理论和代码-RNN-代码

学习根据sin预测cos的函数曲线，序列长度20，具体见代码

# coding=utf-8

import os
os.environ['KERAS_BACKEND'] = 'tensorflow'

import matplotlib.pyplot as plt
import numpy as np

from keras.models import Sequential
from keras.layers import LSTM, TimeDistributed, Dense
from keras.optimizers import Adam

BATCH_START = 0
TIME_STEPS = 20
BATCH_SIZE = 50

INPUT_DIM = 1
OUTPUT_SIZE = 1
UNITS = 20

LR = 0.008

def get_batch():
    global BATCH_START
    xs = np.arange(BATCH_START, BATCH_START + BATCH_SIZE * TIME_STEPS).reshape(BATCH_SIZE, TIME_STEPS) * np.pi / 180
    #print xs[0, :]
    seq = np.sin(xs)
    res = np.cos(xs)
    #print res[0, :]
    BATCH_START += TIME_STEPS
    #plt.plot(xs[0, :], res[0, :], 'r')#, xs[0, :], seq[0, :], 'b--')
    #plt.plot(xs[0, :], res[0, :], 'r', label='cos')
    #plt.plot(xs[0, :], seq[0, :], 'b', label='sin')
    #plt.show()
    #print seq[:, :, np.newaxis].shape
    #print res[:, :, np.newaxis].shape
    #print xs.shape
    return [seq[:, :, np.newaxis], res[:, :, np.newaxis], xs]

#get_batch()

model = Sequential()
# build LSTM RNN
model.add(LSTM(
    batch_input_shape=(BATCH_SIZE, TIME_STEPS, INPUT_DIM), # TIME_STEPS is input_length, INPUT_DIM is input_dim
    #return_sequences is true, (BATCH_SIZE, TIME_STEPS, INPUT_DIM)--->(BATCH_SIZE, TIME_STEPS, UNITS);
    #return_sequences is false, (BATCH_SIZE, TIME_STEPS, INPUT_DIM)-->(BATCH_SIZE, UNITS)
    output_dim=UNITS,
    return_sequences = True, # True output at all steps/all sequences, False output as last step/last output
    stateful = True)) # True the final state or batch1 is feed into the initial state of the batch2
# add output layer
model.add(TimeDistributed(Dense(OUTPUT_SIZE))) # 
model.compile(optimizer=Adam(LR), loss='mse')

for step in range(200):
    x_batch, y_batch, xs = get_batch()
    cost = model.train_on_batch(x_batch, y_batch)
    predict = model.predict(x_batch, BATCH_SIZE)

    plt.plot(xs[0, :], y_batch[0].flatten(), 'r', xs[0, :], predict.flatten()[:TIME_STEPS], 'b--')
    plt.ylim(-1.5, 1.5)
    plt.draw()
    plt.pause(0.2)
    if step % 10 == 0:
        print('cost:', cost)