在之前调参记录的基础上，首先，大幅度削减了自适应参数化ReLU中全连接神经元的个数，想着可以减轻训练的难度，也可以减少过拟合；然后，将Epoch增加到1000个，继续测试ResNet+自适应参数化ReLU激活函数在Cifar10上的效果。
自适应参数化ReLU激活函数的基本原理如下：
Keras程序：
#!/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, DOI: 10.1109/TIE.2020.2972458,
Date of Publication: 13 February 2020
@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, Lambda
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)
def cal_mean(inputs):
outputs = K.mean(inputs, axis=1, keepdims=True)
return outputs
# 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 300 epoches
def scheduler(epoch):
if epoch % 300 == 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 = Lambda(cal_mean)(GlobalAveragePooling2D()(pos_input))
scales_n = Lambda(cal_mean)(GlobalAveragePooling2D()(neg_input))
scales = Concatenate()([scales_n, scales_p])
scales = Dense(2, 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(1, 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((1,1,1))(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 = aprelu(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 = aprelu(residual)
residual = Conv2D(out_channels, 3, padding='same', kernel_initializer='he_normal',
kernel_regularizer=l2(1e-4))(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 = aprelu(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=1000,
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])实验结果：
Using TensorFlow backend.
x_train shape: (50000, 32, 32, 3)
50000 train samples
10000 test samples
Epoch 1/1000
413s 826ms/step - loss: 6.7483 - acc: 0.3694 - val_loss: 5.9301 - val_acc: 0.4997
Epoch 2/1000
330s 659ms/step - loss: 5.5312 - acc: 0.4941 - val_loss: 4.9301 - val_acc: 0.5792
Epoch 3/1000
330s 659ms/step - loss: 4.6439 - acc: 0.5593 - val_loss: 4.0912 - val_acc: 0.6487
Epoch 4/1000
329s 659ms/step - loss: 3.9263 - acc: 0.6101 - val_loss: 3.4891 - val_acc: 0.6778
Epoch 5/1000
329s 659ms/step - loss: 3.3595 - acc: 0.6415 - val_loss: 2.9574 - val_acc: 0.7220
Epoch 6/1000
334s 668ms/step - loss: 2.8985 - acc: 0.6704 - val_loss: 2.5622 - val_acc: 0.7322
Epoch 7/1000
330s 660ms/step - loss: 2.5229 - acc: 0.6917 - val_loss: 2.2063 - val_acc: 0.7612
Epoch 8/1000
329s 659ms/step - loss: 2.2063 - acc: 0.7167 - val_loss: 1.9074 - val_acc: 0.7861
Epoch 9/1000
329s 659ms/step - loss: 1.9480 - acc: 0.7339 - val_loss: 1.6790 - val_acc: 0.8023
Epoch 10/1000
330s 659ms/step - loss: 1.7331 - acc: 0.7527 - val_loss: 1.5017 - val_acc: 0.8113
Epoch 11/1000
330s 659ms/step - loss: 1.5656 - acc: 0.7652 - val_loss: 1.3648 - val_acc: 0.8174
Epoch 12/1000
330s 659ms/step - loss: 1.4253 - acc: 0.7756 - val_loss: 1.2259 - val_acc: 0.8327
Epoch 13/1000
329s 659ms/step - loss: 1.3032 - acc: 0.7877 - val_loss: 1.1070 - val_acc: 0.8442
Epoch 14/1000
330s 659ms/step - loss: 1.2136 - acc: 0.7941 - val_loss: 1.0533 - val_acc: 0.8408
Epoch 15/1000
329s 659ms/step - loss: 1.1363 - acc: 0.8009 - val_loss: 0.9825 - val_acc: 0.8517
Epoch 16/1000
331s 661ms/step - loss: 1.0679 - acc: 0.8111 - val_loss: 0.9233 - val_acc: 0.8563
Epoch 17/1000
330s 660ms/step - loss: 1.0143 - acc: 0.8168 - val_loss: 0.8913 - val_acc: 0.8569
Epoch 18/1000
330s 660ms/step - loss: 0.9731 - acc: 0.8215 - val_loss: 0.8578 - val_acc: 0.8561
Epoch 19/1000
330s 660ms/step - loss: 0.9397 - acc: 0.8261 - val_loss: 0.8226 - val_acc: 0.8675
Epoch 20/1000
330s 659ms/step - loss: 0.9073 - acc: 0.8301 - val_loss: 0.7973 - val_acc: 0.8697
Epoch 21/1000
329s 659ms/step - loss: 0.8823 - acc: 0.8342 - val_loss: 0.7837 - val_acc: 0.8693
Epoch 22/1000
331s 662ms/step - loss: 0.8688 - acc: 0.8369 - val_loss: 0.7693 - val_acc: 0.8736
Epoch 23/1000
332s 665ms/step - loss: 0.8479 - acc: 0.8425 - val_loss: 0.7626 - val_acc: 0.8730
Epoch 24/1000
329s 659ms/step - loss: 0.8329 - acc: 0.8453 - val_loss: 0.7745 - val_acc: 0.8671
Epoch 25/1000
330s 660ms/step - loss: 0.8271 - acc: 0.8462 - val_loss: 0.7271 - val_acc: 0.8833
Epoch 26/1000
331s 661ms/step - loss: 0.8121 - acc: 0.8516 - val_loss: 0.7155 - val_acc: 0.8883
Epoch 27/1000
330s 659ms/step - loss: 0.8008 - acc: 0.8543 - val_loss: 0.7522 - val_acc: 0.8707
Epoch 28/1000
330s 659ms/step - loss: 0.7959 - acc: 0.8558 - val_loss: 0.7200 - val_acc: 0.8799
Epoch 29/1000
329s 659ms/step - loss: 0.7937 - acc: 0.8563 - val_loss: 0.7196 - val_acc: 0.8855
Epoch 30/1000
329s 659ms/step - loss: 0.7843 - acc: 0.8602 - val_loss: 0.7285 - val_acc: 0.8808
Epoch 31/1000
330s 659ms/step - loss: 0.7791 - acc: 0.8606 - val_loss: 0.7120 - val_acc: 0.8879
Epoch 32/1000
329s 658ms/step - loss: 0.7781 - acc: 0.8612 - val_loss: 0.7139 - val_acc: 0.8840
Epoch 33/1000
329s 659ms/step - loss: 0.7650 - acc: 0.8660 - val_loss: 0.7243 - val_acc: 0.8814
Epoch 34/1000
329s 658ms/step - loss: 0.7686 - acc: 0.8669 - val_loss: 0.7223 - val_acc: 0.8811
Epoch 35/1000
329s 659ms/step - loss: 0.7694 - acc: 0.8662 - val_loss: 0.6995 - val_acc: 0.8938
Epoch 36/1000
329s 659ms/step - loss: 0.7618 - acc: 0.8679 - val_loss: 0.7032 - val_acc: 0.8919
Epoch 37/1000
329s 658ms/step - loss: 0.7542 - acc: 0.8709 - val_loss: 0.7115 - val_acc: 0.8868
Epoch 38/1000
329s 659ms/step - loss: 0.7525 - acc: 0.8719 - val_loss: 0.7282 - val_acc: 0.8791
Epoch 39/1000
329s 658ms/step - loss: 0.7528 - acc: 0.8727 - val_loss: 0.7046 - val_acc: 0.8898
Epoch 40/1000
329s 658ms/step - loss: 0.7476 - acc: 0.8748 - val_loss: 0.7071 - val_acc: 0.8927
Epoch 41/1000
329s 659ms/step - loss: 0.7429 - acc: 0.8762 - val_loss: 0.7195 - val_acc: 0.8852
Epoch 42/1000
329s 658ms/step - loss: 0.7487 - acc: 0.8758 - val_loss: 0.7064 - val_acc: 0.8936
Epoch 43/1000
329s 658ms/step - loss: 0.7423 - acc: 0.8769 - val_loss: 0.6941 - val_acc: 0.8959
Epoch 44/1000
330s 660ms/step - loss: 0.7357 - acc: 0.8802 - val_loss: 0.7097 - val_acc: 0.8921
Epoch 45/1000
329s 659ms/step - loss: 0.7385 - acc: 0.8796 - val_loss: 0.6845 - val_acc: 0.8997
Epoch 46/1000
332s 664ms/step - loss: 0.7331 - acc: 0.8821 - val_loss: 0.7265 - val_acc: 0.8870
Epoch 47/1000
329s 658ms/step - loss: 0.7358 - acc: 0.8804 - val_loss: 0.7062 - val_acc: 0.8947
Epoch 48/1000
329s 658ms/step - loss: 0.7357 - acc: 0.8812 - val_loss: 0.7007 - val_acc: 0.8932
Epoch 49/1000
329s 658ms/step - loss: 0.7338 - acc: 0.8832 - val_loss: 0.6983 - val_acc: 0.8989
Epoch 50/1000
330s 659ms/step - loss: 0.7268 - acc: 0.8852 - val_loss: 0.7136 - val_acc: 0.8940
Epoch 51/1000
329s 659ms/step - loss: 0.7327 - acc: 0.8845 - val_loss: 0.7076 - val_acc: 0.8957
Epoch 52/1000
329s 658ms/step - loss: 0.7255 - acc: 0.8858 - val_loss: 0.6909 - val_acc: 0.8995
Epoch 53/1000
329s 658ms/step - loss: 0.7298 - acc: 0.8860 - val_loss: 0.7088 - val_acc: 0.8956
Epoch 54/1000
331s 661ms/step - loss: 0.7237 - acc: 0.8884 - val_loss: 0.7069 - val_acc: 0.8960
Epoch 55/1000
329s 659ms/step - loss: 0.7288 - acc: 0.8863 - val_loss: 0.6929 - val_acc: 0.8984
Epoch 56/1000
329s 658ms/step - loss: 0.7208 - acc: 0.8890 - val_loss: 0.6925 - val_acc: 0.9017
Epoch 57/1000
330s 659ms/step - loss: 0.7187 - acc: 0.8898 - val_loss: 0.7150 - val_acc: 0.8944
Epoch 58/1000
330s 659ms/step - loss: 0.7148 - acc: 0.8920 - val_loss: 0.6766 - val_acc: 0.9060
Epoch 59/1000
329s 659ms/step - loss: 0.7185 - acc: 0.8905 - val_loss: 0.7031 - val_acc: 0.8951
Epoch 60/1000
329s 658ms/step - loss: 0.7192 - acc: 0.8904 - val_loss: 0.6956 - val_acc: 0.9020
Epoch 61/1000
329s 659ms/step - loss: 0.7209 - acc: 0.8895 - val_loss: 0.6967 - val_acc: 0.8997
Epoch 62/1000
329s 658ms/step - loss: 0.7083 - acc: 0.8936 - val_loss: 0.6806 - val_acc: 0.9048
Epoch 63/1000
329s 658ms/step - loss: 0.7123 - acc: 0.8926 - val_loss: 0.6832 - val_acc: 0.9042
Epoch 64/1000
329s 658ms/step - loss: 0.7085 - acc: 0.8939 - val_loss: 0.7000 - val_acc: 0.8973
Epoch 65/1000
330s 660ms/step - loss: 0.7049 - acc: 0.8963 - val_loss: 0.6993 - val_acc: 0.8969
Epoch 66/1000
330s 660ms/step - loss: 0.7062 - acc: 0.8950 - val_loss: 0.6977 - val_acc: 0.9044
Epoch 67/1000
329s 659ms/step - loss: 0.7127 - acc: 0.8938 - val_loss: 0.7069 - val_acc: 0.8977
Epoch 68/1000
329s 657ms/step - loss: 0.7010 - acc: 0.8960 - val_loss: 0.7052 - val_acc: 0.8994
Epoch 69/1000
329s 658ms/step - loss: 0.7096 - acc: 0.8934 - val_loss: 0.6982 - val_acc: 0.9027
Epoch 70/1000
329s 658ms/step - loss: 0.7118 - acc: 0.8947 - val_loss: 0.6850 - val_acc: 0.9053
Epoch 71/1000
329s 658ms/step - loss: 0.7076 - acc: 0.8953 - val_loss: 0.6798 - val_acc: 0.9053
Epoch 72/1000
329s 658ms/step - loss: 0.6998 - acc: 0.8981 - val_loss: 0.7312 - val_acc: 0.8905
Epoch 73/1000
329s 658ms/step - loss: 0.7049 - acc: 0.8969 - val_loss: 0.7042 - val_acc: 0.8996
Epoch 74/1000
329s 658ms/step - loss: 0.7094 - acc: 0.8959 - val_loss: 0.6987 - val_acc: 0.9020
Epoch 75/1000
329s 658ms/step - loss: 0.7045 - acc: 0.8963 - val_loss: 0.6909 - val_acc: 0.9056
Epoch 76/1000
329s 658ms/step - loss: 0.7028 - acc: 0.8980 - val_loss: 0.7257 - val_acc: 0.8918
Epoch 77/1000
329s 657ms/step - loss: 0.7059 - acc: 0.8956 - val_loss: 0.6813 - val_acc: 0.9097
Epoch 78/1000
329s 658ms/step - loss: 0.6996 - acc: 0.9006 - val_loss: 0.6942 - val_acc: 0.9081
Epoch 79/1000
329s 658ms/step - loss: 0.6988 - acc: 0.8996 - val_loss: 0.6801 - val_acc: 0.9059
Epoch 80/1000
329s 657ms/step - loss: 0.6983 - acc: 0.8990 - val_loss: 0.6759 - val_acc: 0.9081
Epoch 81/1000
329s 658ms/step - loss: 0.6968 - acc: 0.9012 - val_loss: 0.6868 - val_acc: 0.9052
Epoch 82/1000
329s 658ms/step - loss: 0.6976 - acc: 0.9000 - val_loss: 0.6945 - val_acc: 0.9028
Epoch 83/1000
329s 658ms/step - loss: 0.6974 - acc: 0.8999 - val_loss: 0.7020 - val_acc: 0.9010
Epoch 84/1000
329s 658ms/step - loss: 0.6984 - acc: 0.9003 - val_loss: 0.7220 - val_acc: 0.8956
Epoch 85/1000
329s 658ms/step - loss: 0.7054 - acc: 0.8974 - val_loss: 0.6846 - val_acc: 0.9082
Epoch 86/1000
329s 657ms/step - loss: 0.6974 - acc: 0.9005 - val_loss: 0.6880 - val_acc: 0.9085
Epoch 87/1000
329s 658ms/step - loss: 0.6989 - acc: 0.8998 - val_loss: 0.6956 - val_acc: 0.9035
Epoch 88/1000
329s 658ms/step - loss: 0.7002 - acc: 0.8998 - val_loss: 0.6852 - val_acc: 0.9056
Epoch 89/1000
329s 657ms/step - loss: 0.7005 - acc: 0.9004 - val_loss: 0.6891 - val_acc: 0.9077
Epoch 90/1000
329s 658ms/step - loss: 0.6945 - acc: 0.9030 - val_loss: 0.6880 - val_acc: 0.9071
Epoch 91/1000
329s 658ms/step - loss: 0.6999 - acc: 0.9011 - val_loss: 0.6868 - val_acc: 0.9064
Epoch 92/1000
329s 658ms/step - loss: 0.6976 - acc: 0.9028 - val_loss: 0.6964 - val_acc: 0.9022
Epoch 93/1000
329s 658ms/step - loss: 0.6923 - acc: 0.9045 - val_loss: 0.6712 - val_acc: 0.9126
Epoch 94/1000
329s 657ms/step - loss: 0.6981 - acc: 0.9020 - val_loss: 0.6813 - val_acc: 0.9084
Epoch 95/1000
329s 657ms/step - loss: 0.6952 - acc: 0.9029 - val_loss: 0.6713 - val_acc: 0.9135
Epoch 96/1000
329s 657ms/step - loss: 0.6958 - acc: 0.9032 - val_loss: 0.6835 - val_acc: 0.9084
Epoch 97/1000
329s 657ms/step - loss: 0.6889 - acc: 0.9045 - val_loss: 0.6818 - val_acc: 0.9053
Epoch 98/1000
329s 657ms/step - loss: 0.6997 - acc: 0.9009 - val_loss: 0.6886 - val_acc: 0.9075
Epoch 99/1000
329s 657ms/step - loss: 0.6948 - acc: 0.9029 - val_loss: 0.6872 - val_acc: 0.9057
Epoch 100/1000
329s 657ms/step - loss: 0.6925 - acc: 0.9044 - val_loss: 0.6992 - val_acc: 0.9051
Epoch 101/1000
329s 657ms/step - loss: 0.6948 - acc: 0.9033 - val_loss: 0.7153 - val_acc: 0.8990
Epoch 102/1000
328s 657ms/step - loss: 0.6998 - acc: 0.9025 - val_loss: 0.6735 - val_acc: 0.9155
Epoch 103/1000
329s 657ms/step - loss: 0.6931 - acc: 0.9037 - val_loss: 0.6907 - val_acc: 0.9052
Epoch 104/1000
329s 657ms/step - loss: 0.6941 - acc: 0.9048 - val_loss: 0.7109 - val_acc: 0.9046
Epoch 105/1000
329s 658ms/step - loss: 0.6902 - acc: 0.9066 - val_loss: 0.6744 - val_acc: 0.9114
Epoch 106/1000
329s 657ms/step - loss: 0.6944 - acc: 0.9038 - val_loss: 0.6758 - val_acc: 0.9119
Epoch 107/1000
329s 658ms/step - loss: 0.6895 - acc: 0.9036 - val_loss: 0.7041 - val_acc: 0.9024
Epoch 108/1000
329s 657ms/step - loss: 0.6976 - acc: 0.9026 - val_loss: 0.6959 - val_acc: 0.9065
Epoch 109/1000
329s 657ms/step - loss: 0.6890 - acc: 0.9063 - val_loss: 0.6866 - val_acc: 0.9125
Epoch 110/1000
329s 657ms/step - loss: 0.6897 - acc: 0.9062 - val_loss: 0.6786 - val_acc: 0.9150
Epoch 111/1000
329s 657ms/step - loss: 0.6941 - acc: 0.9061 - val_loss: 0.6870 - val_acc: 0.9100
Epoch 112/1000
329s 657ms/step - loss: 0.6916 - acc: 0.9052 - val_loss: 0.6836 - val_acc: 0.9069
Epoch 113/1000
329s 657ms/step - loss: 0.6908 - acc: 0.9040 - val_loss: 0.6841 - val_acc: 0.9108
Epoch 114/1000
329s 657ms/step - loss: 0.6907 - acc: 0.9059 - val_loss: 0.6831 - val_acc: 0.9102
Epoch 115/1000
328s 657ms/step - loss: 0.6904 - acc: 0.9063 - val_loss: 0.6852 - val_acc: 0.9101
Epoch 116/1000
329s 657ms/step - loss: 0.6940 - acc: 0.9051 - val_loss: 0.6977 - val_acc: 0.9069
Epoch 117/1000
329s 657ms/step - loss: 0.6909 - acc: 0.9075 - val_loss: 0.6904 - val_acc: 0.9110
Epoch 118/1000
329s 658ms/step - loss: 0.6888 - acc: 0.9066 - val_loss: 0.6845 - val_acc: 0.9144
Epoch 119/1000
328s 657ms/step - loss: 0.6890 - acc: 0.9058 - val_loss: 0.6954 - val_acc: 0.9076
Epoch 120/1000
329s 657ms/step - loss: 0.6893 - acc: 0.9059 - val_loss: 0.6790 - val_acc: 0.9125
Epoch 121/1000
329s 657ms/step - loss: 0.6840 - acc: 0.9082 - val_loss: 0.6813 - val_acc: 0.9088
Epoch 122/1000
329s 657ms/step - loss: 0.6966 - acc: 0.9032 - val_loss: 0.6816 - val_acc: 0.9079
Epoch 123/1000
329s 657ms/step - loss: 0.6869 - acc: 0.9093 - val_loss: 0.6970 - val_acc: 0.9057
Epoch 124/1000
329s 657ms/step - loss: 0.6907 - acc: 0.9052 - val_loss: 0.6734 - val_acc: 0.9143
Epoch 125/1000
328s 657ms/step - loss: 0.6886 - acc: 0.9074 - val_loss: 0.6811 - val_acc: 0.9102
Epoch 126/1000
329s 657ms/step - loss: 0.6862 - acc: 0.9082 - val_loss: 0.7166 - val_acc: 0.8981
Epoch 127/1000
329s 657ms/step - loss: 0.6882 - acc: 0.9078 - val_loss: 0.6780 - val_acc: 0.9125
Epoch 128/1000
329s 657ms/step - loss: 0.6972 - acc: 0.9036 - val_loss: 0.6861 - val_acc: 0.9110
Epoch 129/1000
329s 657ms/step - loss: 0.6848 - acc: 0.9099 - val_loss: 0.6905 - val_acc: 0.9069
Epoch 130/1000
329s 657ms/step - loss: 0.6868 - acc: 0.9072 - val_loss: 0.6773 - val_acc: 0.9150
Epoch 131/1000
329s 657ms/step - loss: 0.6915 - acc: 0.9070 - val_loss: 0.6852 - val_acc: 0.9113
Epoch 132/1000
329s 657ms/step - loss: 0.6879 - acc: 0.9075 - val_loss: 0.6821 - val_acc: 0.9116
Epoch 133/1000
328s 657ms/step - loss: 0.6903 - acc: 0.9068 - val_loss: 0.6854 - val_acc: 0.9102
Epoch 134/1000
329s 657ms/step - loss: 0.6888 - acc: 0.9075 - val_loss: 0.6987 - val_acc: 0.9077
Epoch 135/1000
329s 658ms/step - loss: 0.6888 - acc: 0.9074 - val_loss: 0.7070 - val_acc: 0.9055
Epoch 136/1000
329s 657ms/step - loss: 0.6909 - acc: 0.9089 - val_loss: 0.6839 - val_acc: 0.9133
Epoch 137/1000
329s 657ms/step - loss: 0.6866 - acc: 0.9087 - val_loss: 0.6873 - val_acc: 0.9091
Epoch 138/1000
328s 657ms/step - loss: 0.6916 - acc: 0.9067 - val_loss: 0.6984 - val_acc: 0.9073
Epoch 139/1000
329s 658ms/step - loss: 0.6905 - acc: 0.9070 - val_loss: 0.6842 - val_acc: 0.9137
Epoch 140/1000
331s 662ms/step - loss: 0.6863 - acc: 0.9084 - val_loss: 0.6936 - val_acc: 0.9110
Epoch 141/1000
329s 657ms/step - loss: 0.6863 - acc: 0.9090 - val_loss: 0.7201 - val_acc: 0.9043
Epoch 142/1000
328s 657ms/step - loss: 0.6855 - acc: 0.9089 - val_loss: 0.6890 - val_acc: 0.9100
Epoch 143/1000
329s 657ms/step - loss: 0.6881 - acc: 0.9071 - val_loss: 0.6997 - val_acc: 0.9022
Epoch 144/1000
329s 657ms/step - loss: 0.6860 - acc: 0.9087 - val_loss: 0.6794 - val_acc: 0.9121
Epoch 145/1000
328s 657ms/step - loss: 0.6878 - acc: 0.9079 - val_loss: 0.6919 - val_acc: 0.9096
Epoch 146/1000
329s 657ms/step - loss: 0.6860 - acc: 0.9076 - val_loss: 0.6870 - val_acc: 0.9115
Epoch 147/1000
328s 657ms/step - loss: 0.6836 - acc: 0.9112 - val_loss: 0.6757 - val_acc: 0.9140
Epoch 148/1000
329s 658ms/step - loss: 0.6876 - acc: 0.9068 - val_loss: 0.6843 - val_acc: 0.9123
Epoch 149/1000
329s 657ms/step - loss: 0.6857 - acc: 0.9086 - val_loss: 0.6851 - val_acc: 0.9114
Epoch 150/1000
329s 659ms/step - loss: 0.6832 - acc: 0.9096 - val_loss: 0.6798 - val_acc: 0.9142
Epoch 151/1000
329s 659ms/step - loss: 0.6887 - acc: 0.9081 - val_loss: 0.6761 - val_acc: 0.9161
Epoch 152/1000
329s 658ms/step - loss: 0.6850 - acc: 0.9090 - val_loss: 0.7138 - val_acc: 0.9042
Epoch 153/1000
332s 664ms/step - loss: 0.6845 - acc: 0.9101 - val_loss: 0.6783 - val_acc: 0.9137
Epoch 154/1000
330s 659ms/step - loss: 0.6879 - acc: 0.9095 - val_loss: 0.6898 - val_acc: 0.9070
Epoch 155/1000
329s 657ms/step - loss: 0.6820 - acc: 0.9106 - val_loss: 0.6964 - val_acc: 0.9076
Epoch 156/1000
328s 657ms/step - loss: 0.6884 - acc: 0.9078 - val_loss: 0.6919 - val_acc: 0.9089
Epoch 157/1000
328s 657ms/step - loss: 0.6860 - acc: 0.9095 - val_loss: 0.6758 - val_acc: 0.9164
Epoch 158/1000
329s 658ms/step - loss: 0.6853 - acc: 0.9101 - val_loss: 0.6822 - val_acc: 0.9149
Epoch 159/1000
328s 657ms/step - loss: 0.6880 - acc: 0.9086 - val_loss: 0.6964 - val_acc: 0.9101
Epoch 160/1000
328s 657ms/step - loss: 0.6822 - acc: 0.9103 - val_loss: 0.6767 - val_acc: 0.9140
Epoch 161/1000
329s 657ms/step - loss: 0.6826 - acc: 0.9107 - val_loss: 0.6733 - val_acc: 0.9170
Epoch 162/1000
329s 658ms/step - loss: 0.6848 - acc: 0.9104 - val_loss: 0.7111 - val_acc: 0.9036
Epoch 163/1000
329s 657ms/step - loss: 0.6824 - acc: 0.9113 - val_loss: 0.6874 - val_acc: 0.9096
Epoch 164/1000
329s 657ms/step - loss: 0.6804 - acc: 0.9120 - val_loss: 0.6969 - val_acc: 0.9083
Epoch 165/1000
329s 657ms/step - loss: 0.6812 - acc: 0.9108 - val_loss: 0.6867 - val_acc: 0.9105
Epoch 166/1000
331s 663ms/step - loss: 0.6861 - acc: 0.9094 - val_loss: 0.7172 - val_acc: 0.9049
Epoch 167/1000
329s 658ms/step - loss: 0.6872 - acc: 0.9109 - val_loss: 0.7027 - val_acc: 0.9063
Epoch 168/1000
329s 658ms/step - loss: 0.6847 - acc: 0.9110 - val_loss: 0.6851 - val_acc: 0.9128
Epoch 169/1000
329s 657ms/step - loss: 0.6849 - acc: 0.9109 - val_loss: 0.6786 - val_acc: 0.9133
Epoch 170/1000
329s 657ms/step - loss: 0.6848 - acc: 0.9109 - val_loss: 0.6923 - val_acc: 0.9104
Epoch 171/1000
328s 657ms/step - loss: 0.6871 - acc: 0.9104 - val_loss: 0.7007 - val_acc: 0.9083
Epoch 172/1000
328s 657ms/step - loss: 0.6823 - acc: 0.9113 - val_loss: 0.6857 - val_acc: 0.9152
Epoch 173/1000
328s 657ms/step - loss: 0.6827 - acc: 0.9107 - val_loss: 0.6829 - val_acc: 0.9136
Epoch 174/1000
329s 657ms/step - loss: 0.6789 - acc: 0.9118 - val_loss: 0.6929 - val_acc: 0.9096
Epoch 175/1000
329s 658ms/step - loss: 0.6860 - acc: 0.9090 - val_loss: 0.6831 - val_acc: 0.9144
Epoch 176/1000
329s 657ms/step - loss: 0.6852 - acc: 0.9091 - val_loss: 0.6786 - val_acc: 0.9154
Epoch 177/1000
328s 657ms/step - loss: 0.6829 - acc: 0.9107 - val_loss: 0.6841 - val_acc: 0.9128
Epoch 178/1000
329s 657ms/step - loss: 0.6827 - acc: 0.9114 - val_loss: 0.6962 - val_acc: 0.9081
Epoch 179/1000
329s 658ms/step - loss: 0.6837 - acc: 0.9109 - val_loss: 0.6851 - val_acc: 0.9122
Epoch 180/1000
329s 657ms/step - loss: 0.6833 - acc: 0.9107 - val_loss: 0.6863 - val_acc: 0.9092
Epoch 181/1000
330s 659ms/step - loss: 0.6830 - acc: 0.9103 - val_loss: 0.7133 - val_acc: 0.9031
Epoch 182/1000
331s 661ms/step - loss: 0.6879 - acc: 0.9112 - val_loss: 0.6916 - val_acc: 0.9104
Epoch 183/1000
329s 657ms/step - loss: 0.6829 - acc: 0.9098 - val_loss: 0.6945 - val_acc: 0.9089
Epoch 184/1000
328s 657ms/step - loss: 0.6882 - acc: 0.9102 - val_loss: 0.6908 - val_acc: 0.9093
Epoch 185/1000
329s 657ms/step - loss: 0.6816 - acc: 0.9113 - val_loss: 0.6898 - val_acc: 0.9138
Epoch 186/1000
331s 662ms/step - loss: 0.6875 - acc: 0.9095 - val_loss: 0.6690 - val_acc: 0.9201
Epoch 187/1000
329s 659ms/step - loss: 0.6821 - acc: 0.9122 - val_loss: 0.6973 - val_acc: 0.9113
Epoch 188/1000
332s 665ms/step - loss: 0.6854 - acc: 0.9116 - val_loss: 0.6866 - val_acc: 0.9150
Epoch 189/1000
329s 657ms/step - loss: 0.6810 - acc: 0.9117 - val_loss: 0.6731 - val_acc: 0.9184
Epoch 190/1000
329s 657ms/step - loss: 0.6845 - acc: 0.9116 - val_loss: 0.6797 - val_acc: 0.9123
Epoch 191/1000
329s 658ms/step - loss: 0.6849 - acc: 0.9117 - val_loss: 0.6768 - val_acc: 0.9151
Epoch 192/1000
329s 657ms/step - loss: 0.6824 - acc: 0.9126 - val_loss: 0.6717 - val_acc: 0.9180
Epoch 193/1000
329s 657ms/step - loss: 0.6831 - acc: 0.9115 - val_loss: 0.6978 - val_acc: 0.9090
Epoch 194/1000
328s 657ms/step - loss: 0.6871 - acc: 0.9088 - val_loss: 0.6844 - val_acc: 0.9113
Epoch 195/1000
328s 657ms/step - loss: 0.6831 - acc: 0.9120 - val_loss: 0.6968 - val_acc: 0.9118
Epoch 196/1000
329s 657ms/step - loss: 0.6804 - acc: 0.9127 - val_loss: 0.6671 - val_acc: 0.9193
Epoch 197/1000
329s 657ms/step - loss: 0.6792 - acc: 0.9127 - val_loss: 0.6900 - val_acc: 0.9150
Epoch 198/1000
329s 657ms/step - loss: 0.6877 - acc: 0.9094 - val_loss: 0.6898 - val_acc: 0.9117
Epoch 199/1000
329s 657ms/step - loss: 0.6789 - acc: 0.9133 - val_loss: 0.6828 - val_acc: 0.9161
Epoch 200/1000
329s 657ms/step - loss: 0.6834 - acc: 0.9110 - val_loss: 0.6733 - val_acc: 0.9192
Epoch 201/1000
329s 657ms/step - loss: 0.6806 - acc: 0.9131 - val_loss: 0.6693 - val_acc: 0.9196
Epoch 202/1000
331s 662ms/step - loss: 0.6782 - acc: 0.9119 - val_loss: 0.6918 - val_acc: 0.9103
Epoch 203/1000
329s 658ms/step - loss: 0.6822 - acc: 0.9116 - val_loss: 0.7031 - val_acc: 0.9100
Epoch 204/1000
329s 658ms/step - loss: 0.6835 - acc: 0.9123 - val_loss: 0.6968 - val_acc: 0.9090
Epoch 205/1000
329s 658ms/step - loss: 0.6831 - acc: 0.9104 - val_loss: 0.6719 - val_acc: 0.9204
Epoch 206/1000
329s 658ms/step - loss: 0.6806 - acc: 0.9128 - val_loss: 0.6813 - val_acc: 0.9143
Epoch 207/1000
329s 657ms/step - loss: 0.6824 - acc: 0.9116 - val_loss: 0.6830 - val_acc: 0.9131
Epoch 208/1000
329s 658ms/step - loss: 0.6815 - acc: 0.9117 - val_loss: 0.6724 - val_acc: 0.9192
Epoch 209/1000
329s 658ms/step - loss: 0.6813 - acc: 0.9121 - val_loss: 0.6961 - val_acc: 0.9085
Epoch 210/1000
329s 658ms/step - loss: 0.6853 - acc: 0.9102 - val_loss: 0.6890 - val_acc: 0.9138
Epoch 211/1000
329s 658ms/step - loss: 0.6803 - acc: 0.9133 - val_loss: 0.6917 - val_acc: 0.9103
Epoch 212/1000
329s 658ms/step - loss: 0.6768 - acc: 0.9141 - val_loss: 0.6763 - val_acc: 0.9183
Epoch 213/1000
329s 658ms/step - loss: 0.6801 - acc: 0.9120 - val_loss: 0.6743 - val_acc: 0.9174
Epoch 214/1000
329s 658ms/step - loss: 0.6824 - acc: 0.9124 - val_loss: 0.6973 - val_acc: 0.9117
Epoch 215/1000
329s 657ms/step - loss: 0.6877 - acc: 0.9110 - val_loss: 0.6758 - val_acc: 0.9157
Epoch 216/1000
328s 657ms/step - loss: 0.6793 - acc: 0.9135 - val_loss: 0.6639 - val_acc: 0.9190
Epoch 217/1000
329s 657ms/step - loss: 0.6824 - acc: 0.9123 - val_loss: 0.6938 - val_acc: 0.9093
Epoch 218/1000
328s 657ms/step - loss: 0.6773 - acc: 0.9131 - val_loss: 0.6859 - val_acc: 0.9124
Epoch 219/1000
328s 657ms/step - loss: 0.6818 - acc: 0.9137 - val_loss: 0.7022 - val_acc: 0.9091
Epoch 220/1000
329s 657ms/step - loss: 0.6819 - acc: 0.9131 - val_loss: 0.6826 - val_acc: 0.9152
Epoch 221/1000
329s 659ms/step - loss: 0.6844 - acc: 0.9125 - val_loss: 0.6899 - val_acc: 0.9094
Epoch 222/1000
329s 658ms/step - loss: 0.6814 - acc: 0.9133 - val_loss: 0.6745 - val_acc: 0.9186
Epoch 223/1000
329s 658ms/step - loss: 0.6784 - acc: 0.9130 - val_loss: 0.7058 - val_acc: 0.9094
Epoch 224/1000
329s 658ms/step - loss: 0.6797 - acc: 0.9141 - val_loss: 0.6742 - val_acc: 0.9172
Epoch 225/1000
329s 658ms/step - loss: 0.6802 - acc: 0.9136 - val_loss: 0.6896 - val_acc: 0.9130
Epoch 226/1000
329s 658ms/step - loss: 0.6821 - acc: 0.9122 - val_loss: 0.6661 - val_acc: 0.9212
Epoch 227/1000
329s 657ms/step - loss: 0.6805 - acc: 0.9134 - val_loss: 0.6954 - val_acc: 0.9120
Epoch 228/1000
329s 657ms/step - loss: 0.6832 - acc: 0.9128 - val_loss: 0.6933 - val_acc: 0.9108
Epoch 229/1000
329s 658ms/step - loss: 0.6776 - acc: 0.9143 - val_loss: 0.6777 - val_acc: 0.9166
Epoch 230/1000
329s 658ms/step - loss: 0.6824 - acc: 0.9132 - val_loss: 0.7123 - val_acc: 0.9043
Epoch 231/1000
329s 658ms/step - loss: 0.6842 - acc: 0.9124 - val_loss: 0.6955 - val_acc: 0.9111
Epoch 232/1000
329s 657ms/step - loss: 0.6811 - acc: 0.9134 - val_loss: 0.6866 - val_acc: 0.9119
Epoch 233/1000
329s 658ms/step - loss: 0.6755 - acc: 0.9149 - val_loss: 0.7080 - val_acc: 0.9069
Epoch 234/1000
329s 658ms/step - loss: 0.6836 - acc: 0.9119 - val_loss: 0.6990 - val_acc: 0.9106
Epoch 235/1000
329s 658ms/step - loss: 0.6786 - acc: 0.9129 - val_loss: 0.6783 - val_acc: 0.9172
Epoch 236/1000
329s 658ms/step - loss: 0.6811 - acc: 0.9129 - val_loss: 0.6744 - val_acc: 0.9152
Epoch 237/1000
329s 658ms/step - loss: 0.6788 - acc: 0.9139 - val_loss: 0.6812 - val_acc: 0.9161
Epoch 238/1000
329s 658ms/step - loss: 0.6810 - acc: 0.9137 - val_loss: 0.6842 - val_acc: 0.9156
Epoch 239/1000
329s 658ms/step - loss: 0.6890 - acc: 0.9106 - val_loss: 0.6681 - val_acc: 0.9207
Epoch 240/1000
330s 660ms/step - loss: 0.6851 - acc: 0.9133 - val_loss: 0.6813 - val_acc: 0.9190
Epoch 241/1000
329s 659ms/step - loss: 0.6842 - acc: 0.9133 - val_loss: 0.6929 - val_acc: 0.9123
Epoch 242/1000
329s 658ms/step - loss: 0.6745 - acc: 0.9149 - val_loss: 0.6817 - val_acc: 0.9129
Epoch 243/1000
331s 662ms/step - loss: 0.6758 - acc: 0.9148 - val_loss: 0.6810 - val_acc: 0.9154
Epoch 244/1000
329s 658ms/step - loss: 0.6828 - acc: 0.9134 - val_loss: 0.6866 - val_acc: 0.9152
Epoch 245/1000
329s 659ms/step - loss: 0.6802 - acc: 0.9132 - val_loss: 0.6830 - val_acc: 0.9168
Epoch 246/1000
328s 657ms/step - loss: 0.6801 - acc: 0.9139 - val_loss: 0.6893 - val_acc: 0.9139
Epoch 247/1000
328s 657ms/step - loss: 0.6770 - acc: 0.9159 - val_loss: 0.6869 - val_acc: 0.9138
Epoch 248/1000
328s 657ms/step - loss: 0.6771 - acc: 0.9144 - val_loss: 0.6803 - val_acc: 0.9158
Epoch 249/1000
329s 657ms/step - loss: 0.6786 - acc: 0.9138 - val_loss: 0.6931 - val_acc: 0.9087
Epoch 250/1000
328s 657ms/step - loss: 0.6824 - acc: 0.9119 - val_loss: 0.6867 - val_acc: 0.9128
Epoch 251/1000
329s 657ms/step - loss: 0.6748 - acc: 0.9161 - val_loss: 0.6715 - val_acc: 0.9160
Epoch 252/1000
329s 658ms/step - loss: 0.6768 - acc: 0.9149 - val_loss: 0.6969 - val_acc: 0.9104
Epoch 253/1000
331s 663ms/step - loss: 0.6777 - acc: 0.9133 - val_loss: 0.6803 - val_acc: 0.9169
Epoch 254/1000
329s 657ms/step - loss: 0.6785 - acc: 0.9152 - val_loss: 0.6824 - val_acc: 0.9167
Epoch 255/1000
329s 657ms/step - loss: 0.6842 - acc: 0.9134 - val_loss: 0.6889 - val_acc: 0.9136
Epoch 256/1000
328s 657ms/step - loss: 0.6805 - acc: 0.9144 - val_loss: 0.6850 - val_acc: 0.9122
Epoch 257/1000
329s 657ms/step - loss: 0.6799 - acc: 0.9143 - val_loss: 0.6681 - val_acc: 0.9209
Epoch 258/1000
329s 657ms/step - loss: 0.6744 - acc: 0.9152 - val_loss: 0.6823 - val_acc: 0.9170
Epoch 259/1000
328s 657ms/step - loss: 0.6804 - acc: 0.9139 - val_loss: 0.6833 - val_acc: 0.9150
Epoch 260/1000
332s 663ms/step - loss: 0.6776 - acc: 0.9148 - val_loss: 0.6851 - val_acc: 0.9137
Epoch 261/1000
329s 657ms/step - loss: 0.6786 - acc: 0.9142 - val_loss: 0.6813 - val_acc: 0.9166
Epoch 262/1000
328s 657ms/step - loss: 0.6834 - acc: 0.9129 - val_loss: 0.6840 - val_acc: 0.9174
Epoch 263/1000
328s 657ms/step - loss: 0.6761 - acc: 0.9163 - val_loss: 0.6774 - val_acc: 0.9140
Epoch 264/1000
328s 657ms/step - loss: 0.6754 - acc: 0.9142 - val_loss: 0.6793 - val_acc: 0.9166
Epoch 265/1000
332s 663ms/step - loss: 0.6817 - acc: 0.9137 - val_loss: 0.7218 - val_acc: 0.9043
Epoch 266/1000
329s 657ms/step - loss: 0.6812 - acc: 0.9142 - val_loss: 0.6807 - val_acc: 0.9142
Epoch 267/1000
328s 656ms/step - loss: 0.6719 - acc: 0.9164 - val_loss: 0.6801 - val_acc: 0.9183
Epoch 268/1000
330s 659ms/step - loss: 0.6768 - acc: 0.9142 - val_loss: 0.6874 - val_acc: 0.9134
Epoch 269/1000
329s 658ms/step - loss: 0.6747 - acc: 0.9157 - val_loss: 0.6954 - val_acc: 0.9114
Epoch 270/1000
329s 658ms/step - loss: 0.6737 - acc: 0.9150 - val_loss: 0.6955 - val_acc: 0.9102
Epoch 271/1000
328s 656ms/step - loss: 0.6770 - acc: 0.9148 - val_loss: 0.6853 - val_acc: 0.9104
Epoch 272/1000
328s 656ms/step - loss: 0.6785 - acc: 0.9139 - val_loss: 0.6960 - val_acc: 0.9134
Epoch 273/1000
328s 656ms/step - loss: 0.6830 - acc: 0.9136 - val_loss: 0.6851 - val_acc: 0.9141
Epoch 274/1000
328s 656ms/step - loss: 0.6784 - acc: 0.9143 - val_loss: 0.6921 - val_acc: 0.9119
Epoch 275/1000
328s 656ms/step - loss: 0.6786 - acc: 0.9135 - val_loss: 0.6974 - val_acc: 0.9074
Epoch 276/1000
329s 659ms/step - loss: 0.6739 - acc: 0.9164 - val_loss: 0.6864 - val_acc: 0.9137
Epoch 277/1000
329s 658ms/step - loss: 0.6805 - acc: 0.9154 - val_loss: 0.7002 - val_acc: 0.9091
Epoch 278/1000
329s 658ms/step - loss: 0.6839 - acc: 0.9110 - val_loss: 0.6873 - val_acc: 0.9170
Epoch 279/1000
329s 657ms/step - loss: 0.6806 - acc: 0.9142 - val_loss: 0.6887 - val_acc: 0.9139
Epoch 280/1000
329s 658ms/step - loss: 0.6830 - acc: 0.9149 - val_loss: 0.6858 - val_acc: 0.9136
Epoch 281/1000
329s 658ms/step - loss: 0.6815 - acc: 0.9150 - val_loss: 0.7072 - val_acc: 0.9102
Epoch 282/1000
329s 658ms/step - loss: 0.6824 - acc: 0.9150 - val_loss: 0.6886 - val_acc: 0.9141
Epoch 283/1000
329s 658ms/step - loss: 0.6759 - acc: 0.9165 - val_loss: 0.6684 - val_acc: 0.9154
Epoch 284/1000
329s 658ms/step - loss: 0.6777 - acc: 0.9159 - val_loss: 0.6890 - val_acc: 0.9133
Epoch 285/1000
329s 658ms/step - loss: 0.6739 - acc: 0.9146 - val_loss: 0.6872 - val_acc: 0.9134
Epoch 286/1000
329s 657ms/step - loss: 0.6761 - acc: 0.9160 - val_loss: 0.6999 - val_acc: 0.9128
Epoch 287/1000
329s 658ms/step - loss: 0.6830 - acc: 0.9131 - val_loss: 0.7030 - val_acc: 0.9074
Epoch 288/1000
329s 659ms/step - loss: 0.6833 - acc: 0.9146 - val_loss: 0.6731 - val_acc: 0.9201
Epoch 289/1000
329s 657ms/step - loss: 0.6811 - acc: 0.9156 - val_loss: 0.6852 - val_acc: 0.9137
Epoch 290/1000
329s 658ms/step - loss: 0.6783 - acc: 0.9158 - val_loss: 0.6965 - val_acc: 0.9127
Epoch 291/1000
329s 658ms/step - loss: 0.6793 - acc: 0.9147 - val_loss: 0.7039 - val_acc: 0.9103
Epoch 292/1000
329s 657ms/step - loss: 0.6802 - acc: 0.9141 - val_loss: 0.6988 - val_acc: 0.9117
Epoch 293/1000
329s 658ms/step - loss: 0.6731 - acc: 0.9181 - val_loss: 0.6805 - val_acc: 0.9176
Epoch 294/1000
329s 657ms/step - loss: 0.6777 - acc: 0.9160 - val_loss: 0.6879 - val_acc: 0.9152
Epoch 295/1000
329s 658ms/step - loss: 0.6797 - acc: 0.9135 - val_loss: 0.7164 - val_acc: 0.9047
Epoch 296/1000
329s 658ms/step - loss: 0.6779 - acc: 0.9155 - val_loss: 0.7147 - val_acc: 0.9053
Epoch 297/1000
329s 658ms/step - loss: 0.6819 - acc: 0.9145 - val_loss: 0.6997 - val_acc: 0.9125
Epoch 298/1000
329s 658ms/step - loss: 0.6778 - acc: 0.9159 - val_loss: 0.6868 - val_acc: 0.9186
Epoch 299/1000
329s 658ms/step - loss: 0.6737 - acc: 0.9162 - val_loss: 0.7111 - val_acc: 0.9053
Epoch 300/1000
329s 658ms/step - loss: 0.6785 - acc: 0.9150 - val_loss: 0.6837 - val_acc: 0.9176
Epoch 301/1000
lr changed to 0.010000000149011612
329s 657ms/step - loss: 0.5770 - acc: 0.9504 - val_loss: 0.5992 - val_acc: 0.9421
Epoch 302/1000
329s 657ms/step - loss: 0.5208 - acc: 0.9675 - val_loss: 0.5883 - val_acc: 0.9469
Epoch 303/1000
329s 657ms/step - loss: 0.4996 - acc: 0.9727 - val_loss: 0.5778 - val_acc: 0.9466
Epoch 304/1000
329s 658ms/step - loss: 0.4859 - acc: 0.9737 - val_loss: 0.5660 - val_acc: 0.9480
Epoch 305/1000
329s 657ms/step - loss: 0.4716 - acc: 0.9773 - val_loss: 0.5623 - val_acc: 0.9476
Epoch 306/1000
329s 657ms/step - loss: 0.4561 - acc: 0.9802 - val_loss: 0.5534 - val_acc: 0.9496
Epoch 307/1000
329s 657ms/step - loss: 0.4473 - acc: 0.9807 - val_loss: 0.5434 - val_acc: 0.9488
Epoch 308/1000
329s 657ms/step - loss: 0.4373 - acc: 0.9814 - val_loss: 0.5413 - val_acc: 0.9494
Epoch 309/1000
329s 658ms/step - loss: 0.4271 - acc: 0.9827 - val_loss: 0.5359 - val_acc: 0.9481
Epoch 310/1000
329s 657ms/step - loss: 0.4173 - acc: 0.9838 - val_loss: 0.5266 - val_acc: 0.9509
Epoch 311/1000
329s 657ms/step - loss: 0.4078 - acc: 0.9856 - val_loss: 0.5195 - val_acc: 0.9496
Epoch 312/1000
329s 657ms/step - loss: 0.4022 - acc: 0.9846 - val_loss: 0.5200 - val_acc: 0.9493
Epoch 313/1000
329s 658ms/step - loss: 0.3934 - acc: 0.9860 - val_loss: 0.5136 - val_acc: 0.9505
Epoch 314/1000
329s 657ms/step - loss: 0.3863 - acc: 0.9862 - val_loss: 0.5125 - val_acc: 0.9480
Epoch 315/1000
329s 657ms/step - loss: 0.3784 - acc: 0.9875 - val_loss: 0.5041 - val_acc: 0.9503
Epoch 316/1000
329s 657ms/step - loss: 0.3740 - acc: 0.9866 - val_loss: 0.5026 - val_acc: 0.9485
Epoch 317/1000
328s 657ms/step - loss: 0.3689 - acc: 0.9867 - val_loss: 0.4907 - val_acc: 0.9518
Epoch 318/1000
329s 657ms/step - loss: 0.3607 - acc: 0.9878 - val_loss: 0.4934 - val_acc: 0.9506
Epoch 319/1000
330s 659ms/step - loss: 0.3558 - acc: 0.9873 - val_loss: 0.4836 - val_acc: 0.9503
Epoch 320/1000
329s 657ms/step - loss: 0.3465 - acc: 0.9890 - val_loss: 0.4787 - val_acc: 0.9514
Epoch 321/1000
328s 657ms/step - loss: 0.3427 - acc: 0.9891 - val_loss: 0.4784 - val_acc: 0.9509
Epoch 322/1000
331s 663ms/step - loss: 0.3382 - acc: 0.9890 - val_loss: 0.4716 - val_acc: 0.9506
Epoch 323/1000
329s 658ms/step - loss: 0.3314 - acc: 0.9894 - val_loss: 0.4708 - val_acc: 0.9509
Epoch 324/1000
329s 657ms/step - loss: 0.3292 - acc: 0.9892 - val_loss: 0.4612 - val_acc: 0.9534
Epoch 325/1000
329s 657ms/step - loss: 0.3212 - acc: 0.9900 - val_loss: 0.4520 - val_acc: 0.9509
Epoch 326/1000
328s 657ms/step - loss: 0.3152 - acc: 0.9903 - val_loss: 0.4612 - val_acc: 0.9474
Epoch 327/1000
332s 663ms/step - loss: 0.3132 - acc: 0.9898 - val_loss: 0.4641 - val_acc: 0.9496
...
Epoch 577/1000
329s 658ms/step - loss: 0.1779 - acc: 0.9889 - val_loss: 0.3452 - val_acc: 0.9418
Epoch 578/1000
329s 657ms/step - loss: 0.1758 - acc: 0.9896 - val_loss: 0.3450 - val_acc: 0.9440
Epoch 579/1000
329s 657ms/step - loss: 0.1775 - acc: 0.9887 - val_loss: 0.3361 - val_acc: 0.9458
Epoch 580/1000
329s 658ms/step - loss: 0.1750 - acc: 0.9899 - val_loss: 0.3418 - val_acc: 0.9479
Epoch 581/1000
329s 658ms/step - loss: 0.1721 - acc: 0.9912 - val_loss: 0.3549 - val_acc: 0.9425
Epoch 582/1000
329s 657ms/step - loss: 0.1765 - acc: 0.9894 - val_loss: 0.3358 - val_acc: 0.9446
Epoch 583/1000
329s 658ms/step - loss: 0.1760 - acc: 0.9897 - val_loss: 0.3409 - val_acc: 0.9451
Epoch 584/1000
329s 658ms/step - loss: 0.1755 - acc: 0.9893 - val_loss: 0.3453 - val_acc: 0.9425
Epoch 585/1000
329s 658ms/step - loss: 0.1788 - acc: 0.9884 - val_loss: 0.3428 - val_acc: 0.9421
Epoch 586/1000
329s 658ms/step - loss: 0.1772 - acc: 0.9885 - val_loss: 0.3503 - val_acc: 0.9408
Epoch 587/1000
329s 657ms/step - loss: 0.1774 - acc: 0.9885 - val_loss: 0.3441 - val_acc: 0.9444
Epoch 588/1000
329s 657ms/step - loss: 0.1730 - acc: 0.9909 - val_loss: 0.3379 - val_acc: 0.9461
Epoch 589/1000
329s 658ms/step - loss: 0.1787 - acc: 0.9879 - val_loss: 0.3538 - val_acc: 0.9408
Epoch 590/1000
329s 658ms/step - loss: 0.1769 - acc: 0.9891 - val_loss: 0.3490 - val_acc: 0.9435
Epoch 591/1000
329s 658ms/step - loss: 0.1751 - acc: 0.9898 - val_loss: 0.3517 - val_acc: 0.9405
Epoch 592/1000
329s 658ms/step - loss: 0.1757 - acc: 0.9897 - val_loss: 0.3406 - val_acc: 0.9428
Epoch 593/1000
329s 658ms/step - loss: 0.1771 - acc: 0.9891 - val_loss: 0.3388 - val_acc: 0.9441
Epoch 594/1000
329s 658ms/step - loss: 0.1757 - acc: 0.9898 - val_loss: 0.3501 - val_acc: 0.9415
Epoch 595/1000
329s 657ms/step - loss: 0.1755 - acc: 0.9896 - val_loss: 0.3473 - val_acc: 0.9416
Epoch 596/1000
329s 658ms/step - loss: 0.1728 - acc: 0.9901 - val_loss: 0.3366 - val_acc: 0.9443
Epoch 597/1000
329s 657ms/step - loss: 0.1748 - acc: 0.9896 - val_loss: 0.3571 - val_acc: 0.9413
Epoch 598/1000
329s 658ms/step - loss: 0.1767 - acc: 0.9889 - val_loss: 0.3485 - val_acc: 0.9421
Epoch 599/1000
329s 658ms/step - loss: 0.1778 - acc: 0.9887 - val_loss: 0.3409 - val_acc: 0.9422
Epoch 600/1000
329s 658ms/step - loss: 0.1759 - acc: 0.9894 - val_loss: 0.3471 - val_acc: 0.9428
Epoch 601/1000
lr changed to 0.0009999999776482583
329s 657ms/step - loss: 0.1637 - acc: 0.9939 - val_loss: 0.3166 - val_acc: 0.9501
Epoch 602/1000
329s 657ms/step - loss: 0.1559 - acc: 0.9966 - val_loss: 0.3119 - val_acc: 0.9506
Epoch 603/1000
329s 657ms/step - loss: 0.1527 - acc: 0.9974 - val_loss: 0.3101 - val_acc: 0.9506
Epoch 604/1000
329s 658ms/step - loss: 0.1519 - acc: 0.9974 - val_loss: 0.3096 - val_acc: 0.9528
Epoch 605/1000
329s 658ms/step - loss: 0.1508 - acc: 0.9977 - val_loss: 0.3065 - val_acc: 0.9534
Epoch 606/1000
329s 658ms/step - loss: 0.1498 - acc: 0.9977 - val_loss: 0.3029 - val_acc: 0.9537
Epoch 607/1000
329s 657ms/step - loss: 0.1486 - acc: 0.9983 - val_loss: 0.3023 - val_acc: 0.9550
Epoch 608/1000
329s 658ms/step - loss: 0.1479 - acc: 0.9984 - val_loss: 0.3038 - val_acc: 0.9547
Epoch 609/1000
329s 658ms/step - loss: 0.1481 - acc: 0.9982 - val_loss: 0.3042 - val_acc: 0.9552
Epoch 610/1000
329s 658ms/step - loss: 0.1477 - acc: 0.9984 - val_loss: 0.3046 - val_acc: 0.9551
Epoch 611/1000
329s 658ms/step - loss: 0.1469 - acc: 0.9985 - val_loss: 0.3042 - val_acc: 0.9545
Epoch 612/1000
329s 657ms/step - loss: 0.1466 - acc: 0.9984 - val_loss: 0.3041 - val_acc: 0.9544
Epoch 613/1000
329s 658ms/step - loss: 0.1461 - acc: 0.9983 - val_loss: 0.3028 - val_acc: 0.9550
Epoch 614/1000
329s 658ms/step - loss: 0.1454 - acc: 0.9988 - val_loss: 0.3013 - val_acc: 0.9547
Epoch 615/1000
329s 657ms/step - loss: 0.1443 - acc: 0.9989 - val_loss: 0.3034 - val_acc: 0.9545
Epoch 616/1000
329s 657ms/step - loss: 0.1449 - acc: 0.9985 - val_loss: 0.3020 - val_acc: 0.9547
Epoch 617/1000
329s 658ms/step - loss: 0.1442 - acc: 0.9989 - val_loss: 0.3005 - val_acc: 0.9547
Epoch 618/1000
329s 658ms/step - loss: 0.1442 - acc: 0.9987 - val_loss: 0.3007 - val_acc: 0.9550
Epoch 619/1000
329s 657ms/step - loss: 0.1433 - acc: 0.9990 - val_loss: 0.3015 - val_acc: 0.9558
Epoch 620/1000
329s 658ms/step - loss: 0.1436 - acc: 0.9990 - val_loss: 0.3018 - val_acc: 0.9557
Epoch 621/1000
329s 658ms/step - loss: 0.1430 - acc: 0.9990 - val_loss: 0.3001 - val_acc: 0.9552
Epoch 622/1000
329s 657ms/step - loss: 0.1427 - acc: 0.9990 - val_loss: 0.2994 - val_acc: 0.9552
Epoch 623/1000
329s 658ms/step - loss: 0.1423 - acc: 0.9989 - val_loss: 0.2989 - val_acc: 0.9548
Epoch 624/1000
329s 658ms/step - loss: 0.1419 - acc: 0.9991 - val_loss: 0.2995 - val_acc: 0.9553
Epoch 625/1000
329s 658ms/step - loss: 0.1414 - acc: 0.9990 - val_loss: 0.3002 - val_acc: 0.9558
Epoch 626/1000
329s 657ms/step - loss: 0.1413 - acc: 0.9990 - val_loss: 0.3006 - val_acc: 0.9551
Epoch 627/1000
329s 657ms/step - loss: 0.1409 - acc: 0.9991 - val_loss: 0.2997 - val_acc: 0.9548
Epoch 628/1000
329s 657ms/step - loss: 0.1405 - acc: 0.9993 - val_loss: 0.3015 - val_acc: 0.9542
Epoch 629/1000
329s 657ms/step - loss: 0.1409 - acc: 0.9988 - val_loss: 0.3005 - val_acc: 0.9548
Epoch 630/1000
329s 658ms/step - loss: 0.1404 - acc: 0.9989 - val_loss: 0.3032 - val_acc: 0.9534
Epoch 631/1000
329s 658ms/step - loss: 0.1400 - acc: 0.9991 - val_loss: 0.3029 - val_acc: 0.9541
Epoch 632/1000
329s 658ms/step - loss: 0.1391 - acc: 0.9993 - val_loss: 0.3021 - val_acc: 0.9540
Epoch 633/1000
329s 658ms/step - loss: 0.1392 - acc: 0.9991 - val_loss: 0.3025 - val_acc: 0.9535
Epoch 634/1000
329s 657ms/step - loss: 0.1388 - acc: 0.9992 - val_loss: 0.3026 - val_acc: 0.9539
Epoch 635/1000
329s 658ms/step - loss: 0.1386 - acc: 0.9993 - val_loss: 0.3009 - val_acc: 0.9538
Epoch 636/1000
329s 658ms/step - loss: 0.1379 - acc: 0.9993 - val_loss: 0.3001 - val_acc: 0.9541
Epoch 637/1000
329s 658ms/step - loss: 0.1376 - acc: 0.9994 - val_loss: 0.3000 - val_acc: 0.9550
...
Epoch 893/1000
334s 667ms/step - loss: 0.0918 - acc: 0.9997 - val_loss: 0.2585 - val_acc: 0.9570
Epoch 894/1000
334s 667ms/step - loss: 0.0919 - acc: 0.9997 - val_loss: 0.2587 - val_acc: 0.9571
Epoch 895/1000
334s 667ms/step - loss: 0.0913 - acc: 0.9998 - val_loss: 0.2579 - val_acc: 0.9568
Epoch 896/1000
334s 668ms/step - loss: 0.0916 - acc: 0.9996 - val_loss: 0.2551 - val_acc: 0.9581
Epoch 897/1000
334s 667ms/step - loss: 0.0915 - acc: 0.9995 - val_loss: 0.2544 - val_acc: 0.9578
Epoch 898/1000
334s 667ms/step - loss: 0.0919 - acc: 0.9994 - val_loss: 0.2601 - val_acc: 0.9558
Epoch 899/1000
333s 667ms/step - loss: 0.0914 - acc: 0.9996 - val_loss: 0.2554 - val_acc: 0.9575
Epoch 900/1000
334s 667ms/step - loss: 0.0909 - acc: 0.9995 - val_loss: 0.2557 - val_acc: 0.9576
Epoch 901/1000
lr changed to 9.999999310821295e-05
333s 667ms/step - loss: 0.0910 - acc: 0.9995 - val_loss: 0.2555 - val_acc: 0.9575
Epoch 902/1000
330s 660ms/step - loss: 0.0911 - acc: 0.9995 - val_loss: 0.2548 - val_acc: 0.9576
Epoch 903/1000
327s 654ms/step - loss: 0.0912 - acc: 0.9995 - val_loss: 0.2552 - val_acc: 0.9576
Epoch 904/1000
327s 653ms/step - loss: 0.0910 - acc: 0.9995 - val_loss: 0.2545 - val_acc: 0.9572
Epoch 905/1000
327s 654ms/step - loss: 0.0904 - acc: 0.9998 - val_loss: 0.2548 - val_acc: 0.9575
Epoch 906/1000
327s 653ms/step - loss: 0.0905 - acc: 0.9998 - val_loss: 0.2549 - val_acc: 0.9572
Epoch 907/1000
327s 654ms/step - loss: 0.0908 - acc: 0.9995 - val_loss: 0.2546 - val_acc: 0.9573
Epoch 908/1000
327s 654ms/step - loss: 0.0908 - acc: 0.9995 - val_loss: 0.2542 - val_acc: 0.9573
Epoch 909/1000
327s 654ms/step - loss: 0.0910 - acc: 0.9994 - val_loss: 0.2539 - val_acc: 0.9578
Epoch 910/1000
327s 654ms/step - loss: 0.0905 - acc: 0.9996 - val_loss: 0.2538 - val_acc: 0.9575
Epoch 911/1000
327s 653ms/step - loss: 0.0908 - acc: 0.9996 - val_loss: 0.2537 - val_acc: 0.9574
Epoch 912/1000
327s 654ms/step - loss: 0.0906 - acc: 0.9996 - val_loss: 0.2534 - val_acc: 0.9573
Epoch 913/1000
327s 653ms/step - loss: 0.0903 - acc: 0.9998 - val_loss: 0.2538 - val_acc: 0.9575
Epoch 914/1000
331s 662ms/step - loss: 0.0905 - acc: 0.9996 - val_loss: 0.2542 - val_acc: 0.9577
Epoch 915/1000
333s 667ms/step - loss: 0.0904 - acc: 0.9997 - val_loss: 0.2544 - val_acc: 0.9572
Epoch 916/1000
331s 662ms/step - loss: 0.0905 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9574
Epoch 917/1000
327s 654ms/step - loss: 0.0905 - acc: 0.9995 - val_loss: 0.2543 - val_acc: 0.9576
Epoch 918/1000
327s 653ms/step - loss: 0.0901 - acc: 0.9998 - val_loss: 0.2541 - val_acc: 0.9578
Epoch 919/1000
327s 654ms/step - loss: 0.0904 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9578
Epoch 920/1000
327s 653ms/step - loss: 0.0902 - acc: 0.9998 - val_loss: 0.2545 - val_acc: 0.9575
Epoch 921/1000
327s 654ms/step - loss: 0.0903 - acc: 0.9997 - val_loss: 0.2544 - val_acc: 0.9578
Epoch 922/1000
327s 653ms/step - loss: 0.0903 - acc: 0.9996 - val_loss: 0.2542 - val_acc: 0.9575
Epoch 923/1000
327s 654ms/step - loss: 0.0903 - acc: 0.9997 - val_loss: 0.2544 - val_acc: 0.9577
Epoch 924/1000
327s 654ms/step - loss: 0.0905 - acc: 0.9996 - val_loss: 0.2548 - val_acc: 0.9573
Epoch 925/1000
327s 654ms/step - loss: 0.0904 - acc: 0.9997 - val_loss: 0.2547 - val_acc: 0.9574
Epoch 926/1000
327s 654ms/step - loss: 0.0903 - acc: 0.9997 - val_loss: 0.2546 - val_acc: 0.9575
Epoch 927/1000
327s 653ms/step - loss: 0.0905 - acc: 0.9996 - val_loss: 0.2545 - val_acc: 0.9576
Epoch 928/1000
327s 654ms/step - loss: 0.0901 - acc: 0.9998 - val_loss: 0.2543 - val_acc: 0.9576
Epoch 929/1000
327s 654ms/step - loss: 0.0901 - acc: 0.9998 - val_loss: 0.2544 - val_acc: 0.9579
Epoch 930/1000
327s 654ms/step - loss: 0.0902 - acc: 0.9997 - val_loss: 0.2538 - val_acc: 0.9577
Epoch 931/1000
327s 653ms/step - loss: 0.0904 - acc: 0.9995 - val_loss: 0.2539 - val_acc: 0.9576
Epoch 932/1000
327s 655ms/step - loss: 0.0901 - acc: 0.9997 - val_loss: 0.2535 - val_acc: 0.9577
Epoch 933/1000
327s 654ms/step - loss: 0.0902 - acc: 0.9997 - val_loss: 0.2541 - val_acc: 0.9578
Epoch 934/1000
327s 654ms/step - loss: 0.0899 - acc: 0.9998 - val_loss: 0.2541 - val_acc: 0.9579
Epoch 935/1000
327s 654ms/step - loss: 0.0907 - acc: 0.9995 - val_loss: 0.2542 - val_acc: 0.9579
Epoch 936/1000
327s 654ms/step - loss: 0.0900 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9576
Epoch 937/1000
327s 655ms/step - loss: 0.0900 - acc: 0.9998 - val_loss: 0.2544 - val_acc: 0.9575
Epoch 938/1000
328s 655ms/step - loss: 0.0901 - acc: 0.9997 - val_loss: 0.2544 - val_acc: 0.9577
Epoch 939/1000
329s 659ms/step - loss: 0.0900 - acc: 0.9997 - val_loss: 0.2544 - val_acc: 0.9580
Epoch 940/1000
328s 657ms/step - loss: 0.0901 - acc: 0.9996 - val_loss: 0.2544 - val_acc: 0.9577
Epoch 941/1000
328s 657ms/step - loss: 0.0900 - acc: 0.9997 - val_loss: 0.2541 - val_acc: 0.9575
Epoch 942/1000
333s 666ms/step - loss: 0.0900 - acc: 0.9996 - val_loss: 0.2536 - val_acc: 0.9577
Epoch 943/1000
333s 667ms/step - loss: 0.0901 - acc: 0.9996 - val_loss: 0.2536 - val_acc: 0.9575
Epoch 944/1000
329s 657ms/step - loss: 0.0896 - acc: 0.9998 - val_loss: 0.2536 - val_acc: 0.9575
Epoch 945/1000
326s 653ms/step - loss: 0.0899 - acc: 0.9998 - val_loss: 0.2537 - val_acc: 0.9577
Epoch 946/1000
327s 654ms/step - loss: 0.0899 - acc: 0.9997 - val_loss: 0.2535 - val_acc: 0.9577
Epoch 947/1000
326s 652ms/step - loss: 0.0901 - acc: 0.9996 - val_loss: 0.2533 - val_acc: 0.9580
Epoch 948/1000
326s 652ms/step - loss: 0.0899 - acc: 0.9996 - val_loss: 0.2529 - val_acc: 0.9576
Epoch 949/1000
326s 652ms/step - loss: 0.0901 - acc: 0.9997 - val_loss: 0.2530 - val_acc: 0.9582
Epoch 950/1000
326s 653ms/step - loss: 0.0898 - acc: 0.9998 - val_loss: 0.2528 - val_acc: 0.9578
Epoch 951/1000
326s 652ms/step - loss: 0.0897 - acc: 0.9998 - val_loss: 0.2531 - val_acc: 0.9581
Epoch 952/1000
326s 653ms/step - loss: 0.0894 - acc: 0.9999 - val_loss: 0.2533 - val_acc: 0.9584
Epoch 953/1000
327s 654ms/step - loss: 0.0900 - acc: 0.9997 - val_loss: 0.2534 - val_acc: 0.9580
Epoch 954/1000
327s 654ms/step - loss: 0.0896 - acc: 0.9998 - val_loss: 0.2535 - val_acc: 0.9583
Epoch 955/1000
327s 654ms/step - loss: 0.0900 - acc: 0.9996 - val_loss: 0.2534 - val_acc: 0.9579
Epoch 956/1000
327s 653ms/step - loss: 0.0896 - acc: 0.9998 - val_loss: 0.2533 - val_acc: 0.9581
Epoch 957/1000
327s 654ms/step - loss: 0.0898 - acc: 0.9997 - val_loss: 0.2528 - val_acc: 0.9581
Epoch 958/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9998 - val_loss: 0.2530 - val_acc: 0.9580
Epoch 959/1000
327s 654ms/step - loss: 0.0897 - acc: 0.9997 - val_loss: 0.2537 - val_acc: 0.9582
Epoch 960/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2535 - val_acc: 0.9578
Epoch 961/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9998 - val_loss: 0.2533 - val_acc: 0.9580
Epoch 962/1000
327s 654ms/step - loss: 0.0897 - acc: 0.9997 - val_loss: 0.2529 - val_acc: 0.9579
Epoch 963/1000
327s 653ms/step - loss: 0.0899 - acc: 0.9996 - val_loss: 0.2534 - val_acc: 0.9576
Epoch 964/1000
327s 653ms/step - loss: 0.0896 - acc: 0.9997 - val_loss: 0.2536 - val_acc: 0.9580
Epoch 965/1000
327s 653ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2535 - val_acc: 0.9583
Epoch 966/1000
327s 654ms/step - loss: 0.0898 - acc: 0.9997 - val_loss: 0.2538 - val_acc: 0.9581
Epoch 967/1000
327s 654ms/step - loss: 0.0894 - acc: 0.9997 - val_loss: 0.2536 - val_acc: 0.9582
Epoch 968/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9998 - val_loss: 0.2538 - val_acc: 0.9583
Epoch 969/1000
327s 654ms/step - loss: 0.0896 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9580
Epoch 970/1000
327s 654ms/step - loss: 0.0896 - acc: 0.9996 - val_loss: 0.2545 - val_acc: 0.9584
Epoch 971/1000
327s 655ms/step - loss: 0.0898 - acc: 0.9997 - val_loss: 0.2545 - val_acc: 0.9582
Epoch 972/1000
327s 654ms/step - loss: 0.0897 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9577
Epoch 973/1000
327s 654ms/step - loss: 0.0896 - acc: 0.9996 - val_loss: 0.2541 - val_acc: 0.9581
Epoch 974/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9583
Epoch 975/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2541 - val_acc: 0.9580
Epoch 976/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9580
Epoch 977/1000
327s 654ms/step - loss: 0.0894 - acc: 0.9997 - val_loss: 0.2542 - val_acc: 0.9576
Epoch 978/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9576
Epoch 979/1000
327s 653ms/step - loss: 0.0894 - acc: 0.9997 - val_loss: 0.2546 - val_acc: 0.9576
Epoch 980/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2549 - val_acc: 0.9579
Epoch 981/1000
327s 653ms/step - loss: 0.0896 - acc: 0.9996 - val_loss: 0.2551 - val_acc: 0.9579
Epoch 982/1000
327s 654ms/step - loss: 0.0899 - acc: 0.9996 - val_loss: 0.2546 - val_acc: 0.9584
Epoch 983/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9996 - val_loss: 0.2548 - val_acc: 0.9584
Epoch 984/1000
327s 653ms/step - loss: 0.0894 - acc: 0.9997 - val_loss: 0.2545 - val_acc: 0.9583
Epoch 985/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9997 - val_loss: 0.2545 - val_acc: 0.9581
Epoch 986/1000
327s 654ms/step - loss: 0.0895 - acc: 0.9996 - val_loss: 0.2540 - val_acc: 0.9577
Epoch 987/1000
328s 656ms/step - loss: 0.0894 - acc: 0.9998 - val_loss: 0.2537 - val_acc: 0.9575
Epoch 988/1000
327s 653ms/step - loss: 0.0892 - acc: 0.9998 - val_loss: 0.2540 - val_acc: 0.9579
Epoch 989/1000
327s 653ms/step - loss: 0.0894 - acc: 0.9997 - val_loss: 0.2541 - val_acc: 0.9578
Epoch 990/1000
327s 654ms/step - loss: 0.0893 - acc: 0.9997 - val_loss: 0.2539 - val_acc: 0.9581
Epoch 991/1000
327s 654ms/step - loss: 0.0892 - acc: 0.9997 - val_loss: 0.2543 - val_acc: 0.9578
Epoch 992/1000
327s 654ms/step - loss: 0.0890 - acc: 0.9998 - val_loss: 0.2541 - val_acc: 0.9579
Epoch 993/1000
327s 654ms/step - loss: 0.0892 - acc: 0.9998 - val_loss: 0.2541 - val_acc: 0.9583
Epoch 994/1000
327s 653ms/step - loss: 0.0892 - acc: 0.9997 - val_loss: 0.2542 - val_acc: 0.9580
Epoch 995/1000
326s 653ms/step - loss: 0.0893 - acc: 0.9998 - val_loss: 0.2548 - val_acc: 0.9577
Epoch 996/1000
327s 654ms/step - loss: 0.0892 - acc: 0.9997 - val_loss: 0.2547 - val_acc: 0.9580
Epoch 997/1000
327s 653ms/step - loss: 0.0895 - acc: 0.9996 - val_loss: 0.2540 - val_acc: 0.9582
Epoch 998/1000
326s 652ms/step - loss: 0.0891 - acc: 0.9998 - val_loss: 0.2544 - val_acc: 0.9580
Epoch 999/1000
326s 652ms/step - loss: 0.0890 - acc: 0.9998 - val_loss: 0.2538 - val_acc: 0.9582
Epoch 1000/1000
326s 653ms/step - loss: 0.0894 - acc: 0.9997 - val_loss: 0.2538 - val_acc: 0.9577
Train loss: 0.08789637273550034
Train accuracy: 0.999960000038147
Test loss: 0.25383884519338606
Test accuracy: 0.9577000015974044相较于调参记录24，本次的测试准确率还降低了0.03%。
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, DOI: 10.1109/TIE.2020.2972458, Date of Publication: 13 February 2020
https://ieeexplore.ieee.org/document/8998530
更新：
期间遇到一个问题，将一模一样的程序，拿到另一台硬件和工具包完全相同的电脑上，结果loss在第一个epoch就急剧飙升，然后成为nan。反复尝试了很多次，每次都这样。折腾了两天，更换了很多显卡驱动、Cuda、TensorFlow和Keras的版本，最后也没有解决，放弃了。
本文暂时没有评论，来添加一个吧(●'◡'●)