RNN Classifier 循环神经网络

问题描述

使用RNN对MNIST里面的图片进行分类

关键

SimpleRNN()参数

batch_input_shape
使用状态RNN的注意事项

可以将RNN设置为‘stateful’，意味着由每个batch计算出的状态都会被重用于初始化下一个batch的初始状态。状态RNN假设连续的两个batch之中，相同下标的元素有一一映射关系。

要启用状态RNN，请在实例化层对象时指定参数stateful=True，并在Sequential模型使用固定大小的batch：通过在模型的第一层传入batch_size=(…)和input_shape来实现。在函数式模型中，对所有的输入都要指定相同的batch_size。

如果要将循环层的状态重置，请调用.reset_states()，对模型调用将重置模型中所有状态RNN的状态。对单个层调用则只重置该层的状态。

（samples，timesteps，input_dim）

代码

'''
RNN Classifier 循环神经网络
'''
import numpy as np
np.random.seed(1337)

from keras.datasets import mnist
from keras.utils import  np_utils
from keras.models import Sequential
from keras.layers import SimpleRNN, Activation, Dense
from keras.optimizers import  Adam

time_step = 28
input_size = 28
batch_size = 50
output_size = 10
cell_size = 50
LR = 0.001

(X_train, y_train), (X_test, y_test) = mnist.load_data()

X_train = X_train.reshape(-1, 28, 28) / 255.      # normalize
X_test = X_test.reshape(-1, 28, 28) / 255.        # normalize
y_train = np_utils.to_categorical(y_train, num_classes=10)
y_test = np_utils.to_categorical(y_test, num_classes=10)

model = Sequential()
model.add(
    SimpleRNN(
        batch_input_shape=(None, time_step, input_size),
        units=cell_size
    )
)

model.add(
    Dense(output_size)
)

model.add(Activation('softmax'))

adam = Adam(LR)
model.compile(
    optimizer=adam,
    loss='categorical_crossentropy',
    metrics=['accuracy']
)
model.summary()

model.fit(X_train, y_train, batch_size=batch_size, epochs=2, verbose=2, validation_data=(X_test, y_test))

结果

Model: "sequential_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
simple_rnn_2 (SimpleRNN)     (None, 50)                3950      
_________________________________________________________________
dense_2 (Dense)              (None, 10)                510       
_________________________________________________________________
activation_2 (Activation)    (None, 10)                0         
=================================================================
Total params: 4,460
Trainable params: 4,460
Non-trainable params: 0
_________________________________________________________________
Train on 60000 samples, validate on 10000 samples
Epoch 1/2
 - 12s - loss: 0.6643 - accuracy: 0.7966 - val_loss: 0.4501 - val_accuracy: 0.8550
Epoch 2/2
 - 9s - loss: 0.3220 - accuracy: 0.9087 - val_loss: 0.2445 - val_accuracy: 0.9359