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%
自適應引數化ReLU是一種動態ReLU(Dynamic ReLU)啟用函式 ,在2019年5月3日投稿到IEEE Transactions on Industrial Electronics,在2020年1月24日(農曆新年初一)錄用,在 2020年2月13日在IEEE官網釋出預覽版。
本文在調參記錄23的基礎上,增加摺積核的個數,最少是64個,最多是256個,繼續測試深度殘差網路+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啟用函式被放在了殘差模組的第二個卷積層之後,它的基本原理如下:
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() 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(64, 3, padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(inputs) net = residual_block(net, 20, 64, downsample=False) net = residual_block(net, 1, 128, downsample=True) net = residual_block(net, 19, 128, downsample=False) net = residual_block(net, 1, 256, downsample=True) net = residual_block(net, 19, 256, 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 281s 562ms/step - loss: 9.6683 - acc: 0.2858 - val_loss: 8.5491 - val_acc: 0.4224 Epoch 2/500 236s 471ms/step - loss: 7.8652 - acc: 0.4406 - val_loss: 7.0180 - val_acc: 0.5270 Epoch 3/500 235s 471ms/step - loss: 6.5241 - acc: 0.5264 - val_loss: 5.7927 - val_acc: 0.6159 Epoch 4/500 235s 471ms/step - loss: 5.4217 - acc: 0.6013 - val_loss: 4.7898 - val_acc: 0.6878 Epoch 5/500 235s 471ms/step - loss: 4.5434 - acc: 0.6542 - val_loss: 4.0362 - val_acc: 0.7256 Epoch 6/500 235s 470ms/step - loss: 3.8297 - acc: 0.6947 - val_loss: 3.3928 - val_acc: 0.7654 Epoch 7/500 235s 471ms/step - loss: 3.2680 - acc: 0.7257 - val_loss: 2.8972 - val_acc: 0.7805 Epoch 8/500 236s 471ms/step - loss: 2.8023 - acc: 0.7493 - val_loss: 2.4718 - val_acc: 0.8117 Epoch 9/500 235s 471ms/step - loss: 2.4351 - acc: 0.7652 - val_loss: 2.1518 - val_acc: 0.8216 Epoch 10/500 235s 470ms/step - loss: 2.1298 - acc: 0.7822 - val_loss: 1.8664 - val_acc: 0.8355 Epoch 11/500 235s 470ms/step - loss: 1.8768 - acc: 0.7961 - val_loss: 1.6576 - val_acc: 0.8407 Epoch 12/500 235s 470ms/step - loss: 1.6745 - acc: 0.8071 - val_loss: 1.4888 - val_acc: 0.8456 Epoch 13/500 235s 471ms/step - loss: 1.5155 - acc: 0.8139 - val_loss: 1.3255 - val_acc: 0.8598 Epoch 14/500 235s 471ms/step - loss: 1.3782 - acc: 0.8230 - val_loss: 1.2249 - val_acc: 0.8616 Epoch 15/500 235s 471ms/step - loss: 1.2630 - acc: 0.8293 - val_loss: 1.1236 - val_acc: 0.8655 Epoch 16/500 235s 471ms/step - loss: 1.1829 - acc: 0.8342 - val_loss: 1.0384 - val_acc: 0.8768 Epoch 17/500 235s 470ms/step - loss: 1.1094 - acc: 0.8389 - val_loss: 0.9748 - val_acc: 0.8752 Epoch 18/500 236s 471ms/step - loss: 1.0510 - acc: 0.8448 - val_loss: 0.9660 - val_acc: 0.8697 Epoch 19/500 235s 471ms/step - loss: 1.0037 - acc: 0.8472 - val_loss: 0.9055 - val_acc: 0.8760 Epoch 20/500 235s 471ms/step - loss: 0.9615 - acc: 0.8520 - val_loss: 0.8935 - val_acc: 0.8711 Epoch 21/500 235s 471ms/step - loss: 0.9345 - acc: 0.8545 - val_loss: 0.8621 - val_acc: 0.8743 Epoch 22/500 235s 470ms/step - loss: 0.9044 - acc: 0.8589 - val_loss: 0.8440 - val_acc: 0.8776 Epoch 23/500 235s 470ms/step - loss: 0.8816 - acc: 0.8625 - val_loss: 0.8310 - val_acc: 0.8792 Epoch 24/500 235s 470ms/step - loss: 0.8640 - acc: 0.8659 - val_loss: 0.8157 - val_acc: 0.8820 Epoch 25/500 235s 470ms/step - loss: 0.8446 - acc: 0.8696 - val_loss: 0.7921 - val_acc: 0.8873 Epoch 26/500 235s 470ms/step - loss: 0.8283 - acc: 0.8716 - val_loss: 0.7739 - val_acc: 0.8934 Epoch 27/500 235s 470ms/step - loss: 0.8212 - acc: 0.8720 - val_loss: 0.7726 - val_acc: 0.8885 Epoch 28/500 235s 471ms/step - loss: 0.8089 - acc: 0.8743 - val_loss: 0.7783 - val_acc: 0.8855 Epoch 29/500 235s 470ms/step - loss: 0.7970 - acc: 0.8775 - val_loss: 0.7350 - val_acc: 0.8988 Epoch 30/500 235s 470ms/step - loss: 0.7911 - acc: 0.8792 - val_loss: 0.7695 - val_acc: 0.8860 Epoch 31/500 235s 470ms/step - loss: 0.7846 - acc: 0.8802 - val_loss: 0.7392 - val_acc: 0.8989 Epoch 32/500 235s 471ms/step - loss: 0.7784 - acc: 0.8814 - val_loss: 0.7618 - val_acc: 0.8888 Epoch 33/500 235s 470ms/step - loss: 0.7724 - acc: 0.8842 - val_loss: 0.7547 - val_acc: 0.8937 Epoch 34/500 235s 470ms/step - loss: 0.7680 - acc: 0.8856 - val_loss: 0.7400 - val_acc: 0.8941 Epoch 35/500 235s 470ms/step - loss: 0.7646 - acc: 0.8865 - val_loss: 0.7079 - val_acc: 0.9096 Epoch 36/500 235s 470ms/step - loss: 0.7567 - acc: 0.8889 - val_loss: 0.7297 - val_acc: 0.8991 Epoch 37/500 235s 471ms/step - loss: 0.7518 - acc: 0.8920 - val_loss: 0.7265 - val_acc: 0.9011 Epoch 38/500 235s 470ms/step - loss: 0.7499 - acc: 0.8911 - val_loss: 0.7068 - val_acc: 0.9108 Epoch 39/500 235s 470ms/step - loss: 0.7455 - acc: 0.8927 - val_loss: 0.7524 - val_acc: 0.8939 Epoch 40/500 235s 470ms/step - loss: 0.7451 - acc: 0.8926 - val_loss: 0.7293 - val_acc: 0.9007 Epoch 41/500 235s 471ms/step - loss: 0.7434 - acc: 0.8951 - val_loss: 0.6985 - val_acc: 0.9097 Epoch 42/500 235s 470ms/step - loss: 0.7439 - acc: 0.8933 - val_loss: 0.7252 - val_acc: 0.9018 Epoch 43/500 235s 470ms/step - loss: 0.7433 - acc: 0.8952 - val_loss: 0.7304 - val_acc: 0.9006 Epoch 44/500 235s 470ms/step - loss: 0.7393 - acc: 0.8958 - val_loss: 0.6997 - val_acc: 0.9134 Epoch 45/500 235s 470ms/step - loss: 0.7348 - acc: 0.8992 - val_loss: 0.7287 - val_acc: 0.9035 Epoch 46/500 235s 470ms/step - loss: 0.7373 - acc: 0.8976 - val_loss: 0.7235 - val_acc: 0.9036 Epoch 47/500 235s 470ms/step - loss: 0.7382 - acc: 0.8974 - val_loss: 0.7178 - val_acc: 0.9081 Epoch 48/500 235s 470ms/step - loss: 0.7363 - acc: 0.8975 - val_loss: 0.7247 - val_acc: 0.9044 Epoch 49/500 235s 470ms/step - loss: 0.7306 - acc: 0.9009 - val_loss: 0.7328 - val_acc: 0.9006 Epoch 50/500 235s 470ms/step - loss: 0.7356 - acc: 0.9003 - val_loss: 0.7096 - val_acc: 0.9114 Epoch 51/500 235s 470ms/step - loss: 0.7282 - acc: 0.9029 - val_loss: 0.7156 - val_acc: 0.9076 Epoch 52/500 235s 470ms/step - loss: 0.7286 - acc: 0.9014 - val_loss: 0.7233 - val_acc: 0.9046 Epoch 53/500 235s 470ms/step - loss: 0.7304 - acc: 0.9016 - val_loss: 0.7087 - val_acc: 0.9088 Epoch 54/500 235s 470ms/step - loss: 0.7261 - acc: 0.9030 - val_loss: 0.7202 - val_acc: 0.9085 Epoch 55/500 235s 470ms/step - loss: 0.7257 - acc: 0.9034 - val_loss: 0.7138 - val_acc: 0.9095 Epoch 56/500 235s 470ms/step - loss: 0.7230 - acc: 0.9043 - val_loss: 0.7196 - val_acc: 0.9084 Epoch 57/500 235s 470ms/step - loss: 0.7212 - acc: 0.9048 - val_loss: 0.7094 - val_acc: 0.9098 Epoch 58/500 236s 473ms/step - loss: 0.7247 - acc: 0.9037 - val_loss: 0.7177 - val_acc: 0.9101 Epoch 59/500 236s 473ms/step - loss: 0.7183 - acc: 0.9067 - val_loss: 0.7385 - val_acc: 0.9026 Epoch 60/500 236s 472ms/step - loss: 0.7224 - acc: 0.9044 - val_loss: 0.7005 - val_acc: 0.9120 Epoch 61/500 236s 472ms/step - loss: 0.7185 - acc: 0.9050 - val_loss: 0.7287 - val_acc: 0.9067 Epoch 62/500 236s 472ms/step - loss: 0.7237 - acc: 0.9049 - val_loss: 0.6969 - val_acc: 0.9160 Epoch 63/500 236s 472ms/step - loss: 0.7177 - acc: 0.9074 - val_loss: 0.7044 - val_acc: 0.9117 Epoch 64/500 236s 472ms/step - loss: 0.7140 - acc: 0.9096 - val_loss: 0.7135 - val_acc: 0.9089 Epoch 65/500 236s 472ms/step - loss: 0.7120 - acc: 0.9074 - val_loss: 0.7107 - val_acc: 0.9093 Epoch 66/500 236s 472ms/step - loss: 0.7148 - acc: 0.9074 - val_loss: 0.7084 - val_acc: 0.9090 Epoch 67/500 236s 472ms/step - loss: 0.7156 - acc: 0.9081 - val_loss: 0.7086 - val_acc: 0.9132 Epoch 68/500 236s 472ms/step - loss: 0.7180 - acc: 0.9074 - val_loss: 0.7177 - val_acc: 0.9090 Epoch 69/500 237s 473ms/step - loss: 0.7111 - acc: 0.9094 - val_loss: 0.7278 - val_acc: 0.9047 Epoch 70/500 236s 473ms/step - loss: 0.7138 - acc: 0.9093 - val_loss: 0.7179 - val_acc: 0.9090 Epoch 71/500 237s 473ms/step - loss: 0.7165 - acc: 0.9084 - val_loss: 0.7251 - val_acc: 0.9065 Epoch 72/500 236s 473ms/step - loss: 0.7133 - acc: 0.9109 - val_loss: 0.6957 - val_acc: 0.9160 Epoch 73/500 236s 473ms/step - loss: 0.7129 - acc: 0.9106 - val_loss: 0.7008 - val_acc: 0.9154 Epoch 74/500 237s 473ms/step - loss: 0.7109 - acc: 0.9110 - val_loss: 0.7126 - val_acc: 0.9121 Epoch 75/500 236s 473ms/step - loss: 0.7143 - acc: 0.9105 - val_loss: 0.7286 - val_acc: 0.9061 Epoch 76/500 236s 472ms/step - loss: 0.7091 - acc: 0.9125 - val_loss: 0.7024 - val_acc: 0.9149 Epoch 77/500 236s 472ms/step - loss: 0.7129 - acc: 0.9104 - val_loss: 0.7176 - val_acc: 0.9106 Epoch 78/500 236s 472ms/step - loss: 0.7108 - acc: 0.9118 - val_loss: 0.6977 - val_acc: 0.9191 Epoch 79/500 237s 473ms/step - loss: 0.7088 - acc: 0.9123 - val_loss: 0.7253 - val_acc: 0.9069 Epoch 80/500 237s 473ms/step - loss: 0.7138 - acc: 0.9106 - val_loss: 0.7052 - val_acc: 0.9192 Epoch 81/500 236s 472ms/step - loss: 0.7120 - acc: 0.9130 - val_loss: 0.7205 - val_acc: 0.9113 Epoch 82/500 236s 472ms/step - loss: 0.7099 - acc: 0.9129 - val_loss: 0.7249 - val_acc: 0.9120 Epoch 83/500 236s 472ms/step - loss: 0.7076 - acc: 0.9117 - val_loss: 0.7329 - val_acc: 0.9060 Epoch 84/500 236s 472ms/step - loss: 0.7150 - acc: 0.9105 - val_loss: 0.6931 - val_acc: 0.9205 Epoch 85/500 236s 472ms/step - loss: 0.7062 - acc: 0.9159 - val_loss: 0.7189 - val_acc: 0.9107 Epoch 86/500 236s 472ms/step - loss: 0.7111 - acc: 0.9117 - val_loss: 0.6910 - val_acc: 0.9204 Epoch 87/500 236s 472ms/step - loss: 0.7070 - acc: 0.9138 - val_loss: 0.6921 - val_acc: 0.9201 Epoch 88/500 236s 472ms/step - loss: 0.7050 - acc: 0.9142 - val_loss: 0.6977 - val_acc: 0.9186 Epoch 89/500 236s 472ms/step - loss: 0.7056 - acc: 0.9136 - val_loss: 0.7174 - val_acc: 0.9109 Epoch 90/500 236s 472ms/step - loss: 0.7032 - acc: 0.9154 - val_loss: 0.6996 - val_acc: 0.9184 Epoch 91/500 236s 472ms/step - loss: 0.7060 - acc: 0.9139 - val_loss: 0.7090 - val_acc: 0.9143 Epoch 92/500 236s 472ms/step - loss: 0.7066 - acc: 0.9130 - val_loss: 0.7228 - val_acc: 0.9114 Epoch 93/500 236s 473ms/step - loss: 0.7063 - acc: 0.9155 - val_loss: 0.7039 - val_acc: 0.9216 Epoch 94/500 236s 473ms/step - loss: 0.7072 - acc: 0.9140 - val_loss: 0.7116 - val_acc: 0.9150 Epoch 95/500 236s 472ms/step - loss: 0.7096 - acc: 0.9143 - val_loss: 0.7216 - val_acc: 0.9109 Epoch 96/500 236s 473ms/step - loss: 0.7006 - acc: 0.9161 - val_loss: 0.7143 - val_acc: 0.9141 Epoch 97/500 236s 472ms/step - loss: 0.7069 - acc: 0.9142 - val_loss: 0.6954 - val_acc: 0.9224 Epoch 98/500 236s 473ms/step - loss: 0.7064 - acc: 0.9151 - val_loss: 0.7273 - val_acc: 0.9081 Epoch 99/500 236s 472ms/step - loss: 0.7038 - acc: 0.9161 - val_loss: 0.7274 - val_acc: 0.9132 Epoch 100/500 236s 472ms/step - loss: 0.7064 - acc: 0.9162 - val_loss: 0.7092 - val_acc: 0.9159 Epoch 101/500 236s 472ms/step - loss: 0.7011 - acc: 0.9169 - val_loss: 0.7331 - val_acc: 0.9080 Epoch 102/500 236s 472ms/step - loss: 0.7047 - acc: 0.9159 - val_loss: 0.7092 - val_acc: 0.9185 Epoch 103/500 237s 473ms/step - loss: 0.6994 - acc: 0.9166 - val_loss: 0.7029 - val_acc: 0.9180 Epoch 104/500 237s 473ms/step - loss: 0.6976 - acc: 0.9186 - val_loss: 0.7215 - val_acc: 0.9083 Epoch 105/500 236s 473ms/step - loss: 0.7002 - acc: 0.9159 - val_loss: 0.7191 - val_acc: 0.9111 Epoch 106/500 236s 472ms/step - loss: 0.7019 - acc: 0.9152 - val_loss: 0.7217 - val_acc: 0.9138 Epoch 107/500 236s 472ms/step - loss: 0.7078 - acc: 0.9154 - val_loss: 0.6926 - val_acc: 0.9249 Epoch 108/500 236s 472ms/step - loss: 0.7069 - acc: 0.9154 - val_loss: 0.7048 - val_acc: 0.9214 Epoch 109/500 236s 472ms/step - loss: 0.6975 - acc: 0.9182 - val_loss: 0.7130 - val_acc: 0.9130 Epoch 110/500 236s 472ms/step - loss: 0.7010 - acc: 0.9168 - val_loss: 0.7074 - val_acc: 0.9140 Epoch 111/500 236s 472ms/step - loss: 0.7020 - acc: 0.9175 - val_loss: 0.7142 - val_acc: 0.9161 Epoch 112/500 236s 473ms/step - loss: 0.6991 - acc: 0.9179 - val_loss: 0.7238 - val_acc: 0.9075 Epoch 113/500 236s 473ms/step - loss: 0.7022 - acc: 0.9165 - val_loss: 0.7162 - val_acc: 0.9165 Epoch 114/500 236s 473ms/step - loss: 0.7006 - acc: 0.9177 - val_loss: 0.7261 - val_acc: 0.9125 Epoch 115/500 237s 473ms/step - loss: 0.6987 - acc: 0.9184 - val_loss: 0.7110 - val_acc: 0.9121 Epoch 116/500 236s 472ms/step - loss: 0.6984 - acc: 0.9169 - val_loss: 0.7012 - val_acc: 0.9221 Epoch 117/500 236s 472ms/step - loss: 0.7002 - acc: 0.9162 - val_loss: 0.7278 - val_acc: 0.9113 Epoch 118/500 236s 472ms/step - loss: 0.6998 - acc: 0.9181 - val_loss: 0.7352 - val_acc: 0.9079 Epoch 119/500 236s 473ms/step - loss: 0.6989 - acc: 0.9187 - val_loss: 0.7147 - val_acc: 0.9162 Epoch 120/500 237s 473ms/step - loss: 0.7025 - acc: 0.9175 - val_loss: 0.7014 - val_acc: 0.9195 Epoch 121/500 236s 473ms/step - loss: 0.7003 - acc: 0.9183 - val_loss: 0.6987 - val_acc: 0.9177 Epoch 122/500 236s 472ms/step - loss: 0.6996 - acc: 0.9172 - val_loss: 0.7206 - val_acc: 0.9146 Epoch 123/500 236s 472ms/step - loss: 0.7033 - acc: 0.9174 - val_loss: 0.7128 - val_acc: 0.9187 Epoch 124/500 236s 472ms/step - loss: 0.6957 - acc: 0.9194 - val_loss: 0.7079 - val_acc: 0.9177 Epoch 125/500 236s 472ms/step - loss: 0.7028 - acc: 0.9171 - val_loss: 0.7080 - val_acc: 0.9200 Epoch 126/500 236s 472ms/step - loss: 0.7005 - acc: 0.9167 - val_loss: 0.7362 - val_acc: 0.9096 Epoch 127/500 236s 472ms/step - loss: 0.7044 - acc: 0.9182 - val_loss: 0.7139 - val_acc: 0.9164 Epoch 128/500 236s 472ms/step - loss: 0.7031 - acc: 0.9184 - val_loss: 0.7105 - val_acc: 0.9162 Epoch 129/500 236s 472ms/step - loss: 0.6979 - acc: 0.9194 - val_loss: 0.7255 - val_acc: 0.9160 Epoch 130/500 236s 472ms/step - loss: 0.7016 - acc: 0.9192 - val_loss: 0.7252 - val_acc: 0.9150 Epoch 131/500 236s 472ms/step - loss: 0.6991 - acc: 0.9208 - val_loss: 0.7086 - val_acc: 0.9204 Epoch 132/500 236s 472ms/step - loss: 0.7005 - acc: 0.9195 - val_loss: 0.7169 - val_acc: 0.9163 Epoch 133/500 236s 472ms/step - loss: 0.7006 - acc: 0.9191 - val_loss: 0.7040 - val_acc: 0.9202 Epoch 134/500 236s 472ms/step - loss: 0.7009 - acc: 0.9196 - val_loss: 0.7241 - val_acc: 0.9139 Epoch 135/500 237s 473ms/step - loss: 0.6915 - acc: 0.9211 - val_loss: 0.7318 - val_acc: 0.9088 Epoch 136/500 236s 472ms/step - loss: 0.7054 - acc: 0.9172 - val_loss: 0.7155 - val_acc: 0.9157 Epoch 137/500 236s 473ms/step - loss: 0.7006 - acc: 0.9191 - val_loss: 0.7221 - val_acc: 0.9152 Epoch 138/500 236s 472ms/step - loss: 0.6938 - acc: 0.9209 - val_loss: 0.7026 - val_acc: 0.9228 Epoch 139/500 236s 472ms/step - loss: 0.6994 - acc: 0.9189 - val_loss: 0.7087 - val_acc: 0.9200 Epoch 140/500 236s 472ms/step - loss: 0.6990 - acc: 0.9194 - val_loss: 0.7284 - val_acc: 0.9125 Epoch 141/500 236s 472ms/step - loss: 0.7056 - acc: 0.9174 - val_loss: 0.7021 - val_acc: 0.9224 Epoch 142/500 236s 472ms/step - loss: 0.6948 - acc: 0.9199 - val_loss: 0.7238 - val_acc: 0.9201 Epoch 143/500 236s 472ms/step - loss: 0.6957 - acc: 0.9199 - val_loss: 0.7073 - val_acc: 0.9164 Epoch 144/500 236s 472ms/step - loss: 0.6981 - acc: 0.9206 - val_loss: 0.7212 - val_acc: 0.9148 Epoch 145/500 236s 472ms/step - loss: 0.6976 - acc: 0.9212 - val_loss: 0.7079 - val_acc: 0.9204 Epoch 146/500 236s 473ms/step - loss: 0.6993 - acc: 0.9197 - val_loss: 0.7286 - val_acc: 0.9154 Epoch 147/500 236s 472ms/step - loss: 0.6891 - acc: 0.9234 - val_loss: 0.7160 - val_acc: 0.9175 Epoch 148/500 236s 472ms/step - loss: 0.6978 - acc: 0.9203 - val_loss: 0.7208 - val_acc: 0.9138 Epoch 149/500 236s 472ms/step - loss: 0.6975 - acc: 0.9196 - val_loss: 0.7227 - val_acc: 0.9138 Epoch 150/500 236s 472ms/step - loss: 0.6930 - acc: 0.9217 - val_loss: 0.7080 - val_acc: 0.9171 Epoch 151/500 lr changed to 0.010000000149011612 236s 472ms/step - loss: 0.5908 - acc: 0.9565 - val_loss: 0.6227 - val_acc: 0.9453 Epoch 152/500 236s 472ms/step - loss: 0.5375 - acc: 0.9734 - val_loss: 0.6086 - val_acc: 0.9479 Epoch 153/500 236s 472ms/step - loss: 0.5165 - acc: 0.9768 - val_loss: 0.5996 - val_acc: 0.9481 Epoch 154/500 236s 472ms/step - loss: 0.5005 - acc: 0.9798 - val_loss: 0.5907 - val_acc: 0.9494 Epoch 155/500 236s 472ms/step - loss: 0.4879 - acc: 0.9815 - val_loss: 0.5831 - val_acc: 0.9512 Epoch 156/500 236s 472ms/step - loss: 0.4739 - acc: 0.9842 - val_loss: 0.5746 - val_acc: 0.9508 Epoch 157/500 236s 472ms/step - loss: 0.4628 - acc: 0.9846 - val_loss: 0.5674 - val_acc: 0.9516 Epoch 158/500 236s 472ms/step - loss: 0.4513 - acc: 0.9866 - val_loss: 0.5614 - val_acc: 0.9513 Epoch 159/500 236s 472ms/step - loss: 0.4419 - acc: 0.9882 - val_loss: 0.5624 - val_acc: 0.9510 Epoch 160/500 236s 472ms/step - loss: 0.4331 - acc: 0.9874 - val_loss: 0.5522 - val_acc: 0.9516 Epoch 161/500 236s 472ms/step - loss: 0.4238 - acc: 0.9895 - val_loss: 0.5444 - val_acc: 0.9517 Epoch 162/500 236s 473ms/step - loss: 0.4177 - acc: 0.9885 - val_loss: 0.5420 - val_acc: 0.9517 Epoch 163/500 236s 472ms/step - loss: 0.4088 - acc: 0.9897 - val_loss: 0.5349 - val_acc: 0.9511 Epoch 164/500 236s 472ms/step - loss: 0.4013 - acc: 0.9897 - val_loss: 0.5262 - val_acc: 0.9537 Epoch 165/500 236s 472ms/step - loss: 0.3953 - acc: 0.9895 - val_loss: 0.5197 - val_acc: 0.9520 Epoch 166/500 236s 472ms/step - loss: 0.3872 - acc: 0.9908 - val_loss: 0.5165 - val_acc: 0.9517 Epoch 167/500 236s 472ms/step - loss: 0.3805 - acc: 0.9911 - val_loss: 0.5151 - val_acc: 0.9519 Epoch 168/500 236s 472ms/step - loss: 0.3732 - acc: 0.9910 - val_loss: 0.5118 - val_acc: 0.9506 Epoch 169/500 236s 472ms/step - loss: 0.3677 - acc: 0.9913 - val_loss: 0.5079 - val_acc: 0.9497 Epoch 170/500 236s 472ms/step - loss: 0.3603 - acc: 0.9913 - val_loss: 0.4986 - val_acc: 0.9512 Epoch 171/500 236s 472ms/step - loss: 0.3551 - acc: 0.9920 - val_loss: 0.4933 - val_acc: 0.9517 Epoch 172/500 236s 472ms/step - loss: 0.3506 - acc: 0.9916 - val_loss: 0.4847 - val_acc: 0.9527 Epoch 173/500 236s 473ms/step - loss: 0.3436 - acc: 0.9921 - val_loss: 0.4808 - val_acc: 0.9528 Epoch 174/500 236s 472ms/step - loss: 0.3374 - acc: 0.9922 - val_loss: 0.4800 - val_acc: 0.9528 Epoch 175/500 236s 472ms/step - loss: 0.3324 - acc: 0.9924 - val_loss: 0.4743 - val_acc: 0.9512 Epoch 176/500 236s 472ms/step - loss: 0.3277 - acc: 0.9925 - val_loss: 0.4741 - val_acc: 0.9505 Epoch 177/500 236s 472ms/step - loss: 0.3222 - acc: 0.9929 - val_loss: 0.4663 - val_acc: 0.9503 Epoch 178/500 236s 472ms/step - loss: 0.3179 - acc: 0.9926 - val_loss: 0.4694 - val_acc: 0.9497 Epoch 179/500 236s 472ms/step - loss: 0.3148 - acc: 0.9922 - val_loss: 0.4584 - val_acc: 0.9502 Epoch 180/500 236s 472ms/step - loss: 0.3082 - acc: 0.9938 - val_loss: 0.4647 - val_acc: 0.9484 Epoch 181/500 236s 471ms/step - loss: 0.3024 - acc: 0.9936 - val_loss: 0.4562 - val_acc: 0.9503 Epoch 182/500 236s 472ms/step - loss: 0.2996 - acc: 0.9931 - val_loss: 0.4528 - val_acc: 0.9518 Epoch 183/500 236s 472ms/step - loss: 0.2949 - acc: 0.9935 - val_loss: 0.4555 - val_acc: 0.9503 Epoch 184/500 236s 472ms/step - loss: 0.2916 - acc: 0.9932 - val_loss: 0.4519 - val_acc: 0.9481 Epoch 185/500 236s 472ms/step - loss: 0.2886 - acc: 0.9929 - val_loss: 0.4428 - val_acc: 0.9491 Epoch 186/500 236s 472ms/step - loss: 0.2857 - acc: 0.9929 - val_loss: 0.4383 - val_acc: 0.9495 Epoch 187/500 236s 472ms/step - loss: 0.2816 - acc: 0.9928 - val_loss: 0.4380 - val_acc: 0.9479 Epoch 188/500 236s 472ms/step - loss: 0.2764 - acc: 0.9935 - val_loss: 0.4373 - val_acc: 0.9479 Epoch 189/500 236s 472ms/step - loss: 0.2742 - acc: 0.9930 - val_loss: 0.4204 - val_acc: 0.9510 Epoch 190/500 236s 472ms/step - loss: 0.2697 - acc: 0.9935 - val_loss: 0.4241 - val_acc: 0.9500 Epoch 191/500 236s 472ms/step - loss: 0.2642 - acc: 0.9940 - val_loss: 0.4287 - val_acc: 0.9486 Epoch 192/500 236s 472ms/step - loss: 0.2631 - acc: 0.9932 - val_loss: 0.4230 - val_acc: 0.9478 Epoch 193/500 236s 472ms/step - loss: 0.2614 - acc: 0.9929 - val_loss: 0.4182 - val_acc: 0.9470 Epoch 194/500 236s 472ms/step - loss: 0.2577 - acc: 0.9929 - val_loss: 0.4143 - val_acc: 0.9465 Epoch 195/500 236s 472ms/step - loss: 0.2551 - acc: 0.9933 - val_loss: 0.4124 - val_acc: 0.9488 Epoch 196/500 236s 472ms/step - loss: 0.2522 - acc: 0.9933 - val_loss: 0.4111 - val_acc: 0.9469 Epoch 197/500 236s 472ms/step - loss: 0.2487 - acc: 0.9930 - val_loss: 0.4062 - val_acc: 0.9515 Epoch 198/500 236s 472ms/step - loss: 0.2474 - acc: 0.9926 - val_loss: 0.4036 - val_acc: 0.9505 Epoch 199/500 236s 472ms/step - loss: 0.2419 - acc: 0.9943 - val_loss: 0.3958 - val_acc: 0.9522 Epoch 200/500 236s 472ms/step - loss: 0.2393 - acc: 0.9939 - val_loss: 0.4007 - val_acc: 0.9509 Epoch 201/500 236s 472ms/step - loss: 0.2399 - acc: 0.9927 - val_loss: 0.3961 - val_acc: 0.9475 Epoch 202/500 236s 472ms/step - loss: 0.2385 - acc: 0.9922 - val_loss: 0.3966 - val_acc: 0.9486 Epoch 203/500 236s 471ms/step - loss: 0.2354 - acc: 0.9926 - val_loss: 0.3975 - val_acc: 0.9502 Epoch 204/500 236s 472ms/step - loss: 0.2314 - acc: 0.9935 - val_loss: 0.3895 - val_acc: 0.9510 Epoch 205/500 236s 472ms/step - loss: 0.2286 - acc: 0.9939 - val_loss: 0.3812 - val_acc: 0.9511 Epoch 206/500 236s 471ms/step - loss: 0.2286 - acc: 0.9926 - val_loss: 0.3841 - val_acc: 0.9500 Epoch 207/500 236s 472ms/step - loss: 0.2286 - acc: 0.9919 - val_loss: 0.3856 - val_acc: 0.9480 Epoch 208/500 236s 472ms/step - loss: 0.2249 - acc: 0.9928 - val_loss: 0.3721 - val_acc: 0.9509 Epoch 209/500 236s 471ms/step - loss: 0.2228 - acc: 0.9927 - val_loss: 0.3787 - val_acc: 0.9504 Epoch 210/500 236s 472ms/step - loss: 0.2222 - acc: 0.9925 - val_loss: 0.3797 - val_acc: 0.9483 Epoch 211/500 236s 471ms/step - loss: 0.2223 - acc: 0.9917 - val_loss: 0.3699 - val_acc: 0.9516 Epoch 212/500 236s 472ms/step - loss: 0.2185 - acc: 0.9925 - val_loss: 0.3764 - val_acc: 0.9500 Epoch 213/500 236s 472ms/step - loss: 0.2165 - acc: 0.9924 - val_loss: 0.3842 - val_acc: 0.9466 Epoch 214/500 236s 472ms/step - loss: 0.2151 - acc: 0.9925 - val_loss: 0.3631 - val_acc: 0.9517 Epoch 215/500 236s 471ms/step - loss: 0.2148 - acc: 0.9921 - val_loss: 0.3738 - val_acc: 0.9485 Epoch 216/500 236s 472ms/step - loss: 0.2137 - acc: 0.9920 - val_loss: 0.3658 - val_acc: 0.9489 Epoch 217/500 236s 472ms/step - loss: 0.2141 - acc: 0.9916 - val_loss: 0.3644 - val_acc: 0.9492 Epoch 218/500 236s 472ms/step - loss: 0.2071 - acc: 0.9929 - val_loss: 0.3660 - val_acc: 0.9489 Epoch 219/500 236s 472ms/step - loss: 0.2108 - acc: 0.9912 - val_loss: 0.3556 - val_acc: 0.9510 Epoch 220/500 236s 471ms/step - loss: 0.2068 - acc: 0.9921 - val_loss: 0.3661 - val_acc: 0.9492 Epoch 221/500 236s 472ms/step - loss: 0.2061 - acc: 0.9921 - val_loss: 0.3592 - val_acc: 0.9478 Epoch 222/500 236s 472ms/step - loss: 0.2073 - acc: 0.9908 - val_loss: 0.3573 - val_acc: 0.9485 Epoch 223/500 236s 472ms/step - loss: 0.2066 - acc: 0.9913 - val_loss: 0.3703 - val_acc: 0.9447 Epoch 224/500 236s 472ms/step - loss: 0.2075 - acc: 0.9903 - val_loss: 0.3552 - val_acc: 0.9499 Epoch 225/500 236s 471ms/step - loss: 0.2015 - acc: 0.9922 - val_loss: 0.3700 - val_acc: 0.9470 Epoch 226/500 236s 472ms/step - loss: 0.2033 - acc: 0.9914 - val_loss: 0.3646 - val_acc: 0.9483 Epoch 227/500 236s 472ms/step - loss: 0.2039 - acc: 0.9908 - val_loss: 0.3619 - val_acc: 0.9483 Epoch 228/500 236s 472ms/step - loss: 0.1979 - acc: 0.9927 - val_loss: 0.3593 - val_acc: 0.9490 Epoch 229/500 236s 472ms/step - loss: 0.1987 - acc: 0.9915 - val_loss: 0.3628 - val_acc: 0.9469 Epoch 230/500 236s 472ms/step - loss: 0.1991 - acc: 0.9916 - val_loss: 0.3524 - val_acc: 0.9471 Epoch 231/500 236s 472ms/step - loss: 0.1983 - acc: 0.9913 - val_loss: 0.3551 - val_acc: 0.9484 Epoch 232/500 236s 472ms/step - loss: 0.1953 - acc: 0.9923 - val_loss: 0.3511 - val_acc: 0.9465 Epoch 233/500 236s 472ms/step - loss: 0.1950 - acc: 0.9917 - val_loss: 0.3513 - val_acc: 0.9472 Epoch 234/500 236s 472ms/step - loss: 0.1930 - acc: 0.9920 - val_loss: 0.3515 - val_acc: 0.9501 Epoch 235/500 236s 472ms/step - loss: 0.1956 - acc: 0.9909 - val_loss: 0.3401 - val_acc: 0.9506 Epoch 236/500 236s 472ms/step - loss: 0.1941 - acc: 0.9909 - val_loss: 0.3536 - val_acc: 0.9471 Epoch 237/500 236s 472ms/step - loss: 0.1961 - acc: 0.9904 - val_loss: 0.3740 - val_acc: 0.9419 Epoch 238/500 236s 471ms/step - loss: 0.1975 - acc: 0.9894 - val_loss: 0.3527 - val_acc: 0.9472 Epoch 239/500 236s 472ms/step - loss: 0.1947 - acc: 0.9908 - val_loss: 0.3384 - val_acc: 0.9506 Epoch 240/500 236s 472ms/step - loss: 0.1910 - acc: 0.9915 - val_loss: 0.3403 - val_acc: 0.9489 Epoch 241/500 236s 472ms/step - loss: 0.1920 - acc: 0.9907 - val_loss: 0.3510 - val_acc: 0.9443 Epoch 242/500 236s 472ms/step - loss: 0.1940 - acc: 0.9899 - val_loss: 0.3439 - val_acc: 0.9487 Epoch 243/500 236s 472ms/step - loss: 0.1885 - acc: 0.9918 - val_loss: 0.3400 - val_acc: 0.9496 Epoch 244/500 236s 471ms/step - loss: 0.1901 - acc: 0.9912 - val_loss: 0.3487 - val_acc: 0.9483 Epoch 245/500 236s 471ms/step - loss: 0.1885 - acc: 0.9916 - val_loss: 0.3321 - val_acc: 0.9526 Epoch 246/500 236s 472ms/step - loss: 0.1887 - acc: 0.9912 - val_loss: 0.3563 - val_acc: 0.9456 Epoch 247/500 236s 471ms/step - loss: 0.1912 - acc: 0.9901 - val_loss: 0.3408 - val_acc: 0.9457 Epoch 248/500 236s 472ms/step - loss: 0.1900 - acc: 0.9907 - val_loss: 0.3511 - val_acc: 0.9462 Epoch 249/500 236s 472ms/step - loss: 0.1902 - acc: 0.9898 - val_loss: 0.3484 - val_acc: 0.9470 Epoch 250/500 236s 472ms/step - loss: 0.1854 - acc: 0.9916 - val_loss: 0.3378 - val_acc: 0.9509 Epoch 251/500 236s 471ms/step - loss: 0.1896 - acc: 0.9899 - val_loss: 0.3470 - val_acc: 0.9466 Epoch 252/500 236s 472ms/step - loss: 0.1885 - acc: 0.9906 - val_loss: 0.3437 - val_acc: 0.9500 Epoch 253/500 236s 472ms/step - loss: 0.1855 - acc: 0.9912 - val_loss: 0.3424 - val_acc: 0.9472 Epoch 254/500 236s 471ms/step - loss: 0.1835 - acc: 0.9914 - val_loss: 0.3507 - val_acc: 0.9477 Epoch 255/500 236s 472ms/step - loss: 0.1887 - acc: 0.9900 - val_loss: 0.3447 - val_acc: 0.9457 Epoch 256/500 236s 472ms/step - loss: 0.1895 - acc: 0.9896 - val_loss: 0.3444 - val_acc: 0.9474 Epoch 257/500 236s 472ms/step - loss: 0.1876 - acc: 0.9904 - val_loss: 0.3446 - val_acc: 0.9448 Epoch 258/500 236s 472ms/step - loss: 0.1864 - acc: 0.9901 - val_loss: 0.3412 - val_acc: 0.9471 Epoch 259/500 236s 471ms/step - loss: 0.1845 - acc: 0.9910 - val_loss: 0.3501 - val_acc: 0.9435 Epoch 260/500 236s 471ms/step - loss: 0.1886 - acc: 0.9891 - val_loss: 0.3388 - val_acc: 0.9468 Epoch 261/500 236s 471ms/step - loss: 0.1876 - acc: 0.9893 - val_loss: 0.3583 - val_acc: 0.9458 Epoch 262/500 236s 471ms/step - loss: 0.1873 - acc: 0.9899 - val_loss: 0.3586 - val_acc: 0.9446 Epoch 263/500 236s 471ms/step - loss: 0.1814 - acc: 0.9917 - val_loss: 0.3447 - val_acc: 0.9480 Epoch 264/500 236s 472ms/step - loss: 0.1851 - acc: 0.9908 - val_loss: 0.3413 - val_acc: 0.9464 Epoch 265/500 236s 471ms/step - loss: 0.1869 - acc: 0.9897 - val_loss: 0.3462 - val_acc: 0.9421 Epoch 266/500 236s 472ms/step - loss: 0.1830 - acc: 0.9908 - val_loss: 0.3249 - val_acc: 0.9518 Epoch 267/500 236s 472ms/step - loss: 0.1854 - acc: 0.9905 - val_loss: 0.3291 - val_acc: 0.9479 Epoch 268/500 236s 471ms/step - loss: 0.1838 - acc: 0.9901 - val_loss: 0.3499 - val_acc: 0.9453 Epoch 269/500 236s 471ms/step - loss: 0.1842 - acc: 0.9903 - val_loss: 0.3493 - val_acc: 0.9468 Epoch 270/500 236s 472ms/step - loss: 0.1858 - acc: 0.9897 - val_loss: 0.3450 - val_acc: 0.9461 Epoch 271/500 236s 471ms/step - loss: 0.1822 - acc: 0.9903 - val_loss: 0.3357 - val_acc: 0.9477 Epoch 272/500 236s 471ms/step - loss: 0.1840 - acc: 0.9904 - val_loss: 0.3432 - val_acc: 0.9452 Epoch 273/500 236s 472ms/step - loss: 0.1873 - acc: 0.9894 - val_loss: 0.3369 - val_acc: 0.9472 Epoch 274/500 236s 472ms/step - loss: 0.1826 - acc: 0.9906 - val_loss: 0.3556 - val_acc: 0.9443 Epoch 275/500 236s 472ms/step - loss: 0.1812 - acc: 0.9905 - val_loss: 0.3580 - val_acc: 0.9474 Epoch 276/500 236s 472ms/step - loss: 0.1835 - acc: 0.9900 - val_loss: 0.3549 - val_acc: 0.9447 Epoch 277/500 236s 471ms/step - loss: 0.1808 - acc: 0.9908 - val_loss: 0.3434 - val_acc: 0.9454 Epoch 278/500 236s 472ms/step - loss: 0.1822 - acc: 0.9904 - val_loss: 0.3398 - val_acc: 0.9476 Epoch 279/500 236s 472ms/step - loss: 0.1825 - acc: 0.9906 - val_loss: 0.3292 - val_acc: 0.9497 Epoch 280/500 236s 471ms/step - loss: 0.1839 - acc: 0.9896 - val_loss: 0.3487 - val_acc: 0.9433 Epoch 281/500 236s 472ms/step - loss: 0.1826 - acc: 0.9905 - val_loss: 0.3433 - val_acc: 0.9464 Epoch 282/500 236s 472ms/step - loss: 0.1808 - acc: 0.9910 - val_loss: 0.3261 - val_acc: 0.9469 Epoch 283/500 236s 472ms/step - loss: 0.1816 - acc: 0.9907 - val_loss: 0.3523 - val_acc: 0.9418 Epoch 284/500 236s 472ms/step - loss: 0.1847 - acc: 0.9895 - val_loss: 0.3297 - val_acc: 0.9477 Epoch 285/500 236s 472ms/step - loss: 0.1813 - acc: 0.9909 - val_loss: 0.3337 - val_acc: 0.9472 Epoch 286/500 236s 472ms/step - loss: 0.1797 - acc: 0.9911 - val_loss: 0.3495 - val_acc: 0.9457 Epoch 287/500 236s 471ms/step - loss: 0.1824 - acc: 0.9905 - val_loss: 0.3378 - val_acc: 0.9471 Epoch 288/500 236s 472ms/step - loss: 0.1815 - acc: 0.9904 - val_loss: 0.3371 - val_acc: 0.9479 Epoch 289/500 236s 472ms/step - loss: 0.1795 - acc: 0.9909 - val_loss: 0.3345 - val_acc: 0.9481 Epoch 290/500 236s 472ms/step - loss: 0.1833 - acc: 0.9896 - val_loss: 0.3221 - val_acc: 0.9505 Epoch 291/500 236s 471ms/step - loss: 0.1798 - acc: 0.9911 - val_loss: 0.3425 - val_acc: 0.9464 Epoch 292/500 236s 471ms/step - loss: 0.1826 - acc: 0.9899 - val_loss: 0.3563 - val_acc: 0.9407 Epoch 293/500 236s 471ms/step - loss: 0.1809 - acc: 0.9906 - val_loss: 0.3398 - val_acc: 0.9463 Epoch 294/500 236s 471ms/step - loss: 0.1835 - acc: 0.9897 - val_loss: 0.3427 - val_acc: 0.9478 Epoch 295/500 236s 471ms/step - loss: 0.1792 - acc: 0.9913 - val_loss: 0.3443 - val_acc: 0.9463 Epoch 296/500 236s 471ms/step - loss: 0.1794 - acc: 0.9907 - val_loss: 0.3456 - val_acc: 0.9457 Epoch 297/500 236s 472ms/step - loss: 0.1751 - acc: 0.9922 - val_loss: 0.3362 - val_acc: 0.9457 Epoch 298/500 236s 472ms/step - loss: 0.1821 - acc: 0.9893 - val_loss: 0.3465 - val_acc: 0.9440 Epoch 299/500 236s 472ms/step - loss: 0.1827 - acc: 0.9893 - val_loss: 0.3462 - val_acc: 0.9443 Epoch 300/500 236s 471ms/step - loss: 0.1793 - acc: 0.9910 - val_loss: 0.3537 - val_acc: 0.9422 Epoch 301/500 lr changed to 0.0009999999776482583 236s 471ms/step - loss: 0.1692 - acc: 0.9943 - val_loss: 0.3201 - val_acc: 0.9535 Epoch 302/500 236s 472ms/step - loss: 0.1627 - acc: 0.9966 - val_loss: 0.3157 - val_acc: 0.9546 Epoch 303/500 236s 471ms/step - loss: 0.1600 - acc: 0.9972 - val_loss: 0.3139 - val_acc: 0.9556 Epoch 304/500 236s 472ms/step - loss: 0.1580 - acc: 0.9980 - val_loss: 0.3134 - val_acc: 0.9543 Epoch 305/500 236s 471ms/step - loss: 0.1571 - acc: 0.9981 - val_loss: 0.3141 - val_acc: 0.9555 Epoch 306/500 236s 472ms/step - loss: 0.1567 - acc: 0.9982 - val_loss: 0.3135 - val_acc: 0.9556 Epoch 307/500 236s 471ms/step - loss: 0.1557 - acc: 0.9983 - val_loss: 0.3120 - val_acc: 0.9546 Epoch 308/500 236s 471ms/step - loss: 0.1544 - acc: 0.9987 - val_loss: 0.3123 - val_acc: 0.9548 Epoch 309/500 236s 472ms/step - loss: 0.1543 - acc: 0.9986 - val_loss: 0.3112 - val_acc: 0.9569 Epoch 310/500 236s 471ms/step - loss: 0.1539 - acc: 0.9988 - val_loss: 0.3101 - val_acc: 0.9567 Epoch 311/500 236s 472ms/step - loss: 0.1534 - acc: 0.9988 - val_loss: 0.3122 - val_acc: 0.9568 Epoch 312/500 236s 471ms/step - loss: 0.1531 - acc: 0.9987 - val_loss: 0.3123 - val_acc: 0.9562 Epoch 313/500 236s 471ms/step - loss: 0.1526 - acc: 0.9985 - val_loss: 0.3129 - val_acc: 0.9559 Epoch 314/500 236s 472ms/step - loss: 0.1526 - acc: 0.9986 - val_loss: 0.3135 - val_acc: 0.9561 Epoch 315/500 236s 471ms/step - loss: 0.1518 - acc: 0.9988 - val_loss: 0.3111 - val_acc: 0.9569 Epoch 316/500 236s 471ms/step - loss: 0.1517 - acc: 0.9986 - val_loss: 0.3121 - val_acc: 0.9563 Epoch 317/500 236s 471ms/step - loss: 0.1511 - acc: 0.9989 - val_loss: 0.3115 - val_acc: 0.9563 Epoch 318/500 236s 471ms/step - loss: 0.1508 - acc: 0.9989 - val_loss: 0.3115 - val_acc: 0.9571 Epoch 319/500 236s 471ms/step - loss: 0.1500 - acc: 0.9990 - val_loss: 0.3102 - val_acc: 0.9576 Epoch 320/500 236s 471ms/step - loss: 0.1502 - acc: 0.9989 - val_loss: 0.3100 - val_acc: 0.9562 Epoch 321/500 236s 472ms/step - loss: 0.1494 - acc: 0.9990 - val_loss: 0.3092 - val_acc: 0.9569 Epoch 322/500 236s 472ms/step - loss: 0.1489 - acc: 0.9990 - val_loss: 0.3092 - val_acc: 0.9562 Epoch 323/500 236s 472ms/step - loss: 0.1492 - acc: 0.9988 - val_loss: 0.3078 - val_acc: 0.9566 Epoch 324/500 236s 472ms/step - loss: 0.1485 - acc: 0.9991 - val_loss: 0.3068 - val_acc: 0.9575 Epoch 325/500 236s 472ms/step - loss: 0.1485 - acc: 0.9991 - val_loss: 0.3066 - val_acc: 0.9567 Epoch 326/500 236s 471ms/step - loss: 0.1479 - acc: 0.9992 - val_loss: 0.3056 - val_acc: 0.9575 Epoch 327/500 236s 471ms/step - loss: 0.1475 - acc: 0.9992 - val_loss: 0.3074 - val_acc: 0.9576 Epoch 328/500 236s 472ms/step - loss: 0.1469 - acc: 0.9994 - val_loss: 0.3073 - val_acc: 0.9571 Epoch 329/500 236s 471ms/step - loss: 0.1472 - acc: 0.9991 - val_loss: 0.3074 - val_acc: 0.9563 Epoch 330/500 236s 471ms/step - loss: 0.1466 - acc: 0.9992 - val_loss: 0.3068 - val_acc: 0.9575 Epoch 331/500 236s 472ms/step - loss: 0.1461 - acc: 0.9993 - val_loss: 0.3067 - val_acc: 0.9570 Epoch 332/500 236s 472ms/step - loss: 0.1464 - acc: 0.9991 - val_loss: 0.3065 - val_acc: 0.9563 Epoch 333/500 236s 472ms/step - loss: 0.1456 - acc: 0.9991 - val_loss: 0.3081 - val_acc: 0.9572 Epoch 334/500 236s 471ms/step - loss: 0.1454 - acc: 0.9993 - val_loss: 0.3080 - val_acc: 0.9569 Epoch 335/500 236s 471ms/step - loss: 0.1451 - acc: 0.9992 - val_loss: 0.3069 - val_acc: 0.9569 Epoch 336/500 236s 471ms/step - loss: 0.1453 - acc: 0.9992 - val_loss: 0.3066 - val_acc: 0.9570 Epoch 337/500 236s 471ms/step - loss: 0.1450 - acc: 0.9992 - val_loss: 0.3060 - val_acc: 0.9572 Epoch 338/500 236s 472ms/step - loss: 0.1440 - acc: 0.9994 - val_loss: 0.3074 - val_acc: 0.9575 Epoch 339/500 236s 471ms/step - loss: 0.1441 - acc: 0.9993 - val_loss: 0.3071 - val_acc: 0.9574 Epoch 340/500 236s 472ms/step - loss: 0.1438 - acc: 0.9994 - val_loss: 0.3070 - val_acc: 0.9579 Epoch 341/500 236s 472ms/step - loss: 0.1435 - acc: 0.9993 - val_loss: 0.3061 - val_acc: 0.9576 Epoch 342/500 236s 472ms/step - loss: 0.1432 - acc: 0.9993 - val_loss: 0.3072 - val_acc: 0.9577 Epoch 343/500 236s 471ms/step - loss: 0.1431 - acc: 0.9992 - val_loss: 0.3061 - val_acc: 0.9573 Epoch 344/500 236s 471ms/step - loss: 0.1426 - acc: 0.9992 - val_loss: 0.3050 - val_acc: 0.9577 Epoch 345/500 236s 471ms/step - loss: 0.1426 - acc: 0.9994 - val_loss: 0.3051 - val_acc: 0.9577 Epoch 346/500 236s 472ms/step - loss: 0.1423 - acc: 0.9992 - val_loss: 0.3049 - val_acc: 0.9583 Epoch 347/500 235s 470ms/step - loss: 0.1421 - acc: 0.9993 - val_loss: 0.3044 - val_acc: 0.9581 Epoch 348/500 235s 469ms/step - loss: 0.1416 - acc: 0.9993 - val_loss: 0.3045 - val_acc: 0.9576 Epoch 349/500 234s 469ms/step - loss: 0.1414 - acc: 0.9994 - val_loss: 0.3044 - val_acc: 0.9573 Epoch 350/500 234s 469ms/step - loss: 0.1409 - acc: 0.9994 - val_loss: 0.3059 - val_acc: 0.9571 Epoch 351/500 234s 469ms/step - loss: 0.1407 - acc: 0.9996 - val_loss: 0.3051 - val_acc: 0.9569 Epoch 352/500 234s 469ms/step - loss: 0.1405 - acc: 0.9993 - val_loss: 0.3048 - val_acc: 0.9571 Epoch 353/500 235s 470ms/step - loss: 0.1405 - acc: 0.9993 - val_loss: 0.3034 - val_acc: 0.9575 Epoch 354/500 235s 469ms/step - loss: 0.1403 - acc: 0.9994 - val_loss: 0.3037 - val_acc: 0.9581 Epoch 355/500 234s 469ms/step - loss: 0.1401 - acc: 0.9993 - val_loss: 0.3037 - val_acc: 0.9583 Epoch 356/500 235s 470ms/step - loss: 0.1394 - acc: 0.9995 - val_loss: 0.3039 - val_acc: 0.9581 Epoch 357/500 235s 469ms/step - loss: 0.1393 - acc: 0.9993 - val_loss: 0.3009 - val_acc: 0.9589 Epoch 358/500 234s 469ms/step - loss: 0.1394 - acc: 0.9991 - val_loss: 0.3027 - val_acc: 0.9585 Epoch 359/500 234s 469ms/step - loss: 0.1389 - acc: 0.9995 - val_loss: 0.3031 - val_acc: 0.9582 Epoch 360/500 234s 469ms/step - loss: 0.1389 - acc: 0.9992 - val_loss: 0.3021 - val_acc: 0.9584 Epoch 361/500 234s 468ms/step - loss: 0.1387 - acc: 0.9994 - val_loss: 0.3027 - val_acc: 0.9595 Epoch 362/500 234s 468ms/step - loss: 0.1383 - acc: 0.9993 - val_loss: 0.3011 - val_acc: 0.9590 Epoch 363/500 234s 468ms/step - loss: 0.1381 - acc: 0.9994 - val_loss: 0.3017 - val_acc: 0.9583 Epoch 364/500 234s 468ms/step - loss: 0.1376 - acc: 0.9992 - val_loss: 0.3024 - val_acc: 0.9582 Epoch 365/500 234s 468ms/step - loss: 0.1372 - acc: 0.9995 - val_loss: 0.3016 - val_acc: 0.9580 Epoch 366/500 234s 468ms/step - loss: 0.1375 - acc: 0.9992 - val_loss: 0.3021 - val_acc: 0.9583 Epoch 367/500 234s 468ms/step - loss: 0.1370 - acc: 0.9994 - val_loss: 0.3022 - val_acc: 0.9576 Epoch 368/500 234s 469ms/step - loss: 0.1369 - acc: 0.9993 - val_loss: 0.3014 - val_acc: 0.9582 Epoch 369/500 234s 468ms/step - loss: 0.1363 - acc: 0.9994 - val_loss: 0.3031 - val_acc: 0.9576 Epoch 370/500 234s 469ms/step - loss: 0.1363 - acc: 0.9995 - val_loss: 0.3005 - val_acc: 0.9578 Epoch 371/500 235s 469ms/step - loss: 0.1358 - acc: 0.9995 - val_loss: 0.3020 - val_acc: 0.9574 Epoch 372/500 235s 469ms/step - loss: 0.1360 - acc: 0.9994 - val_loss: 0.3011 - val_acc: 0.9575 Epoch 373/500 235s 469ms/step - loss: 0.1359 - acc: 0.9994 - val_loss: 0.3002 - val_acc: 0.9576 Epoch 374/500 234s 469ms/step - loss: 0.1353 - acc: 0.9994 - val_loss: 0.3001 - val_acc: 0.9578 Epoch 375/500 234s 469ms/step - loss: 0.1351 - acc: 0.9995 - val_loss: 0.3004 - val_acc: 0.9575 Epoch 376/500 234s 469ms/step - loss: 0.1351 - acc: 0.9993 - val_loss: 0.3003 - val_acc: 0.9591 Epoch 377/500 234s 469ms/step - loss: 0.1347 - acc: 0.9994 - val_loss: 0.2993 - val_acc: 0.9586 Epoch 378/500 234s 468ms/step - loss: 0.1344 - acc: 0.9995 - val_loss: 0.2989 - val_acc: 0.9583 Epoch 379/500 235s 469ms/step - loss: 0.1347 - acc: 0.9992 - val_loss: 0.2994 - val_acc: 0.9584 Epoch 380/500 235s 469ms/step - loss: 0.1338 - acc: 0.9995 - val_loss: 0.2987 - val_acc: 0.9591 Epoch 381/500 234s 468ms/step - loss: 0.1337 - acc: 0.9994 - val_loss: 0.2994 - val_acc: 0.9590 Epoch 382/500 234s 468ms/step - loss: 0.1333 - acc: 0.9995 - val_loss: 0.2998 - val_acc: 0.9585 Epoch 383/500 234s 468ms/step - loss: 0.1332 - acc: 0.9994 - val_loss: 0.3007 - val_acc: 0.9576 Epoch 384/500 234s 468ms/step - loss: 0.1329 - acc: 0.9994 - val_loss: 0.2996 - val_acc: 0.9575 Epoch 385/500 234s 468ms/step - loss: 0.1331 - acc: 0.9993 - val_loss: 0.2985 - val_acc: 0.9582 Epoch 386/500 234s 468ms/step - loss: 0.1327 - acc: 0.9993 - val_loss: 0.2962 - val_acc: 0.9579 Epoch 387/500 234s 468ms/step - loss: 0.1327 - acc: 0.9994 - val_loss: 0.2965 - val_acc: 0.9582 Epoch 388/500 234s 468ms/step - loss: 0.1319 - acc: 0.9995 - val_loss: 0.2953 - val_acc: 0.9579 Epoch 389/500 234s 468ms/step - loss: 0.1319 - acc: 0.9995 - val_loss: 0.2979 - val_acc: 0.9582 Epoch 390/500 234s 468ms/step - loss: 0.1317 - acc: 0.9995 - val_loss: 0.2983 - val_acc: 0.9568 Epoch 391/500 234s 468ms/step - loss: 0.1316 - acc: 0.9994 - val_loss: 0.2973 - val_acc: 0.9570 Epoch 392/500 234s 468ms/step - loss: 0.1313 - acc: 0.9995 - val_loss: 0.2975 - val_acc: 0.9572 Epoch 393/500 234s 468ms/step - loss: 0.1314 - acc: 0.9993 - val_loss: 0.2969 - val_acc: 0.9584 Epoch 394/500 234s 468ms/step - loss: 0.1310 - acc: 0.9994 - val_loss: 0.2972 - val_acc: 0.9568 Epoch 395/500 234s 468ms/step - loss: 0.1312 - acc: 0.9993 - val_loss: 0.2954 - val_acc: 0.9581 Epoch 396/500 234s 468ms/step - loss: 0.1310 - acc: 0.9993 - val_loss: 0.2926 - val_acc: 0.9582 Epoch 397/500 234s 468ms/step - loss: 0.1304 - acc: 0.9995 - val_loss: 0.2932 - val_acc: 0.9585 Epoch 398/500 234s 468ms/step - loss: 0.1304 - acc: 0.9992 - val_loss: 0.2946 - val_acc: 0.9587 Epoch 399/500 234s 468ms/step - loss: 0.1298 - acc: 0.9994 - val_loss: 0.2959 - val_acc: 0.9582 Epoch 400/500 235s 469ms/step - loss: 0.1295 - acc: 0.9996 - val_loss: 0.2945 - val_acc: 0.9581 Epoch 401/500 235s 469ms/step - loss: 0.1296 - acc: 0.9995 - val_loss: 0.2945 - val_acc: 0.9590 Epoch 402/500 235s 469ms/step - loss: 0.1293 - acc: 0.9996 - val_loss: 0.2921 - val_acc: 0.9584 Epoch 403/500 235s 469ms/step - loss: 0.1290 - acc: 0.9994 - val_loss: 0.2917 - val_acc: 0.9583 Epoch 404/500 234s 469ms/step - loss: 0.1287 - acc: 0.9995 - val_loss: 0.2918 - val_acc: 0.9594 Epoch 405/500 234s 469ms/step - loss: 0.1287 - acc: 0.9994 - val_loss: 0.2947 - val_acc: 0.9579 Epoch 406/500 234s 469ms/step - loss: 0.1284 - acc: 0.9995 - val_loss: 0.2959 - val_acc: 0.9577 Epoch 407/500 234s 469ms/step - loss: 0.1284 - acc: 0.9993 - val_loss: 0.2966 - val_acc: 0.9578 Epoch 408/500 234s 469ms/step - loss: 0.1284 - acc: 0.9993 - val_loss: 0.2959 - val_acc: 0.9572 Epoch 409/500 234s 469ms/step - loss: 0.1278 - acc: 0.9995 - val_loss: 0.2936 - val_acc: 0.9573 Epoch 410/500 234s 469ms/step - loss: 0.1278 - acc: 0.9994 - val_loss: 0.2942 - val_acc: 0.9569 Epoch 411/500 234s 469ms/step - loss: 0.1274 - acc: 0.9994 - val_loss: 0.2942 - val_acc: 0.9566 Epoch 412/500 234s 469ms/step - loss: 0.1271 - acc: 0.9994 - val_loss: 0.2928 - val_acc: 0.9579 Epoch 413/500 235s 469ms/step - loss: 0.1271 - acc: 0.9994 - val_loss: 0.2910 - val_acc: 0.9575 Epoch 414/500 234s 469ms/step - loss: 0.1267 - acc: 0.9994 - val_loss: 0.2889 - val_acc: 0.9581 Epoch 415/500 234s 469ms/step - loss: 0.1267 - acc: 0.9994 - val_loss: 0.2906 - val_acc: 0.9583 Epoch 416/500 234s 469ms/step - loss: 0.1263 - acc: 0.9995 - val_loss: 0.2907 - val_acc: 0.9575 Epoch 417/500 234s 469ms/step - loss: 0.1265 - acc: 0.9993 - val_loss: 0.2888 - val_acc: 0.9584 Epoch 418/500 234s 469ms/step - loss: 0.1261 - acc: 0.9994 - val_loss: 0.2894 - val_acc: 0.9584 Epoch 419/500 234s 469ms/step - loss: 0.1255 - acc: 0.9996 - val_loss: 0.2908 - val_acc: 0.9583 Epoch 420/500 234s 469ms/step - loss: 0.1257 - acc: 0.9993 - val_loss: 0.2925 - val_acc: 0.9583 Epoch 421/500 234s 469ms/step - loss: 0.1255 - acc: 0.9993 - val_loss: 0.2911 - val_acc: 0.9584 Epoch 422/500 235s 470ms/step - loss: 0.1253 - acc: 0.9995 - val_loss: 0.2907 - val_acc: 0.9574 Epoch 423/500 235s 469ms/step - loss: 0.1248 - acc: 0.9996 - val_loss: 0.2899 - val_acc: 0.9573 Epoch 424/500 235s 469ms/step - loss: 0.1251 - acc: 0.9993 - val_loss: 0.2932 - val_acc: 0.9582 Epoch 425/500 235s 469ms/step - loss: 0.1246 - acc: 0.9994 - val_loss: 0.2925 - val_acc: 0.9579 Epoch 426/500 234s 469ms/step - loss: 0.1245 - acc: 0.9993 - val_loss: 0.2902 - val_acc: 0.9574 Epoch 427/500 234s 469ms/step - loss: 0.1242 - acc: 0.9995 - val_loss: 0.2919 - val_acc: 0.9570 Epoch 428/500 234s 469ms/step - loss: 0.1237 - acc: 0.9995 - val_loss: 0.2930 - val_acc: 0.9577 Epoch 429/500 235s 469ms/step - loss: 0.1239 - acc: 0.9994 - val_loss: 0.2921 - val_acc: 0.9584 Epoch 430/500 234s 469ms/step - loss: 0.1239 - acc: 0.9994 - val_loss: 0.2924 - val_acc: 0.9575 Epoch 431/500 234s 469ms/step - loss: 0.1237 - acc: 0.9993 - val_loss: 0.2931 - val_acc: 0.9587 Epoch 432/500 235s 470ms/step - loss: 0.1232 - acc: 0.9996 - val_loss: 0.2933 - val_acc: 0.9585 Epoch 433/500 235s 470ms/step - loss: 0.1233 - acc: 0.9995 - val_loss: 0.2922 - val_acc: 0.9584 Epoch 434/500 235s 469ms/step - loss: 0.1226 - acc: 0.9996 - val_loss: 0.2936 - val_acc: 0.9582 Epoch 435/500 234s 469ms/step - loss: 0.1230 - acc: 0.9993 - val_loss: 0.2946 - val_acc: 0.9584 Epoch 436/500 234s 469ms/step - loss: 0.1224 - acc: 0.9996 - val_loss: 0.2939 - val_acc: 0.9573 Epoch 437/500 235s 470ms/step - loss: 0.1227 - acc: 0.9992 - val_loss: 0.2949 - val_acc: 0.9577 Epoch 438/500 235s 469ms/step - loss: 0.1222 - acc: 0.9995 - val_loss: 0.2933 - val_acc: 0.9581 Epoch 439/500 235s 469ms/step - loss: 0.1217 - acc: 0.9995 - val_loss: 0.2919 - val_acc: 0.9574 Epoch 440/500 234s 469ms/step - loss: 0.1215 - acc: 0.9995 - val_loss: 0.2917 - val_acc: 0.9583 Epoch 441/500 235s 469ms/step - loss: 0.1213 - acc: 0.9996 - val_loss: 0.2930 - val_acc: 0.9577 Epoch 442/500 234s 469ms/step - loss: 0.1217 - acc: 0.9993 - val_loss: 0.2951 - val_acc: 0.9571 Epoch 443/500 234s 469ms/step - loss: 0.1212 - acc: 0.9995 - val_loss: 0.2957 - val_acc: 0.9566 Epoch 444/500 234s 469ms/step - loss: 0.1207 - acc: 0.9996 - val_loss: 0.2949 - val_acc: 0.9569 Epoch 445/500 234s 469ms/step - loss: 0.1210 - acc: 0.9994 - val_loss: 0.2923 - val_acc: 0.9569 Epoch 446/500 234s 469ms/step - loss: 0.1206 - acc: 0.9994 - val_loss: 0.2919 - val_acc: 0.9573 Epoch 447/500 234s 469ms/step - loss: 0.1209 - acc: 0.9993 - val_loss: 0.2947 - val_acc: 0.9577 Epoch 448/500 234s 469ms/step - loss: 0.1203 - acc: 0.9995 - val_loss: 0.2923 - val_acc: 0.9581 Epoch 449/500 234s 469ms/step - loss: 0.1203 - acc: 0.9994 - val_loss: 0.2926 - val_acc: 0.9579 Epoch 450/500 235s 469ms/step - loss: 0.1199 - acc: 0.9995 - val_loss: 0.2888 - val_acc: 0.9591 Epoch 451/500 lr changed to 9.999999310821295e-05 234s 469ms/step - loss: 0.1197 - acc: 0.9994 - val_loss: 0.2890 - val_acc: 0.9588 Epoch 452/500 235s 470ms/step - loss: 0.1196 - acc: 0.9996 - val_loss: 0.2891 - val_acc: 0.9587 Epoch 453/500 235s 469ms/step - loss: 0.1197 - acc: 0.9997 - val_loss: 0.2887 - val_acc: 0.9589 Epoch 454/500 235s 470ms/step - loss: 0.1197 - acc: 0.9994 - val_loss: 0.2888 - val_acc: 0.9588 Epoch 455/500 234s 469ms/step - loss: 0.1196 - acc: 0.9995 - val_loss: 0.2885 - val_acc: 0.9592 Epoch 456/500 234s 469ms/step - loss: 0.1196 - acc: 0.9994 - val_loss: 0.2887 - val_acc: 0.9586 Epoch 457/500 235s 469ms/step - loss: 0.1195 - acc: 0.9996 - val_loss: 0.2885 - val_acc: 0.9583 Epoch 458/500 234s 469ms/step - loss: 0.1194 - acc: 0.9997 - val_loss: 0.2886 - val_acc: 0.9585 Epoch 459/500 235s 471ms/step - loss: 0.1193 - acc: 0.9996 - val_loss: 0.2887 - val_acc: 0.9585 Epoch 460/500 235s 470ms/step - loss: 0.1191 - acc: 0.9997 - val_loss: 0.2885 - val_acc: 0.9581 Epoch 461/500 235s 470ms/step - loss: 0.1194 - acc: 0.9995 - val_loss: 0.2887 - val_acc: 0.9585 Epoch 462/500 235s 470ms/step - loss: 0.1192 - acc: 0.9996 - val_loss: 0.2886 - val_acc: 0.9584 Epoch 463/500 235s 469ms/step - loss: 0.1192 - acc: 0.9996 - val_loss: 0.2887 - val_acc: 0.9579 Epoch 464/500 234s 469ms/step - loss: 0.1194 - acc: 0.9996 - val_loss: 0.2891 - val_acc: 0.9583 Epoch 465/500 234s 469ms/step - loss: 0.1194 - acc: 0.9995 - val_loss: 0.2888 - val_acc: 0.9585 Epoch 466/500 234s 469ms/step - loss: 0.1189 - acc: 0.9998 - val_loss: 0.2889 - val_acc: 0.9586 Epoch 467/500 234s 469ms/step - loss: 0.1190 - acc: 0.9997 - val_loss: 0.2885 - val_acc: 0.9584 Epoch 468/500 234s 469ms/step - loss: 0.1192 - acc: 0.9995 - val_loss: 0.2882 - val_acc: 0.9582 Epoch 469/500 235s 469ms/step - loss: 0.1192 - acc: 0.9996 - val_loss: 0.2882 - val_acc: 0.9582 Epoch 470/500 235s 469ms/step - loss: 0.1191 - acc: 0.9996 - val_loss: 0.2885 - val_acc: 0.9579 Epoch 471/500 234s 469ms/step - loss: 0.1193 - acc: 0.9995 - val_loss: 0.2885 - val_acc: 0.9580 Epoch 472/500 235s 471ms/step - loss: 0.1195 - acc: 0.9994 - val_loss: 0.2885 - val_acc: 0.9579 Epoch 473/500 235s 470ms/step - loss: 0.1191 - acc: 0.9996 - val_loss: 0.2883 - val_acc: 0.9578 Epoch 474/500 235s 470ms/step - loss: 0.1195 - acc: 0.9994 - val_loss: 0.2887 - val_acc: 0.9579 Epoch 475/500 235s 469ms/step - loss: 0.1192 - acc: 0.9995 - val_loss: 0.2884 - val_acc: 0.9582 Epoch 476/500 234s 469ms/step - loss: 0.1190 - acc: 0.9996 - val_loss: 0.2885 - val_acc: 0.9581 Epoch 477/500 234s 469ms/step - loss: 0.1189 - acc: 0.9997 - val_loss: 0.2888 - val_acc: 0.9581 Epoch 478/500 234s 469ms/step - loss: 0.1190 - acc: 0.9996 - val_loss: 0.2889 - val_acc: 0.9582 Epoch 479/500 234s 469ms/step - loss: 0.1187 - acc: 0.9997 - val_loss: 0.2887 - val_acc: 0.9584 Epoch 480/500 234s 469ms/step - loss: 0.1188 - acc: 0.9996 - val_loss: 0.2882 - val_acc: 0.9581 Epoch 481/500 235s 469ms/step - loss: 0.1191 - acc: 0.9996 - val_loss: 0.2883 - val_acc: 0.9578 Epoch 482/500 235s 471ms/step - loss: 0.1189 - acc: 0.9996 - val_loss: 0.2880 - val_acc: 0.9581 Epoch 483/500 234s 469ms/step - loss: 0.1186 - acc: 0.9996 - val_loss: 0.2881 - val_acc: 0.9582 Epoch 484/500 234s 469ms/step - loss: 0.1188 - acc: 0.9996 - val_loss: 0.2881 - val_acc: 0.9577 Epoch 485/500 234s 469ms/step - loss: 0.1190 - acc: 0.9996 - val_loss: 0.2884 - val_acc: 0.9578 Epoch 486/500 234s 469ms/step - loss: 0.1186 - acc: 0.9997 - val_loss: 0.2884 - val_acc: 0.9579 Epoch 487/500 234s 469ms/step - loss: 0.1189 - acc: 0.9995 - val_loss: 0.2883 - val_acc: 0.9580 Epoch 488/500 237s 473ms/step - loss: 0.1183 - acc: 0.9997 - val_loss: 0.2883 - val_acc: 0.9580 Epoch 489/500 235s 469ms/step - loss: 0.1186 - acc: 0.9997 - val_loss: 0.2885 - val_acc: 0.9578 Epoch 490/500 234s 468ms/step - loss: 0.1186 - acc: 0.9997 - val_loss: 0.2887 - val_acc: 0.9579 Epoch 491/500 234s 468ms/step - loss: 0.1187 - acc: 0.9996 - val_loss: 0.2884 - val_acc: 0.9581 Epoch 492/500 234s 469ms/step - loss: 0.1187 - acc: 0.9994 - val_loss: 0.2881 - val_acc: 0.9579 Epoch 493/500 234s 469ms/step - loss: 0.1185 - acc: 0.9996 - val_loss: 0.2883 - val_acc: 0.9578 Epoch 494/500 234s 469ms/step - loss: 0.1187 - acc: 0.9997 - val_loss: 0.2881 - val_acc: 0.9579 Epoch 495/500 234s 468ms/step - loss: 0.1187 - acc: 0.9997 - val_loss: 0.2880 - val_acc: 0.9579 Epoch 496/500 234s 469ms/step - loss: 0.1185 - acc: 0.9996 - val_loss: 0.2881 - val_acc: 0.9579 Epoch 497/500 234s 469ms/step - loss: 0.1186 - acc: 0.9996 - val_loss: 0.2881 - val_acc: 0.9582 Epoch 498/500 235s 469ms/step - loss: 0.1187 - acc: 0.9995 - val_loss: 0.2883 - val_acc: 0.9581 Epoch 499/500 234s 469ms/step - loss: 0.1184 - acc: 0.9996 - val_loss: 0.2884 - val_acc: 0.9580 Epoch 500/500 234s 469ms/step - loss: 0.1187 - acc: 0.9996 - val_loss: 0.2881 - val_acc: 0.9580 Train loss: 0.11700774446129798 Train accuracy: 0.999960000038147 Test loss: 0.2880836722254753 Test accuracy: 0.9580000066757202
測試準確率是95.80%,離96%還差一點。
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, Date of Publication: 13 February 2020
https://ieeexplore.ieee.org/document/8998530
————————————————
版權宣告:本文為CSDN博主「dangqing1988」的原創文章,遵循CC 4.0 BY-SA版權協議,轉載請附上原文出處連結及本宣告。
原文連結: https://blog.csdn.net/dangqing1988/article/details/106245196
來自 “ ITPUB部落格 ” ,連結:http://blog.itpub.net/69972329/viewspace-2693939/,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22)Cifar10~95.25%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