您可以使用单个输入节点和单个输出节点训练LSTM网络,以进行时间序列预测,如下所示:
首先,作为一种好习惯,让我们使用Python3的打印功能:
from __future__ import print_function
然后,创建一个简单的时间序列:
data = [1] * 3 + [2] * 3data *= 3print(data)
[1,1,1,2,2,2,1,1,1,2,2,2,1,1,1,2,2,2]
现在,将此时间序列放入一个受监督的数据集中,其中每个样本的目标是下一个样本:
from pybrain.datasets import SequentialDataSetfrom itertools import cycleds = SequentialDataSet(1, 1)for sample, next_sample in zip(data, cycle(data[1:])): ds.addSample(sample, next_sample)
用1个输入节点,5个LSTM单元和1个输出节点构建一个简单的LSTM网络:
from pybrain.tools.shortcuts import buildNetworkfrom pybrain.structure.modules import LSTMLayernet = buildNetwork(1, 5, 1, hiddenclass=LSTMLayer, outputbias=False, recurrent=True)
训练网络:
from pybrain.supervised import RPropMinusTrainerfrom sys import stdouttrainer = RPropMinusTrainer(net, dataset=ds)train_errors = [] # save errors for plotting laterEPOCHS_PER_CYCLE = 5CYCLES = 100EPOCHS = EPOCHS_PER_CYCLE * CYCLESfor i in xrange(CYCLES): trainer.trainEpochs(EPOCHS_PER_CYCLE) train_errors.append(trainer.testonData()) epoch = (i+1) * EPOCHS_PER_CYCLE print("r epoch {}/{}".format(epoch, EPOCHS), end="") stdout.flush()print()print("final error =", train_errors[-1])
绘制错误(请注意,在这个简单的玩具示例中,我们正在同一数据集上进行测试和培训,这当然不是您要为实际项目做的!):
import matplotlib.pyplot as pltplt.plot(range(0, EPOCHS, EPOCHS_PER_CYCLE), train_errors)plt.xlabel('epoch')plt.ylabel('error')plt.show()
现在,要求网络预测下一个样本:
for sample, target in ds.getSequenceIterator(0): print(" sample = %4.1f" % sample) print("predicted next sample = %4.1f" % net.activate(sample)) print(" actual next sample = %4.1f" % target) print()
(以上代码基于
example_rnn.py和PyBrain文档中的示例)
欢迎分享,转载请注明来源:内存溢出
评论列表(0条)