 0.8747 Epoch 23/500 221s 442ms/step - loss: 0.8536 - acc: 0.8403 - val_loss: 0.7988 - val_acc: 0.8599 Epoch 24/500 221s 441ms/step - loss: 0.8457 - acc: 0.8395 - val_loss: 0.7619 - val_acc: 0.8698 Epoch 25/500 221s 441ms/step - loss: 0.8354 - acc: 0.8422 - val_loss: 0.7466 - val_acc: 0.8708 Epoch 26/500 221s 441ms/step - loss: 0.8210 - acc: 0.8449 - val_loss: 0.7481 - val_acc: 0.8714 Epoch 27/500 220s 441ms/step - loss: 0.8155 - acc: 0.8473 - val_loss: 0.7636 - val_acc: 0.8669 Epoch 28/500 220s 441ms/step - loss: 0.8154 - acc: 0.8470 - val_loss: 0.7301 - val_acc: 0.8785 Epoch 29/500 220s 441ms/step - loss: 0.7967 - acc: 0.8537 - val_loss: 0.7206 - val_acc: 0.8811 Epoch 30/500 220s 441ms/step - loss: 0.7961 - acc: 0.8510 - val_loss: 0.7203 - val_acc: 0.8814 Epoch 31/500 221s 441ms/step - loss: 0.7932 - acc: 0.8534 - val_loss: 0.7010 - val_acc: 0.8835 Epoch 32/500 220s 441ms/step - loss: 0.7783 - acc: 0.8585 - val_loss: 0.7239 - val_acc: 0.8797 Epoch 33/500 221s 441ms/step - loss: 0.7744 - acc: 0.8577 - val_loss: 0.7140 - val_acc: 0.8795 Epoch 34/500 221s 442ms/step - loss: 0.7753 - acc: 0.8591 - val_loss: 0.7185 - val_acc: 0.8811 Epoch 35/500 221s 441ms/step - loss: 0.7737 - acc: 0.8575 - val_loss: 0.7251 - val_acc: 0.8752 Epoch 36/500 220s 441ms/step - loss: 0.7666 - acc: 0.8632 - val_loss: 0.7151 - val_acc: 0.8814 Epoch 37/500 221s 441ms/step - loss: 0.7746 - acc: 0.8593 - val_loss: 0.7119 - val_acc: 0.8792 Epoch 38/500 220s 441ms/step - loss: 0.7644 - acc: 0.8631 - val_loss: 0.7091 - val_acc: 0.8819 Epoch 39/500 221s 442ms/step - loss: 0.7620 - acc: 0.8639 - val_loss: 0.7190 - val_acc: 0.8809 Epoch 40/500 221s 441ms/step - loss: 0.7507 - acc: 0.8660 - val_loss: 0.7065 - val_acc: 0.8840 Epoch 41/500 221s 441ms/step - loss: 0.7550 - acc: 0.8658 - val_loss: 0.6998 - val_acc: 0.8858 Epoch 42/500 220s 441ms/step - loss: 0.7546 - acc: 0.8666 - val_loss: 0.7195 - val_acc: 0.8803 Epoch 43/500 221s 441ms/step - loss: 0.7514 - acc: 0.8680 - val_loss: 0.6949 - val_acc: 0.8895 Epoch 44/500 220s 441ms/step - loss: 0.7511 - acc: 0.8661 - val_loss: 0.7011 - val_acc: 0.8872 Epoch 45/500 221s 441ms/step - loss: 0.7431 - acc: 0.8688 - val_loss: 0.7057 - val_acc: 0.8848 Epoch 46/500 221s 441ms/step - loss: 0.7464 - acc: 0.8683 - val_loss: 0.7014 - val_acc: 0.8827 Epoch 47/500 220s 441ms/step - loss: 0.7487 - acc: 0.8678 - val_loss: 0.7002 - val_acc: 0.8859 Epoch 48/500 220s 441ms/step - loss: 0.7453 - acc: 0.8701 - val_loss: 0.6912 - val_acc: 0.8891 Epoch 49/500 220s 441ms/step - loss: 0.7431 - acc: 0.8694 - val_loss: 0.6798 - val_acc: 0.8932 Epoch 50/500 221s 441ms/step - loss: 0.7409 - acc: 0.8726 - val_loss: 0.6813 - val_acc: 0.8949 Epoch 51/500 220s 440ms/step - loss: 0.7370 - acc: 0.8732 - val_loss: 0.7049 - val_acc: 0.8886 Epoch 52/500 220s 441ms/step - loss: 0.7315 - acc: 0.8748 - val_loss: 0.6921 - val_acc: 0.8881 Epoch 53/500 220s 441ms/step - loss: 0.7374 - acc: 0.8728 - val_loss: 0.6728 - val_acc: 0.8990 Epoch 54/500 220s 441ms/step - loss: 0.7326 - acc: 0.8749 - val_loss: 0.6982 - val_acc: 0.8861 Epoch 55/500 221s 441ms/step - loss: 0.7353 - acc: 0.8723 - val_loss: 0.6776 - val_acc: 0.8918 Epoch 56/500 220s 441ms/step - loss: 0.7300 - acc: 0.8752 - val_loss: 0.6822 - val_acc: 0.8923 Epoch 57/500 221s 441ms/step - loss: 0.7321 - acc: 0.8756 - val_loss: 0.6854 - val_acc: 0.8963 Epoch 58/500 221s 441ms/step - loss: 0.7341 - acc: 0.8742 - val_loss: 0.7065 - val_acc: 0.8861 Epoch 59/500 220s 441ms/step - loss: 0.7334 - acc: 0.8749 - val_loss: 0.6815 - val_acc: 0.8960 Epoch 60/500 221s 443ms/step - loss: 0.7250 - acc: 0.8774 - val_loss: 0.6798 - val_acc: 0.8977 Epoch 61/500 222s 444ms/step - loss: 0.7309 - acc: 0.8759 - val_loss: 0.6892 - val_acc: 0.8964 Epoch 62/500 225s 451ms/step - loss: 0.7249 - acc: 0.8784 - val_loss: 0.6967 - val_acc: 0.8923 Epoch 63/500 226s 451ms/step - loss: 0.7291 - acc: 0.8770 - val_loss: 0.7028 - val_acc: 0.8907 Epoch 64/500 225s 451ms/step - loss: 0.7234 - acc: 0.8817 - val_loss: 0.6920 - val_acc: 0.8903 Epoch 65/500 225s 451ms/step - loss: 0.7279 - acc: 0.8787 - val_loss: 0.6723 - val_acc: 0.9003 Epoch 66/500 225s 451ms/step - loss: 0.7229 - acc: 0.8801 - val_loss: 0.6937 - val_acc: 0.8939 Epoch 67/500 225s 450ms/step - loss: 0.7207 - acc: 0.8795 - val_loss: 0.7028 - val_acc: 0.8928 Epoch 68/500 226s 451ms/step - loss: 0.7226 - acc: 0.8804 - val_loss: 0.6830 - val_acc: 0.8942 Epoch 69/500 225s 451ms/step - loss: 0.7210 - acc: 0.8801 - val_loss: 0.6928 - val_acc: 0.8941 Epoch 70/500 225s 451ms/step - loss: 0.7197 - acc: 0.8817 - val_loss: 0.6946 - val_acc: 0.8912 Epoch 71/500 225s 450ms/step - loss: 0.7228 - acc: 0.8799 - val_loss: 0.6721 - val_acc: 0.9010 Epoch 72/500 226s 451ms/step - loss: 0.7195 - acc: 0.8836 - val_loss: 0.6764 - val_acc: 0.9032 Epoch 73/500 225s 450ms/step - loss: 0.7210 - acc: 0.8810 - val_loss: 0.6776 - val_acc: 0.8979 Epoch 74/500 225s 451ms/step - loss: 0.7174 - acc: 0.8819 - val_loss: 0.6784 - val_acc: 0.8965 Epoch 75/500 225s 451ms/step - loss: 0.7144 - acc: 0.8838 - val_loss: 0.6799 - val_acc: 0.8988 Epoch 76/500 225s 451ms/step - loss: 0.7188 - acc: 0.8814 - val_loss: 0.6884 - val_acc: 0.8945 Epoch 77/500 225s 451ms/step - loss: 0.7188 - acc: 0.8833 - val_loss: 0.7054 - val_acc: 0.8915 Epoch 78/500 225s 451ms/step - loss: 0.7147 - acc: 0.8835 - val_loss: 0.6905 - val_acc: 0.8957 Epoch 79/500 225s 451ms/step - loss: 0.7168 - acc: 0.8830 - val_loss: 0.6794 - val_acc: 0.9000 Epoch 80/500 225s 451ms/step - loss: 0.7150 - acc: 0.8829 - val_loss: 0.6843 - val_acc: 0.8957 Epoch 81/500 225s 451ms/step - loss: 0.7102 - acc: 0.8846 - val_loss: 0.6813 - val_acc: 0.8954 Epoch 82/500 226s 451ms/step - loss: 0.7093 - acc: 0.8844 - val_loss: 0.6944 - val_acc: 0.8913 Epoch 83/500 225s 451ms/step - loss: 0.7105 - acc: 0.8840 - val_loss: 0.6791 - val_acc: 0.8964 Epoch 84/500 225s 451ms/step - loss: 0.7068 - acc: 0.8872 - val_loss: 0.6921 - val_acc: 0.8905 Epoch 85/500 225s 451ms/step - loss: 0.7118 - acc: 0.8866 - val_loss: 0.6970 - val_acc: 0.8921 Epoch 86/500 225s 451ms/step - loss: 0.7108 - acc: 0.8842 - val_loss: 0.6891 - val_acc: 0.8955 Epoch 87/500 225s 451ms/step - loss: 0.7105 - acc: 0.8832 - val_loss: 0.6872 - val_acc: 0.8949 Epoch 88/500 225s 451ms/step - loss: 0.7133 - acc: 0.8846 - val_loss: 0.6777 - val_acc: 0.8978 Epoch 89/500 225s 451ms/step - loss: 0.7105 - acc: 0.8853 - val_loss: 0.6784 - val_acc: 0.8953 Epoch 90/500 225s 451ms/step - loss: 0.7031 - acc: 0.8884 - val_loss: 0.6937 - val_acc: 0.8952 Epoch 91/500 225s 451ms/step - loss: 0.7002 - acc: 0.8892 - val_loss: 0.6709 - val_acc: 0.9001 Epoch 92/500 225s 451ms/step - loss: 0.7098 - acc: 0.8863 - val_loss: 0.6674 - val_acc: 0.9002 Epoch 93/500 225s 451ms/step - loss: 0.7034 - acc: 0.8882 - val_loss: 0.7211 - val_acc: 0.8831 Epoch 94/500 225s 450ms/step - loss: 0.7056 - acc: 0.8870 - val_loss: 0.6597 - val_acc: 0.9043 Epoch 95/500 225s 450ms/step - loss: 0.7070 - acc: 0.8861 - val_loss: 0.6682 - val_acc: 0.9026 Epoch 96/500 221s 442ms/step - loss: 0.7015 - acc: 0.8893 - val_loss: 0.6766 - val_acc: 0.9009 Epoch 97/500 224s 448ms/step - loss: 0.7089 - acc: 0.8855 - val_loss: 0.6844 - val_acc: 0.8970 Epoch 98/500 225s 450ms/step - loss: 0.7052 - acc: 0.8885 - val_loss: 0.6668 - val_acc: 0.9040 Epoch 99/500 225s 451ms/step - loss: 0.7072 - acc: 0.8879 - val_loss: 0.6808 - val_acc: 0.8978 Epoch 100/500 225s 451ms/step - loss: 0.7016 - acc: 0.8891 - val_loss: 0.6898 - val_acc: 0.8935 Epoch 101/500 225s 451ms/step - loss: 0.7018 - acc: 0.8888 - val_loss: 0.6803 - val_acc: 0.8980 Epoch 102/500 225s 451ms/step - loss: 0.7099 - acc: 0.8865 - val_loss: 0.6773 - val_acc: 0.8986 Epoch 103/500 225s 451ms/step - loss: 0.7075 - acc: 0.8875 - val_loss: 0.6743 - val_acc: 0.9014 Epoch 104/500 225s 451ms/step - loss: 0.7048 - acc: 0.8881 - val_loss: 0.6627 - val_acc: 0.9064 Epoch 105/500 225s 451ms/step - loss: 0.7041 - acc: 0.8890 - val_loss: 0.6741 - val_acc: 0.9032 Epoch 106/500 226s 451ms/step - loss: 0.7036 - acc: 0.8884 - val_loss: 0.6736 - val_acc: 0.9037 Epoch 107/500 225s 451ms/step - loss: 0.7043 - acc: 0.8882 - val_loss: 0.6758 - val_acc: 0.9005 Epoch 108/500 225s 451ms/step - loss: 0.7024 - acc: 0.8891 - val_loss: 0.6812 - val_acc: 0.8990 Epoch 109/500 225s 450ms/step - loss: 0.7044 - acc: 0.8872 - val_loss: 0.6736 - val_acc: 0.9016 Epoch 110/500 225s 451ms/step - loss: 0.6999 - acc: 0.8913 - val_loss: 0.6756 - val_acc: 0.9007 Epoch 111/500 226s 451ms/step - loss: 0.6951 - acc: 0.8930 - val_loss: 0.6871 - val_acc: 0.8945 Epoch 112/500 226s 451ms/step - loss: 0.6970 - acc: 0.8898 - val_loss: 0.6875 - val_acc: 0.8950 Epoch 113/500 225s 450ms/step - loss: 0.7006 - acc: 0.8902 - val_loss: 0.6711 - val_acc: 0.9032 Epoch 114/500 225s 451ms/step - loss: 0.7000 - acc: 0.8896 - val_loss: 0.6824 - val_acc: 0.8962 Epoch 115/500 225s 450ms/step - loss: 0.6969 - acc: 0.8904 - val_loss: 0.6761 - val_acc: 0.8975 Epoch 116/500 225s 451ms/step - loss: 0.6939 - acc: 0.8913 - val_loss: 0.6924 - val_acc: 0.8974 Epoch 117/500 225s 451ms/step - loss: 0.7028 - acc: 0.8895 - val_loss: 0.6773 - val_acc: 0.9014 Epoch 118/500 225s 450ms/step - loss: 0.6994 - acc: 0.8906 - val_loss: 0.7111 - val_acc: 0.8884 Epoch 119/500 225s 451ms/step - loss: 0.7059 - acc: 0.8889 - val_loss: 0.6947 - val_acc: 0.8955 Epoch 120/500 226s 451ms/step - loss: 0.7000 - acc: 0.8902 - val_loss: 0.6832 - val_acc: 0.8976 Epoch 121/500 225s 451ms/step - loss: 0.6976 - acc: 0.8911 - val_loss: 0.6770 - val_acc: 0.9027 Epoch 122/500 226s 451ms/step - loss: 0.6962 - acc: 0.8918 - val_loss: 0.7034 - val_acc: 0.8925 Epoch 123/500 225s 451ms/step - loss: 0.6908 - acc: 0.8940 - val_loss: 0.6872 - val_acc: 0.8974 Epoch 124/500 225s 450ms/step - loss: 0.7004 - acc: 0.8896 - val_loss: 0.6788 - val_acc: 0.8979 Epoch 125/500 225s 451ms/step - loss: 0.6953 - acc: 0.8924 - val_loss: 0.6973 - val_acc: 0.8933 Epoch 126/500 225s 451ms/step - loss: 0.6998 - acc: 0.8913 - val_loss: 0.6845 - val_acc: 0.8960 Epoch 127/500 225s 451ms/step - loss: 0.7004 - acc: 0.8908 - val_loss: 0.6787 - val_acc: 0.9009 Epoch 128/500 225s 451ms/step - loss: 0.7020 - acc: 0.8898 - val_loss: 0.6899 - val_acc: 0.8970 Epoch 129/500 225s 451ms/step - loss: 0.6948 - acc: 0.8928 - val_loss: 0.6748 - val_acc: 0.9026 Epoch 130/500 221s 443ms/step - loss: 0.6958 - acc: 0.8922 - val_loss: 0.6656 - val_acc: 0.9032 Epoch 131/500 220s 440ms/step - loss: 0.6950 - acc: 0.8929 - val_loss: 0.7022 - val_acc: 0.8930 Epoch 132/500 220s 441ms/step - loss: 0.6973 - acc: 0.8918 - val_loss: 0.6895 - val_acc: 0.8919 Epoch 133/500 220s 441ms/step - loss: 0.6927 - acc: 0.8932 - val_loss: 0.6894 - val_acc: 0.8936 Epoch 134/500 220s 441ms/step - loss: 0.6966 - acc: 0.8910 - val_loss: 0.6820 - val_acc: 0.8989 Epoch 135/500 220s 441ms/step - loss: 0.7001 - acc: 0.8912 - val_loss: 0.6699 - val_acc: 0.9020 Epoch 136/500 221s 441ms/step - loss: 0.6937 - acc: 0.8935 - val_loss: 0.6767 - val_acc: 0.8963 Epoch 137/500 221s 442ms/step - loss: 0.6915 - acc: 0.8933 - val_loss: 0.6711 - val_acc: 0.9046 Epoch 138/500 221s 441ms/step - loss: 0.6897 - acc: 0.8931 - val_loss: 0.6808 - val_acc: 0.8977 Epoch 139/500 220s 440ms/step - loss: 0.6931 - acc: 0.8921 - val_loss: 0.6911 - val_acc: 0.8960 Epoch 140/500 221s 441ms/step - loss: 0.6956 - acc: 0.8927 - val_loss: 0.6699 - val_acc: 0.9041 Epoch 141/500 220s 441ms/step - loss: 0.6909 - acc: 0.8939 - val_loss: 0.6755 - val_acc: 0.8995 Epoch 142/500 221s 441ms/step - loss: 0.6919 - acc: 0.8943 - val_loss: 0.6701 - val_acc: 0.9018 Epoch 143/500 221s 442ms/step - loss: 0.6932 - acc: 0.8930 - val_loss: 0.6764 - val_acc: 0.9030 Epoch 144/500 221s 441ms/step - loss: 0.6964 - acc: 0.8930 - val_loss: 0.6952 - val_acc: 0.8951 Epoch 145/500 220s 441ms/step - loss: 0.6910 - acc: 0.8926 - val_loss: 0.6635 - val_acc: 0.9064 Epoch 146/500 220s 441ms/step - loss: 0.6973 - acc: 0.8925 - val_loss: 0.6861 - val_acc: 0.8976 Epoch 147/500 220s 440ms/step - loss: 0.6910 - acc: 0.8927 - val_loss: 0.6739 - val_acc: 0.9041 Epoch 148/500 220s 441ms/step - loss: 0.6919 - acc: 0.8936 - val_loss: 0.6705 - val_acc: 0.9049 Epoch 149/500 220s 441ms/step - loss: 0.6925 - acc: 0.8936 - val_loss: 0.6694 - val_acc: 0.9025 Epoch 150/500 220s 441ms/step - loss: 0.6944 - acc: 0.8928 - val_loss: 0.6793 - val_acc: 0.8986 Epoch 151/500 lr changed to 0.010000000149011612 220s 441ms/step - loss: 0.5821 - acc: 0.9317 - val_loss: 0.5776 - val_acc: 0.9323 Epoch 152/500 220s 441ms/step - loss: 0.5235 - acc: 0.9495 - val_loss: 0.5587 - val_acc: 0.9370 Epoch 153/500 220s 441ms/step - loss: 0.5024 - acc: 0.9543 - val_loss: 0.5500 - val_acc: 0.9381 Epoch 154/500 221s 441ms/step - loss: 0.4852 - acc: 0.9583 - val_loss: 0.5434 - val_acc: 0.9393 Epoch 155/500 220s 440ms/step - loss: 0.4739 - acc: 0.9607 - val_loss: 0.5420 - val_acc: 0.9374 Epoch 156/500 220s 440ms/step - loss: 0.4595 - acc: 0.9631 - val_loss: 0.5295 - val_acc: 0.9397 Epoch 157/500 221s 441ms/step - loss: 0.4497 - acc: 0.9647 - val_loss: 0.5211 - val_acc: 0.9406 Epoch 158/500 220s 441ms/step - loss: 0.4421 - acc: 0.9653 - val_loss: 0.5143 - val_acc: 0.9411 Epoch 159/500 221s 441ms/step - loss: 0.4317 - acc: 0.9660 - val_loss: 0.5100 - val_acc: 0.9416 Epoch 160/500 221s 441ms/step - loss: 0.4200 - acc: 0.9692 - val_loss: 0.5001 - val_acc: 0.9459 Epoch 161/500 221s 441ms/step - loss: 0.4136 - acc: 0.9686 - val_loss: 0.4992 - val_acc: 0.9447 Epoch 162/500 220s 440ms/step - loss: 0.4050 - acc: 0.9708 - val_loss: 0.4958 - val_acc: 0.9420 Epoch 163/500 220s 441ms/step - loss: 0.4000 - acc: 0.9709 - val_loss: 0.4927 - val_acc: 0.9432 Epoch 164/500 220s 441ms/step - loss: 0.3903 - acc: 0.9721 - val_loss: 0.4920 - val_acc: 0.9431 Epoch 165/500 220s 441ms/step - loss: 0.3828 - acc: 0.9730 - val_loss: 0.4873 - val_acc: 0.9426 Epoch 166/500 220s 441ms/step - loss: 0.3785 - acc: 0.9733 - val_loss: 0.4890 - val_acc: 0.9394 Epoch 167/500 220s 441ms/step - loss: 0.3722 - acc: 0.9735 - val_loss: 0.4821 - val_acc: 0.9408 Epoch 168/500 221s 441ms/step - loss: 0.3642 - acc: 0.9755 - val_loss: 0.4671 - val_acc: 0.9428 Epoch 169/500 220s 441ms/step - loss: 0.3602 - acc: 0.9751 - val_loss: 0.4627 - val_acc: 0.9444 Epoch 170/500 220s 441ms/step - loss: 0.3544 - acc: 0.9756 - val_loss: 0.4749 - val_acc: 0.9396 Epoch 171/500 221s 441ms/step - loss: 0.3498 - acc: 0.9758 - val_loss: 0.4694 - val_acc: 0.9420 Epoch 172/500 221s 442ms/step - loss: 0.3465 - acc: 0.9761 - val_loss: 0.4702 - val_acc: 0.9391 Epoch 173/500 220s 441ms/step - loss: 0.3428 - acc: 0.9761 - val_loss: 0.4564 - val_acc: 0.9429 Epoch 174/500 224s 447ms/step - loss: 0.3382 - acc: 0.9763 - val_loss: 0.4583 - val_acc: 0.9406 Epoch 175/500 221s 442ms/step - loss: 0.3277 - acc: 0.9789 - val_loss: 0.4522 - val_acc: 0.9418 Epoch 176/500 224s 448ms/step - loss: 0.3287 - acc: 0.9769 - val_loss: 0.4466 - val_acc: 0.9420 Epoch 177/500 225s 450ms/step - loss: 0.3249 - acc: 0.9766 - val_loss: 0.4433 - val_acc: 0.9435 Epoch 178/500 225s 450ms/step - loss: 0.3182 - acc: 0.9785 - val_loss: 0.4391 - val_acc: 0.9420 Epoch 179/500 223s 447ms/step - loss: 0.3148 - acc: 0.9781 - val_loss: 0.4420 - val_acc: 0.9387 Epoch 180/500 221s 442ms/step - loss: 0.3119 - acc: 0.9781 - val_loss: 0.4455 - val_acc: 0.9399 Epoch 181/500 222s 443ms/step - loss: 0.3051 - acc: 0.9795 - val_loss: 0.4441 - val_acc: 0.9388 Epoch 182/500 225s 450ms/step - loss: 0.3056 - acc: 0.9792 - val_loss: 0.4391 - val_acc: 0.9390 Epoch 183/500 224s 448ms/step - loss: 0.3042 - acc: 0.9779 - val_loss: 0.4373 - val_acc: 0.9397 Epoch 184/500 221s 442ms/step - loss: 0.2972 - acc: 0.9791 - val_loss: 0.4329 - val_acc: 0.9398 Epoch 185/500 225s 450ms/step - loss: 0.2929 - acc: 0.9800 - val_loss: 0.4325 - val_acc: 0.9401 Epoch 186/500 225s 450ms/step - loss: 0.2905 - acc: 0.9807 - val_loss: 0.4300 - val_acc: 0.9392 Epoch 187/500 225s 450ms/step - loss: 0.2876 - acc: 0.9803 - val_loss: 0.4242 - val_acc: 0.9405 Epoch 188/500 225s 450ms/step - loss: 0.2842 - acc: 0.9806 - val_loss: 0.4245 - val_acc: 0.9396 Epoch 189/500 225s 450ms/step - loss: 0.2856 - acc: 0.9785 - val_loss: 0.4227 - val_acc: 0.9406 Epoch 190/500 224s 448ms/step - loss: 0.2823 - acc: 0.9794 - val_loss: 0.4057 - val_acc: 0.9425 Epoch 191/500 221s 442ms/step - loss: 0.2757 - acc: 0.9811 - val_loss: 0.4065 - val_acc: 0.9422 Epoch 192/500 221s 442ms/step - loss: 0.2775 - acc: 0.9799 - val_loss: 0.4066 - val_acc: 0.9430 Epoch 193/500 222s 445ms/step - loss: 0.2756 - acc: 0.9796 - val_loss: 0.4032 - val_acc: 0.9419 Epoch 194/500 225s 450ms/step - loss: 0.2733 - acc: 0.9795 - val_loss: 0.4105 - val_acc: 0.9391 Epoch 195/500 225s 451ms/step - loss: 0.2689 - acc: 0.9809 - val_loss: 0.4044 - val_acc: 0.9418 Epoch 196/500 225s 450ms/step - loss: 0.2678 - acc: 0.9802 - val_loss: 0.3969 - val_acc: 0.9425 Epoch 197/500 225s 450ms/step - loss: 0.2612 - acc: 0.9825 - val_loss: 0.3984 - val_acc: 0.9437 Epoch 198/500 225s 450ms/step - loss: 0.2686 - acc: 0.9783 - val_loss: 0.4037 - val_acc: 0.9374 Epoch 199/500 225s 450ms/step - loss: 0.2650 - acc: 0.9796 - val_loss: 0.3946 - val_acc: 0.9383 Epoch 200/500 225s 450ms/step - loss: 0.2595 - acc: 0.9804 - val_loss: 0.3933 - val_acc: 0.9401 Epoch 201/500 225s 450ms/step - loss: 0.2575 - acc: 0.9813 - val_loss: 0.3920 - val_acc: 0.9396 Epoch 202/500 225s 450ms/step - loss: 0.2610 - acc: 0.9788 - val_loss: 0.3916 - val_acc: 0.9383 Epoch 203/500 224s 448ms/step - loss: 0.2591 - acc: 0.9796 - val_loss: 0.4071 - val_acc: 0.9366 Epoch 204/500 221s 442ms/step - loss: 0.2575 - acc: 0.9794 - val_loss: 0.3900 - val_acc: 0.9390 Epoch 205/500 224s 448ms/step - loss: 0.2534 - acc: 0.9801 - val_loss: 0.3909 - val_acc: 0.9394 Epoch 206/500 225s 450ms/step - loss: 0.2505 - acc: 0.9813 - val_loss: 0.3957 - val_acc: 0.9407 Epoch 207/500 222s 443ms/step - loss: 0.2473 - acc: 0.9808 - val_loss: 0.3851 - val_acc: 0.9403 Epoch 208/500 220s 441ms/step - loss: 0.2490 - acc: 0.9803 - val_loss: 0.3753 - val_acc: 0.9435 Epoch 209/500 220s 441ms/step - loss: 0.2467 - acc: 0.9808 - val_loss: 0.3765 - val_acc: 0.9431 Epoch 210/500 220s 441ms/step - loss: 0.2457 - acc: 0.9805 - val_loss: 0.3830 - val_acc: 0.9407 Epoch 211/500 220s 441ms/step - loss: 0.2430 - acc: 0.9815 - val_loss: 0.3849 - val_acc: 0.9414 Epoch 212/500 221s 441ms/step - loss: 0.2483 - acc: 0.9789 - val_loss: 0.3818 - val_acc: 0.9407 Epoch 213/500 221s 441ms/step - loss: 0.2394 - acc: 0.9812 - val_loss: 0.3814 - val_acc: 0.9384 Epoch 214/500 220s 441ms/step - loss: 0.2425 - acc: 0.9797 - val_loss: 0.3818 - val_acc: 0.9400 Epoch 215/500 221s 441ms/step - loss: 0.2417 - acc: 0.9802 - val_loss: 0.3813 - val_acc: 0.9381 Epoch 216/500 220s 441ms/step - loss: 0.2433 - acc: 0.9790 - val_loss: 0.3790 - val_acc: 0.9382 Epoch 217/500 220s 441ms/step - loss: 0.2413 - acc: 0.9800 - val_loss: 0.3832 - val_acc: 0.9388 Epoch 218/500 221s 442ms/step - loss: 0.2386 - acc: 0.9798 - val_loss: 0.3793 - val_acc: 0.9377 Epoch 219/500 220s 441ms/step - loss: 0.2396 - acc: 0.9797 - val_loss: 0.3909 - val_acc: 0.9340 Epoch 220/500 221s 441ms/step - loss: 0.2376 - acc: 0.9796 - val_loss: 0.3930 - val_acc: 0.9364 Epoch 221/500 221s 441ms/step - loss: 0.2365 - acc: 0.9806 - val_loss: 0.3738 - val_acc: 0.9370 Epoch 222/500 220s 441ms/step - loss: 0.2398 - acc: 0.9785 - val_loss: 0.3940 - val_acc: 0.9340 Epoch 223/500 220s 441ms/step - loss: 0.2359 - acc: 0.9800 - val_loss: 0.3768 - val_acc: 0.9411 Epoch 224/500 221s 441ms/step - loss: 0.2365 - acc: 0.9795 - val_loss: 0.3841 - val_acc: 0.9354 Epoch 225/500 221s 442ms/step - loss: 0.2353 - acc: 0.9802 - val_loss: 0.3856 - val_acc: 0.9374 Epoch 226/500 221s 441ms/step - loss: 0.2389 - acc: 0.9783 - val_loss: 0.3753 - val_acc: 0.9379 Epoch 227/500 220s 441ms/step - loss: 0.2312 - acc: 0.9809 - val_loss: 0.3766 - val_acc: 0.9403 Epoch 228/500 220s 441ms/step - loss: 0.2394 - acc: 0.9772 - val_loss: 0.3825 - val_acc: 0.9374 Epoch 229/500 220s 440ms/step - loss: 0.2333 - acc: 0.9795 - val_loss: 0.3886 - val_acc: 0.9352 Epoch 230/500 220s 441ms/step - loss: 0.2290 - acc: 0.9804 - val_loss: 0.3754 - val_acc: 0.9375 Epoch 231/500 221s 441ms/step - loss: 0.2297 - acc: 0.9804 - val_loss: 0.3832 - val_acc: 0.9370 Epoch 232/500 221s 442ms/step - loss: 0.2333 - acc: 0.9790 - val_loss: 0.3736 - val_acc: 0.9388 Epoch 233/500 221s 442ms/step - loss: 0.2344 - acc: 0.9781 - val_loss: 0.3842 - val_acc: 0.9363 Epoch 234/500 220s 441ms/step - loss: 0.2314 - acc: 0.9797 - val_loss: 0.3821 - val_acc: 0.9355 Epoch 235/500 220s 440ms/step - loss: 0.2304 - acc: 0.9794 - val_loss: 0.3787 - val_acc: 0.9368 Epoch 236/500 221s 442ms/step - loss: 0.2330 - acc: 0.9784 - val_loss: 0.3721 - val_acc: 0.9369 Epoch 237/500 220s 440ms/step - loss: 0.2317 - acc: 0.9788 - val_loss: 0.3697 - val_acc: 0.9387 Epoch 238/500 221s 441ms/step - loss: 0.2286 - acc: 0.9792 - val_loss: 0.3800 - val_acc: 0.9375 Epoch 239/500 220s 441ms/step - loss: 0.2312 - acc: 0.9788 - val_loss: 0.3691 - val_acc: 0.9399 Epoch 240/500 220s 441ms/step - loss: 0.2300 - acc: 0.9790 - val_loss: 0.3751 - val_acc: 0.9399 Epoch 241/500 220s 441ms/step - loss: 0.2266 - acc: 0.9799 - val_loss: 0.3759 - val_acc: 0.9363 Epoch 242/500 220s 441ms/step - loss: 0.2308 - acc: 0.9785 - val_loss: 0.3801 - val_acc: 0.9365 Epoch 243/500 220s 440ms/step - loss: 0.2270 - acc: 0.9796 - val_loss: 0.3688 - val_acc: 0.9390 Epoch 244/500 220s 441ms/step - loss: 0.2259 - acc: 0.9799 - val_loss: 0.3671 - val_acc: 0.9404 Epoch 245/500 220s 441ms/step - loss: 0.2261 - acc: 0.9811 - val_loss: 0.3679 - val_acc: 0.9365 Epoch 246/500 220s 441ms/step - loss: 0.2266 - acc: 0.9792 - val_loss: 0.3778 - val_acc: 0.9353 Epoch 247/500 221s 441ms/step - loss: 0.2276 - acc: 0.9788 - val_loss: 0.3714 - val_acc: 0.9368 Epoch 248/500 221s 441ms/step - loss: 0.2247 - acc: 0.9798 - val_loss: 0.3816 - val_acc: 0.9332 Epoch 249/500 220s 441ms/step - loss: 0.2263 - acc: 0.9793 - val_loss: 0.3611 - val_acc: 0.9409 Epoch 250/500 220s 441ms/step - loss: 0.2289 - acc: 0.9784 - val_loss: 0.3810 - val_acc: 0.9349 Epoch 251/500 220s 440ms/step - loss: 0.2283 - acc: 0.9776 - val_loss: 0.3684 - val_acc: 0.9353 Epoch 252/500 220s 440ms/step - loss: 0.2269 - acc: 0.9789 - val_loss: 0.3777 - val_acc: 0.9352 Epoch 253/500 220s 441ms/step - loss: 0.2251 - acc: 0.9795 - val_loss: 0.3760 - val_acc: 0.9355 Epoch 254/500 220s 441ms/step - loss: 0.2305 - acc: 0.9773 - val_loss: 0.3834 - val_acc: 0.9354 Epoch 255/500 221s 441ms/step - loss: 0.2241 - acc: 0.9790 - val_loss: 0.3709 - val_acc: 0.9379 Epoch 256/500 220s 440ms/step - loss: 0.2255 - acc: 0.9788 - val_loss: 0.3664 - val_acc: 0.9367 Epoch 257/500 220s 441ms/step - loss: 0.2235 - acc: 0.9799 - val_loss: 0.3739 - val_acc: 0.9364 Epoch 258/500 221s 441ms/step - loss: 0.2268 - acc: 0.9788 - val_loss: 0.3718 - val_acc: 0.9358 Epoch 259/500 220s 440ms/step - loss: 0.2211 - acc: 0.9799 - val_loss: 0.3787 - val_acc: 0.9360 Epoch 260/500 221s 441ms/step - loss: 0.2253 - acc: 0.9784 - val_loss: 0.3616 - val_acc: 0.9384 Epoch 261/500 220s 441ms/step - loss: 0.2215 - acc: 0.9803 - val_loss: 0.3872 - val_acc: 0.9318 Epoch 262/500 220s 441ms/step - loss: 0.2277 - acc: 0.9779 - val_loss: 0.3808 - val_acc: 0.9360 Epoch 263/500 221s 442ms/step - loss: 0.2268 - acc: 0.9779 - val_loss: 0.3859 - val_acc: 0.9343 Epoch 264/500 220s 441ms/step - loss: 0.2246 - acc: 0.9791 - val_loss: 0.3848 - val_acc: 0.9330 Epoch 265/500 221s 441ms/step - loss: 0.2246 - acc: 0.9783 - val_loss: 0.3800 - val_acc: 0.9354 Epoch 266/500 220s 441ms/step - loss: 0.2260 - acc: 0.9786 - val_loss: 0.3780 - val_acc: 0.9337 Epoch 267/500 221s 441ms/step - loss: 0.2216 - acc: 0.9804 - val_loss: 0.3744 - val_acc: 0.9373 Epoch 268/500 221s 441ms/step - loss: 0.2208 - acc: 0.9807 - val_loss: 0.3647 - val_acc: 0.9394 Epoch 269/500 221s 441ms/step - loss: 0.2247 - acc: 0.9789 - val_loss: 0.3728 - val_acc: 0.9348 Epoch 270/500 221s 441ms/step - loss: 0.2190 - acc: 0.9804 - val_loss: 0.3703 - val_acc: 0.9366 Epoch 271/500 221s 441ms/step - loss: 0.2213 - acc: 0.9798 - val_loss: 0.3617 - val_acc: 0.9370 Epoch 272/500 221s 441ms/step - loss: 0.2255 - acc: 0.9776 - val_loss: 0.3695 - val_acc: 0.9377 Epoch 273/500 220s 441ms/step - loss: 0.2245 - acc: 0.9781 - val_loss: 0.3775 - val_acc: 0.9349 Epoch 274/500 221s 441ms/step - loss: 0.2225 - acc: 0.9785 - val_loss: 0.3806 - val_acc: 0.9345 Epoch 275/500 221s 441ms/step - loss: 0.2229 - acc: 0.9794 - val_loss: 0.3718 - val_acc: 0.9373 Epoch 276/500 221s 441ms/step - loss: 0.2195 - acc: 0.9806 - val_loss: 0.3849 - val_acc: 0.9339 Epoch 277/500 221s 441ms/step - loss: 0.2204 - acc: 0.9796 - val_loss: 0.3656 - val_acc: 0.9390 Epoch 278/500 221s 441ms/step - loss: 0.2195 - acc: 0.9800 - val_loss: 0.3760 - val_acc: 0.9374 Epoch 279/500 220s 441ms/step - loss: 0.2240 - acc: 0.9790 - val_loss: 0.3694 - val_acc: 0.9344 Epoch 280/500 220s 441ms/step - loss: 0.2203 - acc: 0.9800 - val_loss: 0.3602 - val_acc: 0.9386 Epoch 281/500 221s 441ms/step - loss: 0.2201 - acc: 0.9801 - val_loss: 0.3794 - val_acc: 0.9354 Epoch 282/500 221s 441ms/step - loss: 0.2208 - acc: 0.9802 - val_loss: 0.3660 - val_acc: 0.9377 Epoch 283/500 220s 441ms/step - loss: 0.2164 - acc: 0.9808 - val_loss: 0.3827 - val_acc: 0.9327 Epoch 284/500 221s 441ms/step - loss: 0.2227 - acc: 0.9785 - val_loss: 0.3633 - val_acc: 0.9401 Epoch 285/500 221s 441ms/step - loss: 0.2184 - acc: 0.9808 - val_loss: 0.3862 - val_acc: 0.9309 Epoch 286/500 221s 441ms/step - loss: 0.2159 - acc: 0.9814 - val_loss: 0.3762 - val_acc: 0.9375 Epoch 287/500 221s 442ms/step - loss: 0.2238 - acc: 0.9775 - val_loss: 0.3692 - val_acc: 0.9336 Epoch 288/500 221s 441ms/step - loss: 0.2228 - acc: 0.9786 - val_loss: 0.3746 - val_acc: 0.9354 Epoch 289/500 220s 441ms/step - loss: 0.2197 - acc: 0.9805 - val_loss: 0.3581 - val_acc: 0.9392 Epoch 290/500 222s 444ms/step - loss: 0.2174 - acc: 0.9806 - val_loss: 0.3626 - val_acc: 0.9376 Epoch 291/500 221s 441ms/step - loss: 0.2201 - acc: 0.9796 - val_loss: 0.3834 - val_acc: 0.9323 Epoch 292/500 221s 442ms/step - loss: 0.2217 - acc: 0.9788 - val_loss: 0.3770 - val_acc: 0.9356 Epoch 293/500 221s 442ms/step - loss: 0.2214 - acc: 0.9791 - val_loss: 0.3685 - val_acc: 0.9359 Epoch 294/500 221s 442ms/step - loss: 0.2186 - acc: 0.9795 - val_loss: 0.3708 - val_acc: 0.9375 Epoch 295/500 221s 442ms/step - loss: 0.2191 - acc: 0.9798 - val_loss: 0.3763 - val_acc: 0.9367 Epoch 296/500 221s 442ms/step - loss: 0.2200 - acc: 0.9803 - val_loss: 0.3730 - val_acc: 0.9362 Epoch 297/500 221s 442ms/step - loss: 0.2207 - acc: 0.9795 - val_loss: 0.3731 - val_acc: 0.9350 Epoch 298/500 221s 441ms/step - loss: 0.2197 - acc: 0.9793 - val_loss: 0.3533 - val_acc: 0.9387 Epoch 299/500 221s 441ms/step - loss: 0.2201 - acc: 0.9797 - val_loss: 0.3747 - val_acc: 0.9365 Epoch 300/500 221s 441ms/step - loss: 0.2160 - acc: 0.9809 - val_loss: 0.3678 - val_acc: 0.9386 Epoch 301/500 lr changed to 0.0009999999776482583 221s 442ms/step - loss: 0.2016 - acc: 0.9862 - val_loss: 0.3429 - val_acc: 0.9460 Epoch 302/500 221s 442ms/step - loss: 0.1867 - acc: 0.9912 - val_loss: 0.3401 - val_acc: 0.9479 Epoch 303/500 221s 442ms/step - loss: 0.1819 - acc: 0.9931 - val_loss: 0.3386 - val_acc: 0.9472 Epoch 304/500 221s 441ms/step - loss: 0.1794 - acc: 0.9943 - val_loss: 0.3365 - val_acc: 0.9486 Epoch 305/500 221s 442ms/step - loss: 0.1787 - acc: 0.9938 - val_loss: 0.3357 - val_acc: 0.9490 Epoch 306/500 221s 442ms/step - loss: 0.1760 - acc: 0.9951 - val_loss: 0.3340 - val_acc: 0.9482 Epoch 307/500 221s 442ms/step - loss: 0.1769 - acc: 0.9943 - val_loss: 0.3335 - val_acc: 0.9489 Epoch 308/500 221s 442ms/step - loss: 0.1746 - acc: 0.9952 - val_loss: 0.3342 - val_acc: 0.9501 Epoch 309/500 220s 441ms/step - loss: 0.1731 - acc: 0.9955 - val_loss: 0.3358 - val_acc: 0.9494 Epoch 310/500 221s 441ms/step - loss: 0.1727 - acc: 0.9957 - val_loss: 0.3339 - val_acc: 0.9501 Epoch 311/500 221s 441ms/step - loss: 0.1720 - acc: 0.9958 - val_loss: 0.3305 - val_acc: 0.9511 Epoch 312/500 221s 442ms/step - loss: 0.1715 - acc: 0.9960 - val_loss: 0.3325 - val_acc: 0.9510 Epoch 313/500 221s 441ms/step - loss: 0.1713 - acc: 0.9956 - val_loss: 0.3348 - val_acc: 0.9495 Epoch 314/500 221s 442ms/step - loss: 0.1697 - acc: 0.9963 - val_loss: 0.3338 - val_acc: 0.9500 Epoch 315/500 221s 441ms/step - loss: 0.1693 - acc: 0.9965 - val_loss: 0.3344 - val_acc: 0.9500 Epoch 316/500 221s 442ms/step - loss: 0.1687 - acc: 0.9960 - val_loss: 0.3332 - val_acc: 0.9507 Epoch 317/500 221s 442ms/step - loss: 0.1673 - acc: 0.9967 - val_loss: 0.3317 - val_acc: 0.9504 Epoch 318/500 221s 442ms/step - loss: 0.1678 - acc: 0.9965 - val_loss: 0.3321 - val_acc: 0.9502 Epoch 319/500 221s 442ms/step - loss: 0.1668 - acc: 0.9968 - val_loss: 0.3320 - val_acc: 0.9495 Epoch 320/500 221s 442ms/step - loss: 0.1671 - acc: 0.9965 - val_loss: 0.3326 - val_acc: 0.9493 Epoch 321/500 221s 442ms/step - loss: 0.1651 - acc: 0.9973 - val_loss: 0.3311 - val_acc: 0.9510 Epoch 322/500 221s 442ms/step - loss: 0.1659 - acc: 0.9967 - val_loss: 0.3320 - val_acc: 0.9498 Epoch 323/500 221s 441ms/step - loss: 0.1659 - acc: 0.9965 - val_loss: 0.3319 - val_acc: 0.9506 Epoch 324/500 221s 441ms/step - loss: 0.1648 - acc: 0.9968 - val_loss: 0.3337 - val_acc: 0.9505 Epoch 325/500 221s 442ms/step - loss: 0.1645 - acc: 0.9967 - val_loss: 0.3342 - val_acc: 0.9495 Epoch 326/500 221s 442ms/step - loss: 0.1640 - acc: 0.9971 - val_loss: 0.3324 - val_acc: 0.9495 Epoch 327/500 221s 442ms/step - loss: 0.1630 - acc: 0.9972 - val_loss: 0.3289 - val_acc: 0.9507 Epoch 328/500 221s 442ms/step - loss: 0.1630 - acc: 0.9972 - val_loss: 0.3306 - val_acc: 0.9512 Epoch 329/500 221s 442ms/step - loss: 0.1636 - acc: 0.9967 - val_loss: 0.3330 - val_acc: 0.9507 Epoch 330/500 221s 442ms/step - loss: 0.1622 - acc: 0.9973 - val_loss: 0.3326 - val_acc: 0.9501 Epoch 331/500 221s 441ms/step - loss: 0.1612 - acc: 0.9975 - val_loss: 0.3305 - val_acc: 0.9514 Epoch 332/500 221s 441ms/step - loss: 0.1600 - acc: 0.9979 - val_loss: 0.3299 - val_acc: 0.9517 Epoch 333/500 221s 441ms/step - loss: 0.1597 - acc: 0.9980 - val_loss: 0.3313 - val_acc: 0.9511 Epoch 334/500 221s 441ms/step - loss: 0.1607 - acc: 0.9972 - val_loss: 0.3278 - val_acc: 0.9517 Epoch 335/500 221s 443ms/step - loss: 0.1607 - acc: 0.9974 - val_loss: 0.3277 - val_acc: 0.9527 Epoch 336/500 221s 441ms/step - loss: 0.1588 - acc: 0.9980 - val_loss: 0.3276 - val_acc: 0.9527 Epoch 337/500 221s 442ms/step - loss: 0.1595 - acc: 0.9973 - val_loss: 0.3257 - val_acc: 0.9520 Epoch 338/500 221s 441ms/step - loss: 0.1585 - acc: 0.9976 - val_loss: 0.3274 - val_acc: 0.9521 Epoch 339/500 221s 442ms/step - loss: 0.1594 - acc: 0.9974 - val_loss: 0.3298 - val_acc: 0.9524 Epoch 340/500 221s 441ms/step - loss: 0.1582 - acc: 0.9977 - val_loss: 0.3282 - val_acc: 0.9530 Epoch 341/500 221s 441ms/step - loss: 0.1590 - acc: 0.9972 - val_loss: 0.3273 - val_acc: 0.9527 Epoch 342/500 221s 441ms/step - loss: 0.1575 - acc: 0.9977 - val_loss: 0.3262 - val_acc: 0.9518 Epoch 343/500 221s 442ms/step - loss: 0.1570 - acc: 0.9979 - val_loss: 0.3263 - val_acc: 0.9515 Epoch 344/500 221s 442ms/step - loss: 0.1574 - acc: 0.9977 - val_loss: 0.3259 - val_acc: 0.9511 Epoch 345/500 221s 442ms/step - loss: 0.1568 - acc: 0.9979 - val_loss: 0.3262 - val_acc: 0.9518 Epoch 346/500 221s 442ms/step - loss: 0.1565 - acc: 0.9978 - val_loss: 0.3262 - val_acc: 0.9531 Epoch 347/500 221s 442ms/step - loss: 0.1558 - acc: 0.9979 - val_loss: 0.3248 - val_acc: 0.9538 Epoch 348/500 221s 442ms/step - loss: 0.1554 - acc: 0.9978 - val_loss: 0.3223 - val_acc: 0.9532 Epoch 349/500 221s 442ms/step - loss: 0.1554 - acc: 0.9977 - val_loss: 0.3242 - val_acc: 0.9528 Epoch 350/500 221s 442ms/step - loss: 0.1551 - acc: 0.9977 - val_loss: 0.3255 - val_acc: 0.9527 Epoch 351/500 221s 442ms/step - loss: 0.1546 - acc: 0.9977 - val_loss: 0.3277 - val_acc: 0.9517 Epoch 352/500 221s 442ms/step - loss: 0.1556 - acc: 0.9975 - val_loss: 0.3230 - val_acc: 0.9514 Epoch 353/500 221s 442ms/step - loss: 0.1546 - acc: 0.9975 - val_loss: 0.3240 - val_acc: 0.9526 Epoch 354/500 221s 442ms/step - loss: 0.1535 - acc: 0.9980 - val_loss: 0.3257 - val_acc: 0.9524 Epoch 355/500 221s 442ms/step - loss: 0.1529 - acc: 0.9980 - val_loss: 0.3260 - val_acc: 0.9530 Epoch 356/500 221s 442ms/step - loss: 0.1540 - acc: 0.9978 - val_loss: 0.3261 - val_acc: 0.9522 Epoch 357/500 221s 442ms/step - loss: 0.1534 - acc: 0.9976 - val_loss: 0.3227 - val_acc: 0.9527 Epoch 358/500 221s 441ms/step - loss: 0.1528 - acc: 0.9980 - val_loss: 0.3226 - val_acc: 0.9525 Epoch 359/500 221s 442ms/step - loss: 0.1526 - acc: 0.9978 - val_loss: 0.3240 - val_acc: 0.9527 Epoch 360/500 221s 442ms/step - loss: 0.1520 - acc: 0.9981 - val_loss: 0.3234 - val_acc: 0.9518 Epoch 361/500 221s 442ms/step - loss: 0.1516 - acc: 0.9980 - val_loss: 0.3245 - val_acc: 0.9508 Epoch 362/500 221s 442ms/step - loss: 0.1522 - acc: 0.9976 - val_loss: 0.3222 - val_acc: 0.9520 Epoch 363/500 221s 441ms/step - loss: 0.1509 - acc: 0.9981 - val_loss: 0.3221 - val_acc: 0.9523 Epoch 364/500 221s 442ms/step - loss: 0.1508 - acc: 0.9980 - val_loss: 0.3249 - val_acc: 0.9522 Epoch 365/500 221s 442ms/step - loss: 0.1503 - acc: 0.9981 - val_loss: 0.3243 - val_acc: 0.9503 Epoch 366/500 221s 441ms/step - loss: 0.1509 - acc: 0.9981 - val_loss: 0.3249 - val_acc: 0.9502 Epoch 367/500 221s 442ms/step - loss: 0.1502 - acc: 0.9979 - val_loss: 0.3246 - val_acc: 0.9506 Epoch 368/500 221s 442ms/step - loss: 0.1499 - acc: 0.9981 - val_loss: 0.3225 - val_acc: 0.9523 Epoch 369/500 221s 441ms/step - loss: 0.1505 - acc: 0.9978 - val_loss: 0.3229 - val_acc: 0.9526 Epoch 370/500 221s 442ms/step - loss: 0.1497 - acc: 0.9980 - val_loss: 0.3216 - val_acc: 0.9522 Epoch 371/500 222s 443ms/step - loss: 0.1492 - acc: 0.9979 - val_loss: 0.3235 - val_acc: 0.9511 Epoch 372/500 224s 448ms/step - loss: 0.1495 - acc: 0.9980 - val_loss: 0.3225 - val_acc: 0.9527 Epoch 373/500 221s 441ms/step - loss: 0.1483 - acc: 0.9982 - val_loss: 0.3261 - val_acc: 0.9520 Epoch 374/500 221s 442ms/step - loss: 0.1478 - acc: 0.9984 - val_loss: 0.3250 - val_acc: 0.9522 Epoch 375/500 221s 442ms/step - loss: 0.1482 - acc: 0.9979 - val_loss: 0.3239 - val_acc: 0.9524 Epoch 376/500 221s 442ms/step - loss: 0.1477 - acc: 0.9981 - val_loss: 0.3247 - val_acc: 0.9528 Epoch 377/500 221s 442ms/step - loss: 0.1483 - acc: 0.9979 - val_loss: 0.3271 - val_acc: 0.9512 Epoch 378/500 221s 442ms/step - loss: 0.1471 - acc: 0.9980 - val_loss: 0.3214 - val_acc: 0.9533 Epoch 379/500 221s 442ms/step - loss: 0.1471 - acc: 0.9979 - val_loss: 0.3224 - val_acc: 0.9527 Epoch 380/500 222s 443ms/step - loss: 0.1468 - acc: 0.9982 - val_loss: 0.3211 - val_acc: 0.9537 Epoch 381/500 221s 442ms/step - loss: 0.1470 - acc: 0.9980 - val_loss: 0.3190 - val_acc: 0.9534 Epoch 382/500 221s 442ms/step - loss: 0.1458 - acc: 0.9983 - val_loss: 0.3221 - val_acc: 0.9538 Epoch 383/500 221s 442ms/step - loss: 0.1459 - acc: 0.9980 - val_loss: 0.3199 - val_acc: 0.9538 Epoch 384/500 221s 442ms/step - loss: 0.1461 - acc: 0.9980 - val_loss: 0.3201 - val_acc: 0.9527 Epoch 385/500 221s 442ms/step - loss: 0.1449 - acc: 0.9984 - val_loss: 0.3206 - val_acc: 0.9523 Epoch 386/500 221s 442ms/step - loss: 0.1465 - acc: 0.9975 - val_loss: 0.3199 - val_acc: 0.9534 Epoch 387/500 221s 441ms/step - loss: 0.1451 - acc: 0.9980 - val_loss: 0.3223 - val_acc: 0.9526 Epoch 388/500 221s 442ms/step - loss: 0.1437 - acc: 0.9985 - val_loss: 0.3214 - val_acc: 0.9523 Epoch 389/500 221s 442ms/step - loss: 0.1439 - acc: 0.9981 - val_loss: 0.3229 - val_acc: 0.9525 Epoch 390/500 222s 443ms/step - loss: 0.1444 - acc: 0.9982 - val_loss: 0.3219 - val_acc: 0.9533 Epoch 391/500 221s 442ms/step - loss: 0.1437 - acc: 0.9982 - val_loss: 0.3200 - val_acc: 0.9524 Epoch 392/500 221s 442ms/step - loss: 0.1433 - acc: 0.9982 - val_loss: 0.3212 - val_acc: 0.9524 Epoch 393/500 221s 442ms/step - loss: 0.1428 - acc: 0.9983 - val_loss: 0.3212 - val_acc: 0.9517 Epoch 394/500 221s 442ms/step - loss: 0.1426 - acc: 0.9983 - val_loss: 0.3209 - val_acc: 0.9521 Epoch 395/500 221s 442ms/step - loss: 0.1425 - acc: 0.9986 - val_loss: 0.3197 - val_acc: 0.9522 Epoch 396/500 221s 442ms/step - loss: 0.1417 - acc: 0.9984 - val_loss: 0.3222 - val_acc: 0.9524 Epoch 397/500 221s 443ms/step - loss: 0.1422 - acc: 0.9982 - val_loss: 0.3201 - val_acc: 0.9524 Epoch 398/500 222s 444ms/step - loss: 0.1424 - acc: 0.9981 - val_loss: 0.3203 - val_acc: 0.9524 Epoch 399/500 221s 442ms/step - loss: 0.1415 - acc: 0.9984 - val_loss: 0.3193 - val_acc: 0.9529 Epoch 400/500 220s 440ms/step - loss: 0.1414 - acc: 0.9983 - val_loss: 0.3211 - val_acc: 0.9527 Epoch 401/500 220s 441ms/step - loss: 0.1407 - acc: 0.9984 - val_loss: 0.3184 - val_acc: 0.9534 Epoch 402/500 221s 442ms/step - loss: 0.1406 - acc: 0.9982 - val_loss: 0.3190 - val_acc: 0.9542 Epoch 403/500 221s 441ms/step - loss: 0.1410 - acc: 0.9983 - val_loss: 0.3228 - val_acc: 0.9530 Epoch 404/500 221s 442ms/step - loss: 0.1403 - acc: 0.9983 - val_loss: 0.3210 - val_acc: 0.9529 Epoch 405/500 221s 442ms/step - loss: 0.1404 - acc: 0.9982 - val_loss: 0.3217 - val_acc: 0.9527 Epoch 406/500 221s 442ms/step - loss: 0.1403 - acc: 0.9983 - val_loss: 0.3186 - val_acc: 0.9525 Epoch 407/500 221s 443ms/step - loss: 0.1408 - acc: 0.9978 - val_loss: 0.3213 - val_acc: 0.9535 Epoch 408/500 221s 443ms/step - loss: 0.1395 - acc: 0.9985 - val_loss: 0.3209 - val_acc: 0.9522 Epoch 409/500 221s 442ms/step - loss: 0.1387 - acc: 0.9987 - val_loss: 0.3221 - val_acc: 0.9511 Epoch 410/500 221s 442ms/step - loss: 0.1384 - acc: 0.9986 - val_loss: 0.3199 - val_acc: 0.9512 Epoch 411/500 221s 442ms/step - loss: 0.1386 - acc: 0.9984 - val_loss: 0.3176 - val_acc: 0.9529 Epoch 412/500 221s 441ms/step - loss: 0.1393 - acc: 0.9982 - val_loss: 0.3221 - val_acc: 0.9518 Epoch 413/500 221s 441ms/step - loss: 0.1383 - acc: 0.9986 - val_loss: 0.3208 - val_acc: 0.9528 Epoch 414/500 221s 442ms/step - loss: 0.1383 - acc: 0.9985 - val_loss: 0.3186 - val_acc: 0.9530 Epoch 415/500 222s 443ms/step - loss: 0.1383 - acc: 0.9983 - val_loss: 0.3219 - val_acc: 0.9522 Epoch 416/500 221s 442ms/step - loss: 0.1377 - acc: 0.9984 - val_loss: 0.3231 - val_acc: 0.9521 Epoch 417/500 221s 442ms/step - loss: 0.1375 - acc: 0.9983 - val_loss: 0.3222 - val_acc: 0.9524 Epoch 418/500 221s 442ms/step - loss: 0.1376 - acc: 0.9983 - val_loss: 0.3229 - val_acc: 0.9523 Epoch 419/500 221s 442ms/step - loss: 0.1376 - acc: 0.9981 - val_loss: 0.3210 - val_acc: 0.9520 Epoch 420/500 221s 442ms/step - loss: 0.1362 - acc: 0.9985 - val_loss: 0.3207 - val_acc: 0.9523 Epoch 421/500 221s 442ms/step - loss: 0.1368 - acc: 0.9983 - val_loss: 0.3181 - val_acc: 0.9524 Epoch 422/500 221s 443ms/step - loss: 0.1359 - acc: 0.9986 - val_loss: 0.3173 - val_acc: 0.9529 Epoch 423/500 221s 442ms/step - loss: 0.1358 - acc: 0.9985 - val_loss: 0.3180 - val_acc: 0.9534 Epoch 424/500 221s 442ms/step - loss: 0.1356 - acc: 0.9986 - val_loss: 0.3145 - val_acc: 0.9527 Epoch 425/500 221s 442ms/step - loss: 0.1356 - acc: 0.9984 - val_loss: 0.3168 - val_acc: 0.9518 Epoch 426/500 221s 442ms/step - loss: 0.1353 - acc: 0.9986 - val_loss: 0.3137 - val_acc: 0.9528 Epoch 427/500 221s 442ms/step - loss: 0.1352 - acc: 0.9983 - val_loss: 0.3135 - val_acc: 0.9522 Epoch 428/500 221s 442ms/step - loss: 0.1349 - acc: 0.9983 - val_loss: 0.3173 - val_acc: 0.9515 Epoch 429/500 221s 441ms/step - loss: 0.1343 - acc: 0.9986 - val_loss: 0.3213 - val_acc: 0.9507 Epoch 430/500 221s 442ms/step - loss: 0.1352 - acc: 0.9982 - val_loss: 0.3126 - val_acc: 0.9522 Epoch 431/500 221s 442ms/step - loss: 0.1349 - acc: 0.9981 - val_loss: 0.3169 - val_acc: 0.9505 Epoch 432/500 221s 441ms/step - loss: 0.1336 - acc: 0.9985 - val_loss: 0.3198 - val_acc: 0.9501 Epoch 433/500 221s 442ms/step - loss: 0.1343 - acc: 0.9982 - val_loss: 0.3190 - val_acc: 0.9510 Epoch 434/500 221s 441ms/step - loss: 0.1337 - acc: 0.9984 - val_loss: 0.3182 - val_acc: 0.9504 Epoch 435/500 221s 442ms/step - loss: 0.1331 - acc: 0.9984 - val_loss: 0.3177 - val_acc: 0.9507 Epoch 436/500 221s 441ms/step - loss: 0.1328 - acc: 0.9986 - val_loss: 0.3173 - val_acc: 0.9512 Epoch 437/500 221s 442ms/step - loss: 0.1331 - acc: 0.9983 - val_loss: 0.3203 - val_acc: 0.9508 Epoch 438/500 221s 442ms/step - loss: 0.1327 - acc: 0.9983 - val_loss: 0.3148 - val_acc: 0.9518 Epoch 439/500 221s 442ms/step - loss: 0.1330 - acc: 0.9983 - val_loss: 0.3128 - val_acc: 0.9517 Epoch 440/500 221s 441ms/step - loss: 0.1329 - acc: 0.9984 - val_loss: 0.3160 - val_acc: 0.9508 Epoch 441/500 221s 442ms/step - loss: 0.1324 - acc: 0.9984 - val_loss: 0.3167 - val_acc: 0.9507 Epoch 442/500 221s 441ms/step - loss: 0.1318 - acc: 0.9986 - val_loss: 0.3176 - val_acc: 0.9513 Epoch 443/500 221s 441ms/step - loss: 0.1317 - acc: 0.9983 - val_loss: 0.3188 - val_acc: 0.9527 Epoch 444/500 221s 442ms/step - loss: 0.1310 - acc: 0.9986 - val_loss: 0.3166 - val_acc: 0.9513 Epoch 445/500 221s 441ms/step - loss: 0.1315 - acc: 0.9982 - val_loss: 0.3168 - val_acc: 0.9513 Epoch 446/500 221s 441ms/step - loss: 0.1311 - acc: 0.9984 - val_loss: 0.3179 - val_acc: 0.9506 Epoch 447/500 221s 442ms/step - loss: 0.1313 - acc: 0.9984 - val_loss: 0.3192 - val_acc: 0.9504 Epoch 448/500 221s 441ms/step - loss: 0.1306 - acc: 0.9985 - val_loss: 0.3191 - val_acc: 0.9512 Epoch 449/500 221s 442ms/step - loss: 0.1302 - acc: 0.9987 - val_loss: 0.3182 - val_acc: 0.9511 Epoch 450/500 221s 442ms/step - loss: 0.1303 - acc: 0.9984 - val_loss: 0.3147 - val_acc: 0.9518 Epoch 451/500 lr changed to 9.999999310821295e-05 221s 441ms/step - loss: 0.1303 - acc: 0.9983 - val_loss: 0.3143 - val_acc: 0.9514 Epoch 452/500 221s 441ms/step - loss: 0.1305 - acc: 0.9984 - val_loss: 0.3135 - val_acc: 0.9513 Epoch 453/500 221s 442ms/step - loss: 0.1298 - acc: 0.9986 - val_loss: 0.3131 - val_acc: 0.9516 Epoch 454/500 221s 442ms/step - loss: 0.1301 - acc: 0.9984 - val_loss: 0.3129 - val_acc: 0.9519 Epoch 455/500 221s 442ms/step - loss: 0.1293 - acc: 0.9986 - val_loss: 0.3130 - val_acc: 0.9518 Epoch 456/500 221s 442ms/step - loss: 0.1295 - acc: 0.9988 - val_loss: 0.3125 - val_acc: 0.9519 Epoch 457/500 221s 441ms/step - loss: 0.1295 - acc: 0.9986 - val_loss: 0.3125 - val_acc: 0.9525 Epoch 458/500 221s 442ms/step - loss: 0.1289 - acc: 0.9989 - val_loss: 0.3124 - val_acc: 0.9525 Epoch 459/500 221s 441ms/step - loss: 0.1294 - acc: 0.9988 - val_loss: 0.3126 - val_acc: 0.9526 Epoch 460/500 221s 441ms/step - loss: 0.1296 - acc: 0.9987 - val_loss: 0.3129 - val_acc: 0.9524 Epoch 461/500 221s 442ms/step - loss: 0.1301 - acc: 0.9985 - val_loss: 0.3133 - val_acc: 0.9526 Epoch 462/500 221s 441ms/step - loss: 0.1293 - acc: 0.9987 - val_loss: 0.3133 - val_acc: 0.9526 Epoch 463/500 221s 442ms/step - loss: 0.1290 - acc: 0.9988 - val_loss: 0.3130 - val_acc: 0.9527 Epoch 464/500 221s 441ms/step - loss: 0.1298 - acc: 0.9984 - val_loss: 0.3126 - val_acc: 0.9530 Epoch 465/500 221s 441ms/step - loss: 0.1290 - acc: 0.9986 - val_loss: 0.3122 - val_acc: 0.9525 Epoch 466/500 221s 442ms/step - loss: 0.1292 - acc: 0.9986 - val_loss: 0.3121 - val_acc: 0.9526 Epoch 467/500 221s 441ms/step - loss: 0.1286 - acc: 0.9989 - val_loss: 0.3122 - val_acc: 0.9524 Epoch 468/500 221s 442ms/step - loss: 0.1288 - acc: 0.9989 - val_loss: 0.3123 - val_acc: 0.9526 Epoch 469/500 221s 441ms/step - loss: 0.1284 - acc: 0.9989 - val_loss: 0.3130 - val_acc: 0.9522 Epoch 470/500 223s 445ms/step - loss: 0.1284 - acc: 0.9989 - val_loss: 0.3136 - val_acc: 0.9522 Epoch 471/500 221s 442ms/step - loss: 0.1282 - acc: 0.9990 - val_loss: 0.3138 - val_acc: 0.9517 Epoch 472/500 221s 442ms/step - loss: 0.1291 - acc: 0.9988 - val_loss: 0.3133 - val_acc: 0.9523 Epoch 473/500 221s 441ms/step - loss: 0.1296 - acc: 0.9984 - val_loss: 0.3130 - val_acc: 0.9524 Epoch 474/500 221s 441ms/step - loss: 0.1284 - acc: 0.9988 - val_loss: 0.3128 - val_acc: 0.9527 Epoch 475/500 221s 441ms/step - loss: 0.1283 - acc: 0.9989 - val_loss: 0.3126 - val_acc: 0.9523 Epoch 476/500 221s 442ms/step - loss: 0.1290 - acc: 0.9987 - val_loss: 0.3125 - val_acc: 0.9524 Epoch 477/500 221s 442ms/step - loss: 0.1287 - acc: 0.9988 - val_loss: 0.3121 - val_acc: 0.9521 Epoch 478/500 221s 442ms/step - loss: 0.1291 - acc: 0.9986 - val_loss: 0.3123 - val_acc: 0.9521 Epoch 479/500 221s 442ms/step - loss: 0.1292 - acc: 0.9986 - val_loss: 0.3124 - val_acc: 0.9522 Epoch 480/500 221s 442ms/step - loss: 0.1291 - acc: 0.9986 - val_loss: 0.3123 - val_acc: 0.9519 Epoch 481/500 221s 442ms/step - loss: 0.1282 - acc: 0.9989 - val_loss: 0.3125 - val_acc: 0.9521 Epoch 482/500 221s 442ms/step - loss: 0.1291 - acc: 0.9988 - val_loss: 0.3125 - val_acc: 0.9522 Epoch 483/500 221s 442ms/step - loss: 0.1286 - acc: 0.9988 - val_loss: 0.3125 - val_acc: 0.9516 Epoch 484/500 220s 441ms/step - loss: 0.1280 - acc: 0.9991 - val_loss: 0.3123 - val_acc: 0.9518 Epoch 485/500 220s 441ms/step - loss: 0.1281 - acc: 0.9989 - val_loss: 0.3128 - val_acc: 0.9519 Epoch 486/500 221s 441ms/step - loss: 0.1281 - acc: 0.9990 - val_loss: 0.3127 - val_acc: 0.9520 Epoch 487/500 221s 441ms/step - loss: 0.1282 - acc: 0.9990 - val_loss: 0.3127 - val_acc: 0.9520 Epoch 488/500 221s 441ms/step - loss: 0.1283 - acc: 0.9988 - val_loss: 0.3129 - val_acc: 0.9520 Epoch 489/500 221s 442ms/step - loss: 0.1282 - acc: 0.9988 - val_loss: 0.3131 - val_acc: 0.9521 Epoch 490/500 221s 441ms/step - loss: 0.1283 - acc: 0.9987 - val_loss: 0.3131 - val_acc: 0.9522 Epoch 491/500 221s 441ms/step - loss: 0.1280 - acc: 0.9990 - val_loss: 0.3133 - val_acc: 0.9526 Epoch 492/500 221s 442ms/step - loss: 0.1281 - acc: 0.9989 - val_loss: 0.3132 - val_acc: 0.9524 Epoch 493/500 221s 441ms/step - loss: 0.1282 - acc: 0.9988 - val_loss: 0.3126 - val_acc: 0.9527 Epoch 494/500 221s 441ms/step - loss: 0.1281 - acc: 0.9989 - val_loss: 0.3125 - val_acc: 0.9522 Epoch 495/500 221s 441ms/step - loss: 0.1278 - acc: 0.9991 - val_loss: 0.3118 - val_acc: 0.9524 Epoch 496/500 221s 441ms/step - loss: 0.1280 - acc: 0.9990 - val_loss: 0.3118 - val_acc: 0.9522 Epoch 497/500 221s 442ms/step - loss: 0.1279 - acc: 0.9988 - val_loss: 0.3118 - val_acc: 0.9527 Epoch 498/500 222s 443ms/step - loss: 0.1275 - acc: 0.9991 - val_loss: 0.3113 - val_acc: 0.9523 Epoch 499/500 221s 443ms/step - loss: 0.1275 - acc: 0.9990 - val_loss: 0.3115 - val_acc: 0.9524 Epoch 500/500 221s 443ms/step - loss: 0.1278 - acc: 0.9989 - val_loss: 0.3119 - val_acc: 0.9525 Train loss: 0.12599779877066614 Train accuracy: 0.9992800006866455 Test loss: 0.3119203564524651 Test accuracy: 0.9525000017881393
相較於調參記錄21的95.12%,只提高了0.13%。
Minghang Zhao, Shisheng Zhong, Xuyun Fu, Baoping Tang, Shaojiang Dong, Michael Pecht, Deep Residual Networks with Adaptively Parametric Rectifier Linear Units for Fault Diagnosis, IEEE Transactions on Industrial Electronics, 2020, DOI: 10.1109/TIE.2020.2972458
https://ieeexplore.ieee.org/document/8998530
來自 “ ITPUB部落格 ” ,連結:http://blog.itpub.net/69972329/viewspace-2692867/,如需轉載,請註明出處,否則將追究法律責任。
相關文章
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄26)Cifar10~95.92%函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄18)Cifar10~94.28%函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄19)Cifar10~93.96%函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄23)Cifar10~95.47%函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄24)Cifar10~95.80%函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄20)Cifar10~94.17%函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄1)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄2)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄3)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄4)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄5)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄6)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄7)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄8)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄9)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄10)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄11)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄12)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄13)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄14)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄15)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄16)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄17)函式
- 深度殘差網路+自適應引數化ReLU啟用函式(調參記錄21)Cifar10~95.12%函式
- 注意力機制下的啟用函式:自適應引數化ReLU函式
- 深度殘差網路(ResNet)
- 深度學習之殘差網路深度學習
- 殘差網路再升級之深度殘差收縮網路(附Keras程式碼)Keras
- 深度殘差收縮網路:(三)網路結構
- 深度學習故障診斷——深度殘差收縮網路深度學習
- 深度殘差收縮網路:(一)背景知識
- 深度殘差收縮網路:(二)整體思路
- 學習筆記16:殘差網路筆記
- 十分鐘弄懂深度殘差收縮網路
- 深度殘差收縮網路:(五)實驗驗證
- 深度殘差收縮網路:(六)程式碼實現
- 從ReLU到GELU,一文概覽神經網路的啟用函式神經網路函式
- PHP函式,引數,可變參函式.PHP函式