本文介绍哈工大团队提出的一种Dynamic ReLU激活函数，即自适应参数化ReLU激活函数，原本是应用在基于一维振动信号的故障诊断，能够让每个样本有自己独特的ReLU参数，在2019年5月3日投稿至IEEE Transactions on Industrial Electronics，2020年1月24日录用，2020年2月13日在IEEE官网公布。
本文在调参记录9的基础上，在数据增强部分添加了shear_range = 30，测试Adaptively Parametric ReLU（APReLU）激活函数在Cifar10图像集上的效果。
Keras里ImageDataGenerator的用法见如下网址：
https://fairyonice.github.io/Learn-about-ImageDataGenerator.html
自适应参数化ReLU激活函数的基本原理见下图：
自适应参数化ReLU：一种Dynamic ReLU激活函数Keras程序如下：
代码语言：python代码运行次数：0复制#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Tue Apr 14 04:17:45 2020
Implemented using TensorFlow 1.0.1 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,
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
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 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 = GlobalAveragePooling2D()(pos_input)
scales_n = GlobalAveragePooling2D()(neg_input)
scales = Concatenate()([scales_n, scales_p])
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('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 = 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(16, 3, padding='same', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(inputs)
net = residual_block(net, 9, 16, downsample=False)
net = residual_block(net, 1, 32, downsample=True)
net = residual_block(net, 8, 32, downsample=False)
net = residual_block(net, 1, 64, downsample=True)
net = residual_block(net, 8, 64, downsample=False)
net = BatchNormalization(momentum=0.9, gamma_regularizer=l2(1e-4))(net)
net = Activation('relu')(net)
net = GlobalAveragePooling2D()(net)
outputs = Dense(10, activation='softmax', kernel_initializer='he_normal', kernel_regularizer=l2(1e-4))(net)
model = Model(inputs=inputs, outputs=outputs)
sgd = optimizers.SGD(lr=0.1, decay=0., momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
# data augmentation
datagen = ImageDataGenerator(
# randomly rotate images in the range (deg 0 to 180)
rotation_range=30,
# 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])实验结果如下：
代码语言：javascript复制x_train shape: (50000, 32, 32, 3)
50000 train samples
10000 test samples
Epoch 1/1000
113s 225ms/step - loss: 3.2549 - acc: 0.4158 - val_loss: 2.7729 - val_acc: 0.5394
Epoch 2/1000
68s 137ms/step - loss: 2.6403 - acc: 0.5484 - val_loss: 2.3416 - val_acc: 0.6117
Epoch 3/1000
69s 138ms/step - loss: 2.2763 - acc: 0.6049 - val_loss: 2.0151 - val_acc: 0.6705
Epoch 4/1000
69s 137ms/step - loss: 2.0062 - acc: 0.6393 - val_loss: 1.8055 - val_acc: 0.6907
Epoch 5/1000
69s 137ms/step - loss: 1.7997 - acc: 0.6673 - val_loss: 1.6339 - val_acc: 0.7058
Epoch 6/1000
69s 138ms/step - loss: 1.6338 - acc: 0.6849 - val_loss: 1.4391 - val_acc: 0.7345
Epoch 7/1000
69s 138ms/step - loss: 1.4911 - acc: 0.7032 - val_loss: 1.3495 - val_acc: 0.7435
Epoch 8/1000
69s 138ms/step - loss: 1.3733 - acc: 0.7196 - val_loss: 1.2311 - val_acc: 0.7668
Epoch 9/1000
68s 137ms/step - loss: 1.2893 - acc: 0.7308 - val_loss: 1.1543 - val_acc: 0.7741
Epoch 10/1000
68s 137ms/step - loss: 1.2164 - acc: 0.7402 - val_loss: 1.0974 - val_acc: 0.7761
Epoch 11/1000
69s 137ms/step - loss: 1.1580 - acc: 0.7470 - val_loss: 1.0477 - val_acc: 0.7835
Epoch 12/1000
69s 137ms/step - loss: 1.1127 - acc: 0.7519 - val_loss: 1.0269 - val_acc: 0.7813
Epoch 13/1000
69s 138ms/step - loss: 1.0713 - acc: 0.7598 - val_loss: 0.9656 - val_acc: 0.7996
Epoch 14/1000
68s 136ms/step - loss: 1.0369 - acc: 0.7664 - val_loss: 0.9576 - val_acc: 0.7929
Epoch 15/1000
68s 135ms/step - loss: 1.0158 - acc: 0.7677 - val_loss: 0.9189 - val_acc: 0.8064
Epoch 16/1000
68s 135ms/step - loss: 0.9948 - acc: 0.7733 - val_loss: 0.9198 - val_acc: 0.8022
Epoch 17/1000
68s 136ms/step - loss: 0.9720 - acc: 0.7775 - val_loss: 0.9267 - val_acc: 0.7954
Epoch 18/1000
68s 135ms/step - loss: 0.9548 - acc: 0.7813 - val_loss: 0.8897 - val_acc: 0.8043
Epoch 19/1000
68s 135ms/step - loss: 0.9446 - acc: 0.7847 - val_loss: 0.8642 - val_acc: 0.8104
Epoch 20/1000
68s 135ms/step - loss: 0.9290 - acc: 0.7873 - val_loss: 0.8666 - val_acc: 0.8119
Epoch 21/1000
68s 135ms/step - loss: 0.9131 - acc: 0.7913 - val_loss: 0.8433 - val_acc: 0.8202
Epoch 22/1000
68s 135ms/step - loss: 0.9099 - acc: 0.7912 - val_loss: 0.8735 - val_acc: 0.8077
Epoch 23/1000
67s 135ms/step - loss: 0.9000 - acc: 0.7956 - val_loss: 0.8418 - val_acc: 0.8150
Epoch 24/1000
68s 135ms/step - loss: 0.8962 - acc: 0.7966 - val_loss: 0.8452 - val_acc: 0.8181
Epoch 25/1000
68s 135ms/step - loss: 0.8874 - acc: 0.7994 - val_loss: 0.8209 - val_acc: 0.8242
Epoch 26/1000
68s 136ms/step - loss: 0.8810 - acc: 0.8021 - val_loss: 0.8378 - val_acc: 0.8202
Epoch 27/1000
68s 135ms/step - loss: 0.8764 - acc: 0.8026 - val_loss: 0.8474 - val_acc: 0.8173
Epoch 28/1000
67s 135ms/step - loss: 0.8706 - acc: 0.8040 - val_loss: 0.8239 - val_acc: 0.8230
Epoch 29/1000
68s 135ms/step - loss: 0.8655 - acc: 0.8075 - val_loss: 0.8163 - val_acc: 0.8244
Epoch 30/1000
68s 135ms/step - loss: 0.8600 - acc: 0.8074 - val_loss: 0.8065 - val_acc: 0.8288
Epoch 31/1000
68s 135ms/step - loss: 0.8544 - acc: 0.8113 - val_loss: 0.8080 - val_acc: 0.8306
Epoch 32/1000
68s 135ms/step - loss: 0.8510 - acc: 0.8121 - val_loss: 0.8152 - val_acc: 0.8304
Epoch 33/1000
68s 135ms/step - loss: 0.8464 - acc: 0.8142 - val_loss: 0.7827 - val_acc: 0.8387
Epoch 34/1000
68s 135ms/step - loss: 0.8429 - acc: 0.8166 - val_loss: 0.7738 - val_acc: 0.8453
Epoch 35/1000
68s 135ms/step - loss: 0.8366 - acc: 0.8160 - val_loss: 0.7855 - val_acc: 0.8388
Epoch 36/1000
68s 135ms/step - loss: 0.8352 - acc: 0.8191 - val_loss: 0.7651 - val_acc: 0.8468
Epoch 37/1000
68s 135ms/step - loss: 0.8292 - acc: 0.8212 - val_loss: 0.7620 - val_acc: 0.8470
Epoch 38/1000
68s 135ms/step - loss: 0.8319 - acc: 0.8208 - val_loss: 0.7890 - val_acc: 0.8376
Epoch 39/1000
68s 136ms/step - loss: 0.8239 - acc: 0.8256 - val_loss: 0.7870 - val_acc: 0.8370
Epoch 40/1000
68s 135ms/step - loss: 0.8266 - acc: 0.8216 - val_loss: 0.7975 - val_acc: 0.8331
Epoch 41/1000
68s 135ms/step - loss: 0.8209 - acc: 0.8239 - val_loss: 0.7982 - val_acc: 0.8334
Epoch 42/1000
68s 135ms/step - loss: 0.8135 - acc: 0.8276 - val_loss: 0.7722 - val_acc: 0.8427
Epoch 43/1000
68s 135ms/step - loss: 0.8115 - acc: 0.8280 - val_loss: 0.7658 - val_acc: 0.8430
Epoch 44/1000
67s 135ms/step - loss: 0.8166 - acc: 0.8259 - val_loss: 0.7388 - val_acc: 0.8559
Epoch 45/1000
67s 135ms/step - loss: 0.8108 - acc: 0.8293 - val_loss: 0.7728 - val_acc: 0.8436
Epoch 46/1000
68s 135ms/step - loss: 0.8046 - acc: 0.8303 - val_loss: 0.7684 - val_acc: 0.8434
Epoch 47/1000
68s 136ms/step - loss: 0.8055 - acc: 0.8322 - val_loss: 0.7478 - val_acc: 0.8511
Epoch 48/1000
68s 135ms/step - loss: 0.8100 - acc: 0.8290 - val_loss: 0.7644 - val_acc: 0.8445
Epoch 49/1000
68s 135ms/step - loss: 0.8027 - acc: 0.8325 - val_loss: 0.7449 - val_acc: 0.8545
Epoch 50/1000
67s 135ms/step - loss: 0.8052 - acc: 0.8299 - val_loss: 0.7941 - val_acc: 0.8377
Epoch 51/1000
68s 135ms/step - loss: 0.7969 - acc: 0.8339 - val_loss: 0.7617 - val_acc: 0.8481
Epoch 52/1000
68s 135ms/step - loss: 0.7989 - acc: 0.8335 - val_loss: 0.7559 - val_acc: 0.8550
Epoch 53/1000
68s 136ms/step - loss: 0.7927 - acc: 0.8353 - val_loss: 0.7482 - val_acc: 0.8536
Epoch 54/1000
68s 135ms/step - loss: 0.7931 - acc: 0.8365 - val_loss: 0.7405 - val_acc: 0.8570
Epoch 55/1000
68s 135ms/step - loss: 0.7933 - acc: 0.8372 - val_loss: 0.7541 - val_acc: 0.8535
Epoch 56/1000
68s 135ms/step - loss: 0.7887 - acc: 0.8389 - val_loss: 0.7805 - val_acc: 0.8436
Epoch 57/1000
68s 135ms/step - loss: 0.7877 - acc: 0.8385 - val_loss: 0.7304 - val_acc: 0.8617
Epoch 58/1000
68s 135ms/step - loss: 0.7836 - acc: 0.8404 - val_loss: 0.7630 - val_acc: 0.8480
Epoch 59/1000
68s 135ms/step - loss: 0.7859 - acc: 0.8394 - val_loss: 0.7369 - val_acc: 0.8568
Epoch 60/1000
68s 135ms/step - loss: 0.7864 - acc: 0.8376 - val_loss: 0.7606 - val_acc: 0.8492
Epoch 61/1000
68s 135ms/step - loss: 0.7827 - acc: 0.8401 - val_loss: 0.7497 - val_acc: 0.8524
Epoch 62/1000
68s 135ms/step - loss: 0.7804 - acc: 0.8427 - val_loss: 0.7526 - val_acc: 0.8559
Epoch 63/1000
68s 135ms/step - loss: 0.7766 - acc: 0.8435 - val_loss: 0.7448 - val_acc: 0.8586
Epoch 64/1000
68s 135ms/step - loss: 0.7792 - acc: 0.8419 - val_loss: 0.7605 - val_acc: 0.8511
Epoch 65/1000
68s 135ms/step - loss: 0.7790 - acc: 0.8435 - val_loss: 0.7330 - val_acc: 0.8551
Epoch 66/1000
68s 135ms/step - loss: 0.7748 - acc: 0.8435 - val_loss: 0.7528 - val_acc: 0.8543
Epoch 67/1000
68s 135ms/step - loss: 0.7733 - acc: 0.8452 - val_loss: 0.7330 - val_acc: 0.8585
Epoch 68/1000
68s 135ms/step - loss: 0.7759 - acc: 0.8438 - val_loss: 0.7497 - val_acc: 0.8520
Epoch 69/1000
68s 135ms/step - loss: 0.7680 - acc: 0.8466 - val_loss: 0.7422 - val_acc: 0.8606
Epoch 70/1000
68s 135ms/step - loss: 0.7662 - acc: 0.8473 - val_loss: 0.7185 - val_acc: 0.8633
Epoch 71/1000
68s 135ms/step - loss: 0.7658 - acc: 0.8467 - val_loss: 0.7170 - val_acc: 0.8657
Epoch 72/1000
68s 135ms/step - loss: 0.7681 - acc: 0.8464 - val_loss: 0.7325 - val_acc: 0.8600
Epoch 73/1000
68s 135ms/step - loss: 0.7658 - acc: 0.8477 - val_loss: 0.7109 - val_acc: 0.8662
Epoch 74/1000
68s 135ms/step - loss: 0.7616 - acc: 0.8499 - val_loss: 0.7028 - val_acc: 0.8733
Epoch 75/1000
68s 135ms/step - loss: 0.7621 - acc: 0.8482 - val_loss: 0.7178 - val_acc: 0.8639
Epoch 76/1000
68s 135ms/step - loss: 0.7606 - acc: 0.8496 - val_loss: 0.7096 - val_acc: 0.8674
Epoch 77/1000
68s 135ms/step - loss: 0.7590 - acc: 0.8500 - val_loss: 0.7340 - val_acc: 0.8598
Epoch 78/1000
68s 135ms/step - loss: 0.7639 - acc: 0.8475 - val_loss: 0.7212 - val_acc: 0.8655
Epoch 79/1000
68s 135ms/step - loss: 0.7613 - acc: 0.8477 - val_loss: 0.7171 - val_acc: 0.8702
Epoch 80/1000
67s 135ms/step - loss: 0.7562 - acc: 0.8518 - val_loss: 0.7336 - val_acc: 0.8594
Epoch 81/1000
68s 136ms/step - loss: 0.7532 - acc: 0.8515 - val_loss: 0.7229 - val_acc: 0.8607
Epoch 82/1000
68s 135ms/step - loss: 0.7511 - acc: 0.8541 - val_loss: 0.7062 - val_acc: 0.8688
Epoch 83/1000
68s 135ms/step - loss: 0.7510 - acc: 0.8530 - val_loss: 0.6977 - val_acc: 0.8746
Epoch 84/1000
68s 135ms/step - loss: 0.7562 - acc: 0.8524 - val_loss: 0.7319 - val_acc: 0.8595
Epoch 85/1000
67s 135ms/step - loss: 0.7527 - acc: 0.8530 - val_loss: 0.7161 - val_acc: 0.8660
Epoch 86/1000
67s 135ms/step - loss: 0.7523 - acc: 0.8524 - val_loss: 0.7244 - val_acc: 0.8654
Epoch 87/1000
67s 135ms/step - loss: 0.7505 - acc: 0.8532 - val_loss: 0.7192 - val_acc: 0.8636
Epoch 88/1000
68s 135ms/step - loss: 0.7528 - acc: 0.8516 - val_loss: 0.7316 - val_acc: 0.8645
Epoch 89/1000
68s 135ms/step - loss: 0.7480 - acc: 0.8557 - val_loss: 0.7289 - val_acc: 0.8638
Epoch 90/1000
68s 135ms/step - loss: 0.7435 - acc: 0.8550 - val_loss: 0.7020 - val_acc: 0.8763
Epoch 91/1000
68s 135ms/step - loss: 0.7466 - acc: 0.8563 - val_loss: 0.6977 - val_acc: 0.8750
Epoch 92/1000
68s 135ms/step - loss: 0.7438 - acc: 0.8561 - val_loss: 0.7171 - val_acc: 0.8643
Epoch 93/1000
67s 135ms/step - loss: 0.7438 - acc: 0.8564 - val_loss: 0.7189 - val_acc: 0.8687
Epoch 94/1000
68s 135ms/step - loss: 0.7442 - acc: 0.8566 - val_loss: 0.7072 - val_acc: 0.8685
Epoch 95/1000
68s 135ms/step - loss: 0.7468 - acc: 0.8569 - val_loss: 0.7547 - val_acc: 0.8560
Epoch 96/1000
68s 135ms/step - loss: 0.7468 - acc: 0.8547 - val_loss: 0.7080 - val_acc: 0.8699
Epoch 97/1000
68s 135ms/step - loss: 0.7455 - acc: 0.8559 - val_loss: 0.7020 - val_acc: 0.8711
Epoch 98/1000
68s 135ms/step - loss: 0.7427 - acc: 0.8544 - val_loss: 0.7352 - val_acc: 0.8610
Epoch 99/1000
68s 136ms/step - loss: 0.7424 - acc: 0.8567 - val_loss: 0.7480 - val_acc: 0.8583
Epoch 100/1000
68s 135ms/step - loss: 0.7397 - acc: 0.8579 - val_loss: 0.7151 - val_acc: 0.8650
Epoch 101/1000
68s 135ms/step - loss: 0.7447 - acc: 0.8568 - val_loss: 0.7235 - val_acc: 0.8659
Epoch 102/1000
68s 135ms/step - loss: 0.7367 - acc: 0.8598 - val_loss: 0.7229 - val_acc: 0.8623
Epoch 103/1000
67s 135ms/step - loss: 0.7371 - acc: 0.8586 - val_loss: 0.6899 - val_acc: 0.8769
Epoch 104/1000
68s 135ms/step - loss: 0.7401 - acc: 0.8567 - val_loss: 0.7273 - val_acc: 0.8616
Epoch 105/1000
68s 135ms/step - loss: 0.7382 - acc: 0.8578 - val_loss: 0.7089 - val_acc: 0.8682
Epoch 106/1000
68s 135ms/step - loss: 0.7386 - acc: 0.8580 - val_loss: 0.7158 - val_acc: 0.8659
Epoch 107/1000
67s 135ms/step - loss: 0.7361 - acc: 0.8584 - val_loss: 0.7147 - val_acc: 0.8701
Epoch 108/1000
67s 135ms/step - loss: 0.7408 - acc: 0.8580 - val_loss: 0.7083 - val_acc: 0.8686
Epoch 109/1000
68s 135ms/step - loss: 0.7362 - acc: 0.8599 - val_loss: 0.7096 - val_acc: 0.8703
Epoch 110/1000
67s 135ms/step - loss: 0.7335 - acc: 0.8600 - val_loss: 0.7148 - val_acc: 0.8683
Epoch 111/1000
67s 135ms/step - loss: 0.7334 - acc: 0.8626 - val_loss: 0.7050 - val_acc: 0.8741
Epoch 112/1000
68s 135ms/step - loss: 0.7360 - acc: 0.8586 - val_loss: 0.7150 - val_acc: 0.8682
Epoch 113/1000
68s 136ms/step - loss: 0.7371 - acc: 0.8583 - val_loss: 0.7447 - val_acc: 0.8583
Epoch 114/1000
68s 135ms/step - loss: 0.7352 - acc: 0.8599 - val_loss: 0.6937 - val_acc: 0.8755
Epoch 115/1000
68s 135ms/step - loss: 0.7314 - acc: 0.8604 - val_loss: 0.7140 - val_acc: 0.8684
Epoch 116/1000
68s 135ms/step - loss: 0.7333 - acc: 0.8607 - val_loss: 0.7305 - val_acc: 0.8686
Epoch 117/1000
68s 135ms/step - loss: 0.7277 - acc: 0.8617 - val_loss: 0.7002 - val_acc: 0.8719
Epoch 118/1000
68s 135ms/step - loss: 0.7356 - acc: 0.8580 - val_loss: 0.6926 - val_acc: 0.8763
Epoch 119/1000
68s 135ms/step - loss: 0.7244 - acc: 0.8642 - val_loss: 0.7079 - val_acc: 0.8669
Epoch 120/1000
68s 136ms/step - loss: 0.7302 - acc: 0.8613 - val_loss: 0.7113 - val_acc: 0.8695
Epoch 121/1000
68s 135ms/step - loss: 0.7340 - acc: 0.8608 - val_loss: 0.7415 - val_acc: 0.8554
Epoch 122/1000
68s 135ms/step - loss: 0.7304 - acc: 0.8608 - val_loss: 0.6978 - val_acc: 0.8760
Epoch 123/1000
68s 135ms/step - loss: 0.7263 - acc: 0.8630 - val_loss: 0.6974 - val_acc: 0.8734
Epoch 124/1000
68s 135ms/step - loss: 0.7261 - acc: 0.8625 - val_loss: 0.7109 - val_acc: 0.8715
Epoch 125/1000
67s 135ms/step - loss: 0.7313 - acc: 0.8623 - val_loss: 0.6946 - val_acc: 0.8745
Epoch 126/1000
67s 135ms/step - loss: 0.7277 - acc: 0.8620 - val_loss: 0.7178 - val_acc: 0.8685
Epoch 127/1000
68s 135ms/step - loss: 0.7231 - acc: 0.8653 - val_loss: 0.6999 - val_acc: 0.8762
Epoch 128/1000
68s 135ms/step - loss: 0.7252 - acc: 0.8635 - val_loss: 0.7009 - val_acc: 0.8718
Epoch 129/1000
68s 135ms/step - loss: 0.7284 - acc: 0.8626 - val_loss: 0.7148 - val_acc: 0.8682
Epoch 130/1000
68s 135ms/step - loss: 0.7236 - acc: 0.8646 - val_loss: 0.6945 - val_acc: 0.8746
Epoch 131/1000
68s 135ms/step - loss: 0.7203 - acc: 0.8653 - val_loss: 0.7002 - val_acc: 0.8705
Epoch 132/1000
68s 135ms/step - loss: 0.7248 - acc: 0.8626 - val_loss: 0.7097 - val_acc: 0.8718
Epoch 133/1000
67s 135ms/step - loss: 0.7190 - acc: 0.8660 - val_loss: 0.6993 - val_acc: 0.8722
Epoch 134/1000
68s 136ms/step - loss: 0.7206 - acc: 0.8645 - val_loss: 0.7042 - val_acc: 0.8763
Epoch 135/1000
68s 135ms/step - loss: 0.7248 - acc: 0.8637 - val_loss: 0.6742 - val_acc: 0.8844
Epoch 136/1000
68s 135ms/step - loss: 0.7181 - acc: 0.8650 - val_loss: 0.6972 - val_acc: 0.8721
Epoch 137/1000
67s 135ms/step - loss: 0.7170 - acc: 0.8667 - val_loss: 0.7270 - val_acc: 0.8642
Epoch 138/1000
68s 135ms/step - loss: 0.7209 - acc: 0.8649 - val_loss: 0.7107 - val_acc: 0.8687
Epoch 139/1000
68s 136ms/step - loss: 0.7195 - acc: 0.8652 - val_loss: 0.6993 - val_acc: 0.8752
Epoch 140/1000
68s 135ms/step - loss: 0.7229 - acc: 0.8647 - val_loss: 0.6949 - val_acc: 0.8800
Epoch 141/1000
67s 135ms/step - loss: 0.7154 - acc: 0.8674 - val_loss: 0.6828 - val_acc: 0.8780
Epoch 142/1000
67s 135ms/step - loss: 0.7146 - acc: 0.8675 - val_loss: 0.6799 - val_acc: 0.8818
Epoch 143/1000
68s 135ms/step - loss: 0.7131 - acc: 0.8679 - val_loss: 0.7237 - val_acc: 0.8655
Epoch 144/1000
68s 135ms/step - loss: 0.7167 - acc: 0.8662 - val_loss: 0.7140 - val_acc: 0.8696
Epoch 145/1000
68s 136ms/step - loss: 0.7131 - acc: 0.8677 - val_loss: 0.7086 - val_acc: 0.8696
Epoch 146/1000
67s 135ms/step - loss: 0.7184 - acc: 0.8665 - val_loss: 0.7058 - val_acc: 0.8729
Epoch 147/1000
68s 135ms/step - loss: 0.7179 - acc: 0.8654 - val_loss: 0.7021 - val_acc: 0.8741
Epoch 148/1000
67s 135ms/step - loss: 0.7176 - acc: 0.8671 - val_loss: 0.6892 - val_acc: 0.8795
Epoch 149/1000
68s 135ms/step - loss: 0.7123 - acc: 0.8685 - val_loss: 0.7027 - val_acc: 0.8700
Epoch 150/1000
68s 136ms/step - loss: 0.7146 - acc: 0.8671 - val_loss: 0.6926 - val_acc: 0.8755
Epoch 151/1000
68s 135ms/step - loss: 0.7122 - acc: 0.8651 - val_loss: 0.7179 - val_acc: 0.8685
Epoch 152/1000
68s 136ms/step - loss: 0.7149 - acc: 0.8675 - val_loss: 0.7136 - val_acc: 0.8690
Epoch 153/1000
68s 135ms/step - loss: 0.7141 - acc: 0.8669 - val_loss: 0.7193 - val_acc: 0.8672
Epoch 154/1000
68s 136ms/step - loss: 0.7084 - acc: 0.8684 - val_loss: 0.6779 - val_acc: 0.8826
Epoch 155/1000
67s 135ms/step - loss: 0.7143 - acc: 0.8671 - val_loss: 0.7092 - val_acc: 0.8685
Epoch 156/1000
68s 136ms/step - loss: 0.7118 - acc: 0.8674 - val_loss: 0.7010 - val_acc: 0.8732
Epoch 157/1000
69s 138ms/step - loss: 0.7126 - acc: 0.8677 - val_loss: 0.6918 - val_acc: 0.8766
Epoch 158/1000
68s 137ms/step - loss: 0.7064 - acc: 0.8701 - val_loss: 0.7253 - val_acc: 0.8636
Epoch 159/1000
68s 137ms/step - loss: 0.7107 - acc: 0.8674 - val_loss: 0.7008 - val_acc: 0.8745
Epoch 160/1000
68s 137ms/step - loss: 0.7097 - acc: 0.8698 - val_loss: 0.6922 - val_acc: 0.8771
Epoch 161/1000
68s 137ms/step - loss: 0.7091 - acc: 0.8675 - val_loss: 0.6786 - val_acc: 0.8813
Epoch 162/1000
69s 138ms/step - loss: 0.7117 - acc: 0.8680 - val_loss: 0.7017 - val_acc: 0.8740
Epoch 163/1000
69s 137ms/step - loss: 0.7110 - acc: 0.8681 - val_loss: 0.6862 - val_acc: 0.8800
Epoch 164/1000
68s 137ms/step - loss: 0.7099 - acc: 0.8693 - val_loss: 0.7053 - val_acc: 0.8709
Epoch 165/1000
69s 138ms/step - loss: 0.7104 - acc: 0.8694 - val_loss: 0.6846 - val_acc: 0.8828
Epoch 166/1000
68s 136ms/step - loss: 0.7078 - acc: 0.8715 - val_loss: 0.6968 - val_acc: 0.8749
Epoch 167/1000
68s 136ms/step - loss: 0.7076 - acc: 0.8719 - val_loss: 0.6872 - val_acc: 0.8782
Epoch 168/1000
68s 136ms/step - loss: 0.7099 - acc: 0.8679 - val_loss: 0.6928 - val_acc: 0.8755
Epoch 169/1000
68s 136ms/step - loss: 0.7101 - acc: 0.8678 - val_loss: 0.6947 - val_acc: 0.8786
Epoch 170/1000
68s 137ms/step - loss: 0.7097 - acc: 0.8717 - val_loss: 0.6886 - val_acc: 0.8789
Epoch 171/1000
69s 137ms/step - loss: 0.7070 - acc: 0.8702 - val_loss: 0.6878 - val_acc: 0.8793
Epoch 172/1000
69s 137ms/step - loss: 0.7117 - acc: 0.8679 - val_loss: 0.6783 - val_acc: 0.8836
Epoch 173/1000
68s 137ms/step - loss: 0.7102 - acc: 0.8687 - val_loss: 0.6709 - val_acc: 0.8865
Epoch 174/1000
69s 137ms/step - loss: 0.7038 - acc: 0.8717 - val_loss: 0.6839 - val_acc: 0.8804
Epoch 175/1000
68s 137ms/step - loss: 0.7062 - acc: 0.8713 - val_loss: 0.6934 - val_acc: 0.8780
Epoch 176/1000
68s 137ms/step - loss: 0.7092 - acc: 0.8684 - val_loss: 0.7045 - val_acc: 0.8737
Epoch 177/1000
68s 137ms/step - loss: 0.7048 - acc: 0.8703 - val_loss: 0.6935 - val_acc: 0.8764
Epoch 178/1000
68s 137ms/step - loss: 0.7056 - acc: 0.8713 - val_loss: 0.6825 - val_acc: 0.8800
Epoch 179/1000
69s 137ms/step - loss: 0.7027 - acc: 0.8722 - val_loss: 0.6860 - val_acc: 0.8812
Epoch 180/1000
67s 135ms/step - loss: 0.7056 - acc: 0.8699 - val_loss: 0.6882 - val_acc: 0.8762
Epoch 181/1000
67s 135ms/step - loss: 0.6974 - acc: 0.8745 - val_loss: 0.7030 - val_acc: 0.8704
Epoch 182/1000
67s 135ms/step - loss: 0.7028 - acc: 0.8714 - val_loss: 0.6754 - val_acc: 0.8860
Epoch 183/1000
67s 135ms/step - loss: 0.7022 - acc: 0.8715 - val_loss: 0.6635 - val_acc: 0.8842
Epoch 184/1000
68s 136ms/step - loss: 0.7034 - acc: 0.8704 - val_loss: 0.6905 - val_acc: 0.8762
Epoch 185/1000
67s 135ms/step - loss: 0.7058 - acc: 0.8709 - val_loss: 0.7066 - val_acc: 0.8740
Epoch 186/1000
67s 135ms/step - loss: 0.7016 - acc: 0.8726 - val_loss: 0.6842 - val_acc: 0.8784
Epoch 187/1000
67s 135ms/step - loss: 0.6999 - acc: 0.8719 - val_loss: 0.7051 - val_acc: 0.8731
Epoch 188/1000
67s 135ms/step - loss: 0.7026 - acc: 0.8710 - val_loss: 0.6811 - val_acc: 0.8811
Epoch 189/1000
68s 135ms/step - loss: 0.7040 - acc: 0.8711 - val_loss: 0.6794 - val_acc: 0.8786
Epoch 190/1000
67s 135ms/step - loss: 0.7004 - acc: 0.8728 - val_loss: 0.6594 - val_acc: 0.8916
Epoch 191/1000
68s 136ms/step - loss: 0.6982 - acc: 0.8747 - val_loss: 0.6616 - val_acc: 0.8850
Epoch 192/1000
68s 135ms/step - loss: 0.7036 - acc: 0.8718 - val_loss: 0.6959 - val_acc: 0.8730
Epoch 193/1000
67s 135ms/step - loss: 0.7017 - acc: 0.8708 - val_loss: 0.6671 - val_acc: 0.8862
Epoch 194/1000
67s 135ms/step - loss: 0.6982 - acc: 0.8738 - val_loss: 0.6885 - val_acc: 0.8790
Epoch 195/1000
68s 136ms/step - loss: 0.6996 - acc: 0.8714 - val_loss: 0.6892 - val_acc: 0.8770
Epoch 196/1000
68s 136ms/step - loss: 0.7026 - acc: 0.8706 - val_loss: 0.6824 - val_acc: 0.8792
Epoch 197/1000
68s 136ms/step - loss: 0.7061 - acc: 0.8695 - val_loss: 0.6893 - val_acc: 0.8793
Epoch 198/1000
68s 135ms/step - loss: 0.7023 - acc: 0.8714 - val_loss: 0.6797 - val_acc: 0.8819
Epoch 199/1000
67s 135ms/step - loss: 0.7021 - acc: 0.8726 - val_loss: 0.6969 - val_acc: 0.8754
Epoch 200/1000
68s 136ms/step - loss: 0.7023 - acc: 0.8711 - val_loss: 0.6922 - val_acc: 0.8758
Epoch 201/1000
68s 135ms/step - loss: 0.7050 - acc: 0.8705 - val_loss: 0.6879 - val_acc: 0.8792
Epoch 202/1000
68s 135ms/step - loss: 0.7012 - acc: 0.8713 - val_loss: 0.6756 - val_acc: 0.8845
Epoch 203/1000
68s 136ms/step - loss: 0.7021 - acc: 0.8726 - val_loss: 0.6542 - val_acc: 0.8904
Epoch 204/1000
68s 136ms/step - loss: 0.6981 - acc: 0.8741 - val_loss: 0.7060 - val_acc: 0.8739
Epoch 205/1000
68s 135ms/step - loss: 0.7008 - acc: 0.8718 - val_loss: 0.6938 - val_acc: 0.8741
Epoch 206/1000
68s 136ms/step - loss: 0.6974 - acc: 0.8725 - val_loss: 0.6786 - val_acc: 0.8833
Epoch 207/1000
67s 135ms/step - loss: 0.6938 - acc: 0.8739 - val_loss: 0.6928 - val_acc: 0.8750
Epoch 208/1000
68s 135ms/step - loss: 0.7075 - acc: 0.8690 - val_loss: 0.6770 - val_acc: 0.8806
Epoch 209/1000
68s 136ms/step - loss: 0.6978 - acc: 0.8723 - val_loss: 0.6913 - val_acc: 0.8812
Epoch 210/1000
67s 135ms/step - loss: 0.6974 - acc: 0.8727 - val_loss: 0.6764 - val_acc: 0.8827
...
Epoch 291/1000
69s 138ms/step - loss: 0.6827 - acc: 0.8805 - val_loss: 0.6700 - val_acc: 0.8877
Epoch 292/1000
69s 137ms/step - loss: 0.6875 - acc: 0.8770 - val_loss: 0.6843 - val_acc: 0.8802
Epoch 293/1000
69s 138ms/step - loss: 0.6861 - acc: 0.8795 - val_loss: 0.6889 - val_acc: 0.8812
Epoch 294/1000
68s 137ms/step - loss: 0.6896 - acc: 0.8759 - val_loss: 0.6688 - val_acc: 0.8874
Epoch 295/1000
69s 138ms/step - loss: 0.6792 - acc: 0.8805 - val_loss: 0.6813 - val_acc: 0.8802
Epoch 296/1000
69s 138ms/step - loss: 0.6946 - acc: 0.8733 - val_loss: 0.6697 - val_acc: 0.8858
Epoch 297/1000
69s 138ms/step - loss: 0.6887 - acc: 0.8755 - val_loss: 0.6707 - val_acc: 0.8848
Epoch 298/1000
69s 138ms/step - loss: 0.6875 - acc: 0.8765 - val_loss: 0.7025 - val_acc: 0.8718
Epoch 299/1000
69s 137ms/step - loss: 0.6853 - acc: 0.8789 - val_loss: 0.6842 - val_acc: 0.8805
Epoch 300/1000
69s 138ms/step - loss: 0.6806 - acc: 0.8809 - val_loss: 0.6948 - val_acc: 0.8809
Epoch 301/1000
lr changed to 0.010000000149011612
69s 138ms/step - loss: 0.5763 - acc: 0.9142 - val_loss: 0.5780 - val_acc: 0.9169
Epoch 302/1000
69s 138ms/step - loss: 0.5127 - acc: 0.9355 - val_loss: 0.5618 - val_acc: 0.9209
Epoch 303/1000
68s 137ms/step - loss: 0.4950 - acc: 0.9401 - val_loss: 0.5561 - val_acc: 0.9223
Epoch 304/1000
68s 137ms/step - loss: 0.4744 - acc: 0.9449 - val_loss: 0.5485 - val_acc: 0.9229
Epoch 305/1000
68s 137ms/step - loss: 0.4602 - acc: 0.9489 - val_loss: 0.5469 - val_acc: 0.9206
Epoch 306/1000
69s 137ms/step - loss: 0.4533 - acc: 0.9479 - val_loss: 0.5368 - val_acc: 0.9209
Epoch 307/1000
69s 137ms/step - loss: 0.4463 - acc: 0.9498 - val_loss: 0.5294 - val_acc: 0.9230
Epoch 308/1000
69s 137ms/step - loss: 0.4371 - acc: 0.9508 - val_loss: 0.5304 - val_acc: 0.9228
Epoch 309/1000
69s 137ms/step - loss: 0.4276 - acc: 0.9515 - val_loss: 0.5217 - val_acc: 0.9236
Epoch 310/1000
68s 136ms/step - loss: 0.4185 - acc: 0.9542 - val_loss: 0.5202 - val_acc: 0.9235
Epoch 311/1000
69s 138ms/step - loss: 0.4079 - acc: 0.9563 - val_loss: 0.5213 - val_acc: 0.9224
Epoch 312/1000
69s 137ms/step - loss: 0.4028 - acc: 0.9559 - val_loss: 0.5149 - val_acc: 0.9241
Epoch 313/1000
68s 136ms/step - loss: 0.3940 - acc: 0.9582 - val_loss: 0.5182 - val_acc: 0.9229
Epoch 314/1000
69s 138ms/step - loss: 0.3913 - acc: 0.9584 - val_loss: 0.5063 - val_acc: 0.9222
Epoch 315/1000
69s 138ms/step - loss: 0.3815 - acc: 0.9599 - val_loss: 0.5065 - val_acc: 0.9242
Epoch 316/1000
69s 138ms/step - loss: 0.3779 - acc: 0.9596 - val_loss: 0.5105 - val_acc: 0.9197
Epoch 317/1000
69s 138ms/step - loss: 0.3734 - acc: 0.9607 - val_loss: 0.4951 - val_acc: 0.9242
Epoch 318/1000
69s 138ms/step - loss: 0.3668 - acc: 0.9608 - val_loss: 0.4984 - val_acc: 0.9226
Epoch 319/1000
68s 137ms/step - loss: 0.3600 - acc: 0.9628 - val_loss: 0.5003 - val_acc: 0.9195
Epoch 320/1000
68s 137ms/step - loss: 0.3562 - acc: 0.9622 - val_loss: 0.4927 - val_acc: 0.9206
Epoch 321/1000
69s 138ms/step - loss: 0.3551 - acc: 0.9619 - val_loss: 0.4883 - val_acc: 0.9233
Epoch 322/1000
69s 138ms/step - loss: 0.3467 - acc: 0.9635 - val_loss: 0.4820 - val_acc: 0.9247
Epoch 323/1000
69s 138ms/step - loss: 0.3468 - acc: 0.9621 - val_loss: 0.4795 - val_acc: 0.9225
Epoch 324/1000
68s 136ms/step - loss: 0.3386 - acc: 0.9651 - val_loss: 0.4927 - val_acc: 0.9205
Epoch 325/1000
68s 135ms/step - loss: 0.3368 - acc: 0.9644 - val_loss: 0.4823 - val_acc: 0.9205
Epoch 326/1000
68s 136ms/step - loss: 0.3284 - acc: 0.9667 - val_loss: 0.4691 - val_acc: 0.9236
Epoch 327/1000
69s 138ms/step - loss: 0.3255 - acc: 0.9658 - val_loss: 0.4734 - val_acc: 0.9252
Epoch 328/1000
68s 136ms/step - loss: 0.3255 - acc: 0.9648 - val_loss: 0.4795 - val_acc: 0.9230
Epoch 329/1000
68s 136ms/step - loss: 0.3257 - acc: 0.9638 - val_loss: 0.4681 - val_acc: 0.9223
Epoch 330/1000
68s 136ms/step - loss: 0.3181 - acc: 0.9648 - val_loss: 0.4670 - val_acc: 0.9215
Epoch 331/1000
68s 136ms/step - loss: 0.3138 - acc: 0.9660 - val_loss: 0.4821 - val_acc: 0.9185
Epoch 332/1000
68s 136ms/step - loss: 0.3140 - acc: 0.9648 - val_loss: 0.4727 - val_acc: 0.9202
Epoch 333/1000
69s 137ms/step - loss: 0.3102 - acc: 0.9663 - val_loss: 0.4632 - val_acc: 0.9231
Epoch 334/1000
68s 137ms/step - loss: 0.3085 - acc: 0.9663 - val_loss: 0.4611 - val_acc: 0.9240
Epoch 335/1000
68s 137ms/step - loss: 0.3019 - acc: 0.9679 - val_loss: 0.4614 - val_acc: 0.9238
Epoch 336/1000
69s 138ms/step - loss: 0.3046 - acc: 0.9654 - val_loss: 0.4635 - val_acc: 0.9202
Epoch 337/1000
68s 137ms/step - loss: 0.3015 - acc: 0.9660 - val_loss: 0.4599 - val_acc: 0.9228
Epoch 338/1000
69s 137ms/step - loss: 0.2992 - acc: 0.9662 - val_loss: 0.4577 - val_acc: 0.9207
Epoch 339/1000
69s 138ms/step - loss: 0.2942 - acc: 0.9669 - val_loss: 0.4702 - val_acc: 0.9172
Epoch 340/1000
69s 137ms/step - loss: 0.2924 - acc: 0.9675 - val_loss: 0.4545 - val_acc: 0.9211
...
Epoch 597/1000
68s 135ms/step - loss: 0.2366 - acc: 0.9703 - val_loss: 0.4557 - val_acc: 0.9103
Epoch 598/1000
68s 135ms/step - loss: 0.2399 - acc: 0.9697 - val_loss: 0.4449 - val_acc: 0.9117
Epoch 599/1000
67s 135ms/step - loss: 0.2397 - acc: 0.9689 - val_loss: 0.4359 - val_acc: 0.9147
Epoch 600/1000
68s 136ms/step - loss: 0.2341 - acc: 0.9717 - val_loss: 0.4224 - val_acc: 0.9169
Epoch 601/1000
lr changed to 0.0009999999776482583
68s 136ms/step - loss: 0.2082 - acc: 0.9813 - val_loss: 0.3916 - val_acc: 0.9268
Epoch 602/1000
68s 136ms/step - loss: 0.1952 - acc: 0.9865 - val_loss: 0.3854 - val_acc: 0.9281
Epoch 603/1000
68s 136ms/step - loss: 0.1878 - acc: 0.9881 - val_loss: 0.3852 - val_acc: 0.9299
Epoch 604/1000
68s 136ms/step - loss: 0.1846 - acc: 0.9899 - val_loss: 0.3842 - val_acc: 0.9298
Epoch 605/1000
68s 135ms/step - loss: 0.1826 - acc: 0.9909 - val_loss: 0.3829 - val_acc: 0.9326
Epoch 606/1000
68s 136ms/step - loss: 0.1808 - acc: 0.9912 - val_loss: 0.3838 - val_acc: 0.9305
Epoch 607/1000
68s 136ms/step - loss: 0.1771 - acc: 0.9927 - val_loss: 0.3851 - val_acc: 0.9303
Epoch 608/1000
68s 136ms/step - loss: 0.1768 - acc: 0.9922 - val_loss: 0.3898 - val_acc: 0.9304
Epoch 609/1000
68s 135ms/step - loss: 0.1758 - acc: 0.9926 - val_loss: 0.3878 - val_acc: 0.9309
Epoch 610/1000
68s 136ms/step - loss: 0.1739 - acc: 0.9931 - val_loss: 0.3887 - val_acc: 0.9294
Epoch 611/1000
68s 136ms/step - loss: 0.1731 - acc: 0.9934 - val_loss: 0.3874 - val_acc: 0.9311
Epoch 612/1000
68s 136ms/step - loss: 0.1725 - acc: 0.9935 - val_loss: 0.3898 - val_acc: 0.9297
Epoch 613/1000
68s 135ms/step - loss: 0.1717 - acc: 0.9937 - val_loss: 0.3900 - val_acc: 0.9298
Epoch 614/1000
68s 136ms/step - loss: 0.1705 - acc: 0.9937 - val_loss: 0.3912 - val_acc: 0.9299
Epoch 615/1000
68s 136ms/step - loss: 0.1709 - acc: 0.9934 - val_loss: 0.3898 - val_acc: 0.9307
Epoch 616/1000
68s 136ms/step - loss: 0.1686 - acc: 0.9948 - val_loss: 0.3905 - val_acc: 0.9311
Epoch 617/1000
68s 136ms/step - loss: 0.1695 - acc: 0.9942 - val_loss: 0.3948 - val_acc: 0.9303
Epoch 618/1000
68s 136ms/step - loss: 0.1688 - acc: 0.9941 - val_loss: 0.3936 - val_acc: 0.9298
Epoch 619/1000
68s 136ms/step - loss: 0.1679 - acc: 0.9945 - val_loss: 0.3950 - val_acc: 0.9290
Epoch 620/1000
68s 136ms/step - loss: 0.1675 - acc: 0.9941 - val_loss: 0.3940 - val_acc: 0.9300
Epoch 621/1000
68s 136ms/step - loss: 0.1651 - acc: 0.9949 - val_loss: 0.3956 - val_acc: 0.9309
Epoch 622/1000
68s 136ms/step - loss: 0.1653 - acc: 0.9951 - val_loss: 0.3950 - val_acc: 0.9306
Epoch 623/1000
68s 136ms/step - loss: 0.1656 - acc: 0.9946 - val_loss: 0.3947 - val_acc: 0.9306
Epoch 624/1000
68s 136ms/step - loss: 0.1644 - acc: 0.9949 - val_loss: 0.3946 - val_acc: 0.9304
Epoch 625/1000
68s 136ms/step - loss: 0.1636 - acc: 0.9951 - val_loss: 0.3944 - val_acc: 0.9296
Epoch 626/1000
68s 136ms/step - loss: 0.1630 - acc: 0.9951 - val_loss: 0.3937 - val_acc: 0.9295
Epoch 627/1000
68s 136ms/step - loss: 0.1630 - acc: 0.9953 - val_loss: 0.3959 - val_acc: 0.9296
Epoch 628/1000
68s 136ms/step - loss: 0.1627 - acc: 0.9954 - val_loss: 0.3939 - val_acc: 0.9289
Epoch 629/1000
68s 136ms/step - loss: 0.1630 - acc: 0.9947 - val_loss: 0.3937 - val_acc: 0.9303
Epoch 630/1000
68s 135ms/step - loss: 0.1614 - acc: 0.9958 - val_loss: 0.3909 - val_acc: 0.9316
Epoch 631/1000
68s 137ms/step - loss: 0.1624 - acc: 0.9950 - val_loss: 0.3922 - val_acc: 0.9310
Epoch 632/1000
68s 135ms/step - loss: 0.1611 - acc: 0.9954 - val_loss: 0.3907 - val_acc: 0.9313
Epoch 633/1000
68s 136ms/step - loss: 0.1599 - acc: 0.9955 - val_loss: 0.3893 - val_acc: 0.9295
Epoch 634/1000
68s 136ms/step - loss: 0.1600 - acc: 0.9954 - val_loss: 0.3886 - val_acc: 0.9308
Epoch 635/1000
68s 136ms/step - loss: 0.1593 - acc: 0.9953 - val_loss: 0.3926 - val_acc: 0.9297
Epoch 636/1000
68s 136ms/step - loss: 0.1594 - acc: 0.9950 - val_loss: 0.3945 - val_acc: 0.9289
Epoch 637/1000
68s 136ms/step - loss: 0.1595 - acc: 0.9955 - val_loss: 0.3937 - val_acc: 0.9306
Epoch 638/1000
68s 136ms/step - loss: 0.1591 - acc: 0.9958 - val_loss: 0.3882 - val_acc: 0.9306
Epoch 639/1000
68s 135ms/step - loss: 0.1586 - acc: 0.9959 - val_loss: 0.3893 - val_acc: 0.9309
Epoch 640/1000
68s 136ms/step - loss: 0.1588 - acc: 0.9956 - val_loss: 0.3935 - val_acc: 0.9300
Epoch 641/1000
68s 135ms/step - loss: 0.1571 - acc: 0.9960 - val_loss: 0.3917 - val_acc: 0.9298
Epoch 642/1000
68s 136ms/step - loss: 0.1576 - acc: 0.9956 - val_loss: 0.3945 - val_acc: 0.9284
Epoch 643/1000
68s 136ms/step - loss: 0.1570 - acc: 0.9961 - val_loss: 0.3899 - val_acc: 0.9309
Epoch 644/1000
68s 136ms/step - loss: 0.1565 - acc: 0.9962 - val_loss: 0.3918 - val_acc: 0.9307
Epoch 645/1000
68s 136ms/step - loss: 0.1563 - acc: 0.9956 - val_loss: 0.3940 - val_acc: 0.9307
Epoch 646/1000
68s 136ms/step - loss: 0.1563 - acc: 0.9956 - val_loss: 0.3895 - val_acc: 0.9322
Epoch 647/1000
68s 136ms/step - loss: 0.1555 - acc: 0.9963 - val_loss: 0.3903 - val_acc: 0.9302
Epoch 648/1000
68s 135ms/step - loss: 0.1556 - acc: 0.9958 - val_loss: 0.3926 - val_acc: 0.9307
Epoch 649/1000
68s 135ms/step - loss: 0.1542 - acc: 0.9962 - val_loss: 0.3904 - val_acc: 0.9308
Epoch 650/1000
68s 136ms/step - loss: 0.1552 - acc: 0.9959 - val_loss: 0.3934 - val_acc: 0.9295
Epoch 651/1000
68s 136ms/step - loss: 0.1548 - acc: 0.9959 - val_loss: 0.3921 - val_acc: 0.9307
Epoch 652/1000
68s 136ms/step - loss: 0.1537 - acc: 0.9964 - val_loss: 0.3973 - val_acc: 0.9293
Epoch 653/1000
68s 136ms/step - loss: 0.1540 - acc: 0.9958 - val_loss: 0.3950 - val_acc: 0.9287
Epoch 654/1000
68s 136ms/step - loss: 0.1523 - acc: 0.9965 - val_loss: 0.3956 - val_acc: 0.9296
Epoch 655/1000
68s 137ms/step - loss: 0.1532 - acc: 0.9964 - val_loss: 0.3991 - val_acc: 0.9292
Epoch 656/1000
68s 136ms/step - loss: 0.1538 - acc: 0.9957 - val_loss: 0.3995 - val_acc: 0.9296
Epoch 657/1000
68s 136ms/step - loss: 0.1520 - acc: 0.9966 - val_loss: 0.3988 - val_acc: 0.9310
Epoch 658/1000
68s 136ms/step - loss: 0.1532 - acc: 0.9959 - val_loss: 0.3961 - val_acc: 0.9307
Epoch 659/1000
68s 136ms/step - loss: 0.1526 - acc: 0.9958 - val_loss: 0.3948 - val_acc: 0.9306
Epoch 660/1000
68s 136ms/step - loss: 0.1512 - acc: 0.9965 - val_loss: 0.3947 - val_acc: 0.9309
Epoch 661/1000
68s 136ms/step - loss: 0.1519 - acc: 0.9962 - val_loss: 0.3959 - val_acc: 0.9315
Epoch 662/1000
68s 136ms/step - loss: 0.1510 - acc: 0.9963 - val_loss: 0.3962 - val_acc: 0.9312
Epoch 663/1000
68s 136ms/step - loss: 0.1517 - acc: 0.9960 - val_loss: 0.3939 - val_acc: 0.9304
Epoch 664/1000
68s 135ms/step - loss: 0.1494 - acc: 0.9964 - val_loss: 0.3928 - val_acc: 0.9309
Epoch 665/1000
68s 135ms/step - loss: 0.1492 - acc: 0.9966 - val_loss: 0.3900 - val_acc: 0.9320
Epoch 666/1000
68s 136ms/step - loss: 0.1493 - acc: 0.9963 - val_loss: 0.3907 - val_acc: 0.9312
Epoch 667/1000
68s 136ms/step - loss: 0.1491 - acc: 0.9967 - val_loss: 0.3930 - val_acc: 0.9309
Epoch 668/1000
68s 136ms/step - loss: 0.1494 - acc: 0.9960 - val_loss: 0.3923 - val_acc: 0.9301
Epoch 669/1000
68s 136ms/step - loss: 0.1485 - acc: 0.9966 - val_loss: 0.3941 - val_acc: 0.9308
Epoch 670/1000
68s 135ms/step - loss: 0.1486 - acc: 0.9963 - val_loss: 0.3927 - val_acc: 0.9314
...
Epoch 879/1000
68s 137ms/step - loss: 0.1065 - acc: 0.9977 - val_loss: 0.3767 - val_acc: 0.9280
Epoch 880/1000
69s 138ms/step - loss: 0.1055 - acc: 0.9979 - val_loss: 0.3741 - val_acc: 0.9285
Epoch 881/1000
68s 137ms/step - loss: 0.1056 - acc: 0.9979 - val_loss: 0.3716 - val_acc: 0.9290
Epoch 882/1000
69s 138ms/step - loss: 0.1061 - acc: 0.9977 - val_loss: 0.3736 - val_acc: 0.9295
Epoch 883/1000
69s 138ms/step - loss: 0.1066 - acc: 0.9976 - val_loss: 0.3745 - val_acc: 0.9307
Epoch 884/1000
69s 137ms/step - loss: 0.1059 - acc: 0.9975 - val_loss: 0.3702 - val_acc: 0.9302
Epoch 885/1000
69s 138ms/step - loss: 0.1051 - acc: 0.9979 - val_loss: 0.3656 - val_acc: 0.9311
Epoch 886/1000
68s 137ms/step - loss: 0.1051 - acc: 0.9978 - val_loss: 0.3677 - val_acc: 0.9305
Epoch 887/1000
68s 137ms/step - loss: 0.1062 - acc: 0.9974 - val_loss: 0.3636 - val_acc: 0.9315
Epoch 888/1000
69s 137ms/step - loss: 0.1052 - acc: 0.9977 - val_loss: 0.3710 - val_acc: 0.9295
Epoch 889/1000
68s 137ms/step - loss: 0.1046 - acc: 0.9979 - val_loss: 0.3642 - val_acc: 0.9318
Epoch 890/1000
69s 138ms/step - loss: 0.1051 - acc: 0.9975 - val_loss: 0.3673 - val_acc: 0.9306
Epoch 891/1000
69s 138ms/step - loss: 0.1045 - acc: 0.9978 - val_loss: 0.3681 - val_acc: 0.9299
Epoch 892/1000
68s 137ms/step - loss: 0.1043 - acc: 0.9979 - val_loss: 0.3659 - val_acc: 0.9320
Epoch 893/1000
69s 137ms/step - loss: 0.1040 - acc: 0.9979 - val_loss: 0.3627 - val_acc: 0.9326
Epoch 894/1000
69s 138ms/step - loss: 0.1041 - acc: 0.9976 - val_loss: 0.3698 - val_acc: 0.9301
Epoch 895/1000
68s 137ms/step - loss: 0.1039 - acc: 0.9978 - val_loss: 0.3659 - val_acc: 0.9321
Epoch 896/1000
69s 137ms/step - loss: 0.1040 - acc: 0.9978 - val_loss: 0.3718 - val_acc: 0.9300
Epoch 897/1000
68s 137ms/step - loss: 0.1039 - acc: 0.9977 - val_loss: 0.3728 - val_acc: 0.9311
Epoch 898/1000
68s 137ms/step - loss: 0.1044 - acc: 0.9973 - val_loss: 0.3743 - val_acc: 0.9313
Epoch 899/1000
69s 137ms/step - loss: 0.1036 - acc: 0.9976 - val_loss: 0.3675 - val_acc: 0.9312
Epoch 900/1000
69s 138ms/step - loss: 0.1030 - acc: 0.9979 - val_loss: 0.3730 - val_acc: 0.9313
Epoch 901/1000
lr changed to 9.999999310821295e-05
69s 138ms/step - loss: 0.1023 - acc: 0.9982 - val_loss: 0.3709 - val_acc: 0.9310
Epoch 902/1000
69s 137ms/step - loss: 0.1025 - acc: 0.9979 - val_loss: 0.3690 - val_acc: 0.9311
Epoch 903/1000
68s 137ms/step - loss: 0.1024 - acc: 0.9980 - val_loss: 0.3679 - val_acc: 0.9311
Epoch 904/1000
69s 137ms/step - loss: 0.1020 - acc: 0.9982 - val_loss: 0.3673 - val_acc: 0.9315
Epoch 905/1000
69s 138ms/step - loss: 0.1027 - acc: 0.9979 - val_loss: 0.3672 - val_acc: 0.9310
Epoch 906/1000
69s 138ms/step - loss: 0.1015 - acc: 0.9984 - val_loss: 0.3678 - val_acc: 0.9304
Epoch 907/1000
69s 138ms/step - loss: 0.1016 - acc: 0.9984 - val_loss: 0.3673 - val_acc: 0.9302
Epoch 908/1000
69s 138ms/step - loss: 0.1031 - acc: 0.9977 - val_loss: 0.3667 - val_acc: 0.9307
Epoch 909/1000
69s 139ms/step - loss: 0.1019 - acc: 0.9983 - val_loss: 0.3672 - val_acc: 0.9317
Epoch 910/1000
69s 137ms/step - loss: 0.1018 - acc: 0.9983 - val_loss: 0.3671 - val_acc: 0.9313
Epoch 911/1000
69s 137ms/step - loss: 0.1018 - acc: 0.9982 - val_loss: 0.3669 - val_acc: 0.9309
Epoch 912/1000
69s 137ms/step - loss: 0.1014 - acc: 0.9986 - val_loss: 0.3677 - val_acc: 0.9303
Epoch 913/1000
68s 137ms/step - loss: 0.1015 - acc: 0.9982 - val_loss: 0.3666 - val_acc: 0.9303
Epoch 914/1000
69s 138ms/step - loss: 0.1015 - acc: 0.9984 - val_loss: 0.3659 - val_acc: 0.9309
Epoch 915/1000
69s 138ms/step - loss: 0.1013 - acc: 0.9983 - val_loss: 0.3651 - val_acc: 0.9318
Epoch 916/1000
69s 138ms/step - loss: 0.1014 - acc: 0.9983 - val_loss: 0.3652 - val_acc: 0.9322
Epoch 917/1000
69s 137ms/step - loss: 0.1010 - acc: 0.9984 - val_loss: 0.3648 - val_acc: 0.9322
Epoch 918/1000
68s 137ms/step - loss: 0.1016 - acc: 0.9981 - val_loss: 0.3644 - val_acc: 0.9324
Epoch 919/1000
69s 138ms/step - loss: 0.1013 - acc: 0.9983 - val_loss: 0.3635 - val_acc: 0.9319
...
Epoch 981/1000
69s 137ms/step - loss: 0.0992 - acc: 0.9987 - val_loss: 0.3619 - val_acc: 0.9324
Epoch 982/1000
68s 137ms/step - loss: 0.0994 - acc: 0.9986 - val_loss: 0.3619 - val_acc: 0.9332
Epoch 983/1000
69s 137ms/step - loss: 0.0995 - acc: 0.9986 - val_loss: 0.3617 - val_acc: 0.9329
Epoch 984/1000
68s 137ms/step - loss: 0.0991 - acc: 0.9987 - val_loss: 0.3622 - val_acc: 0.9328
Epoch 985/1000
68s 137ms/step - loss: 0.0991 - acc: 0.9987 - val_loss: 0.3628 - val_acc: 0.9322
Epoch 986/1000
68s 137ms/step - loss: 0.0993 - acc: 0.9987 - val_loss: 0.3625 - val_acc: 0.9319
Epoch 987/1000
68s 137ms/step - loss: 0.0995 - acc: 0.9986 - val_loss: 0.3629 - val_acc: 0.9317
Epoch 988/1000
69s 137ms/step - loss: 0.0993 - acc: 0.9985 - val_loss: 0.3628 - val_acc: 0.9319
Epoch 989/1000
69s 137ms/step - loss: 0.0997 - acc: 0.9984 - val_loss: 0.3624 - val_acc: 0.9322
Epoch 990/1000
69s 138ms/step - loss: 0.0993 - acc: 0.9986 - val_loss: 0.3622 - val_acc: 0.9323
Epoch 991/1000
68s 137ms/step - loss: 0.0993 - acc: 0.9986 - val_loss: 0.3625 - val_acc: 0.9327
Epoch 992/1000
69s 137ms/step - loss: 0.0993 - acc: 0.9988 - val_loss: 0.3630 - val_acc: 0.9325
Epoch 993/1000
68s 137ms/step - loss: 0.0992 - acc: 0.9984 - val_loss: 0.3634 - val_acc: 0.9320
Epoch 994/1000
69s 138ms/step - loss: 0.0991 - acc: 0.9988 - val_loss: 0.3627 - val_acc: 0.9328
Epoch 995/1000
69s 138ms/step - loss: 0.0989 - acc: 0.9989 - val_loss: 0.3637 - val_acc: 0.9321
Epoch 996/1000
69s 138ms/step - loss: 0.0994 - acc: 0.9986 - val_loss: 0.3623 - val_acc: 0.9319
Epoch 997/1000
69s 138ms/step - loss: 0.0987 - acc: 0.9987 - val_loss: 0.3622 - val_acc: 0.9322
Epoch 998/1000
69s 138ms/step - loss: 0.0989 - acc: 0.9988 - val_loss: 0.3621 - val_acc: 0.9325
Epoch 999/1000
69s 138ms/step - loss: 0.0993 - acc: 0.9984 - val_loss: 0.3615 - val_acc: 0.9326
Epoch 1000/1000
69s 138ms/step - loss: 0.0986 - acc: 0.9988 - val_loss: 0.3614 - val_acc: 0.9323
Train loss: 0.09943642792105675
Train accuracy: 0.9982600016593933
Test loss: 0.3614072059094906
Test accuracy: 0.9322999995946885在使用了shear_range = 30的数据增强以后，准确率降了呢。
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
