视频学习来源
https://www.bilibili.com/video/av40787141?from=search&seid=17003307842787199553
笔记
RNN用于图像识别并不是很好
模型保存(结构和参数)
1 需要安装h5py
pip install h5py
2在代码最后一行
model.save(‘model.h5’)
即可在当前目录保存HDF5文件
模型载入
1开头导入包
from keras.models import load_model
2导入模型
model=load_model(‘model.h5’)
模型载入后可接着训练
model.fit(x_train,y_train,batch_size=64,epochs=2)
只保存参数
model.save_weights(‘weights.h5’)
model.load _weights(‘weights.h5’)
只保存网络结构
from keras.models import model_from_json
json_string=model.to_json()
model=model_from_json(json_string)
import numpy as np from keras.datasets import mnist #将会从网络下载mnist数据集 from keras.utils import np_utils from keras.models import Sequential #序列模型 from keras.layers import Dense from keras.layers.recurrent import SimpleRNN #keras中三种RNN SimpleRNN,LSTM,GRU from keras.optimizers import Adam
# 数据长度,一行有28个像素 input_size=28 # 序列长度,一共有28行 time_steps=28 # 隐藏层cell个数 cell_size=50 #载入数据 (x_train,y_train),(x_test,y_test)=mnist.load_data() #查看格式 #(60000,28,28) print('x_shape:',x_train.shape) #(60000) print('y_shape:',y_train.shape) #格式是(60000,28,28) #格式是样本数,time_steps(序列长度),input_size(每一个序列的数据长度) #如果数据是(60000,784)需要转成(60000,28,28) #除以255是做数据归一化处理 x_train=x_train/255.0 #转换数据格式 x_test=x_test/255.0 #转换数据格式 #label标签转换成 one hot 形式 y_train=np_utils.to_categorical(y_train,num_classes=10) #分成10类 y_test=np_utils.to_categorical(y_test,num_classes=10) #分成10类 #定义序列模型 model=Sequential() #循环神经网络 #一个隐藏层 model.add(SimpleRNN( units=cell_size, #输出 input_shape=(time_steps,input_size), #输入 )) #输出层 model.add(Dense(10,activation='softmax')) #定义优化器 #学习速率为10的负4次方 adam=Adam(lr=1e-4) #定义优化器,损失函数,训练效果中计算准确率 model.compile( optimizer=adam, #sgd优化器 loss='categorical_crossentropy', #损失用交叉熵,速度会更快 metrics=['accuracy'], #计算准确率 ) #训练 #六万张,每次训练64张,训练10个周期(六万张全部训练完算一个周期) model.fit(x_train,y_train,batch_size=64,epochs=10) #评估模型 loss,accuracy=model.evaluate(x_test,y_test) print(' test loss',loss) print(' test accuracy',accuracy) loss,accuracy=model.evaluate(x_train,y_train) print(' train loss',loss) print(' train accuracy',accuracy)
x_shape: (60000, 28, 28) y_shape: (60000,) Epoch 1/10 60000/60000 [==============================] - 9s 145us/step - loss: 1.6191 - acc: 0.4629 Epoch 2/10 60000/60000 [==============================] - 9s 156us/step - loss: 0.9580 - acc: 0.7103 Epoch 3/10 60000/60000 [==============================] - 6s 101us/step - loss: 0.7064 - acc: 0.7934 Epoch 4/10 60000/60000 [==============================] - 8s 141us/step - loss: 0.5749 - acc: 0.8344 Epoch 5/10 60000/60000 [==============================] - 8s 128us/step - loss: 0.4999 - acc: 0.8550 Epoch 6/10 60000/60000 [==============================] - 6s 102us/step - loss: 0.4503 - acc: 0.8689 Epoch 7/10 60000/60000 [==============================] - 6s 99us/step - loss: 0.4130 - acc: 0.8808 Epoch 8/10 60000/60000 [==============================] - 6s 95us/step - loss: 0.3838 - acc: 0.8891 Epoch 9/10 60000/60000 [==============================] - 6s 96us/step - loss: 0.3597 - acc: 0.8969 Epoch 10/10 60000/60000 [==============================] - 6s 96us/step - loss: 0.3408 - acc: 0.9020 10000/10000 [==============================] - 1s 73us/step test loss 0.3126664091944695 test accuracy 0.91 60000/60000 [==============================] - 4s 67us/step train loss 0.326995205249389 train accuracy 0.9060166666666667