資源簡介
使用matlab語言,完成遞歸神經(jīng)網(wǎng)絡(luò)程序的編寫(LSTM)
代碼片段和文件信息
%%%?LSTM網(wǎng)絡(luò)結(jié)合實例仿真
%%%?作者:xd.wp
%%%?時間:2016.10.08??12:06
%%?程序說明
%??1、數(shù)據(jù)為7天,四個時間點的空調(diào)功耗,用前三個推測第四個訓(xùn)練,依次類推。第七天作為檢驗
%??2、LSTM網(wǎng)絡(luò)輸入結(jié)點為12,輸出結(jié)點為4個,隱藏結(jié)點18個
clear?all;
clc;
%%?數(shù)據(jù)加載,并歸一化處理
[train_datatest_data]=LSTM_data_process();
data_length=size(train_data1);
data_num=size(train_data2);
%%?網(wǎng)絡(luò)參數(shù)初始化
%?結(jié)點數(shù)設(shè)置
input_num=12;
cell_num=18;
output_num=4;
%?網(wǎng)絡(luò)中門的偏置
bias_input_gate=rand(1cell_num);
bias_forget_gate=rand(1cell_num);
bias_output_gate=rand(1cell_num);
%?ab=1.2;
%?bias_input_gate=ones(1cell_num)/ab;
%?bias_forget_gate=ones(1cell_num)/ab;
%?bias_output_gate=ones(1cell_num)/ab;
%網(wǎng)絡(luò)權(quán)重初始化
ab=20;
weight_input_x=rand(input_numcell_num)/ab;
weight_input_h=rand(output_numcell_num)/ab;
weight_inputgate_x=rand(input_numcell_num)/ab;
weight_inputgate_c=rand(cell_nu
?屬性????????????大小?????日期????時間???名稱
-----------?---------??----------?-----??----
?????文件??????12146??2018-10-11?17:24??LSTM_matlab.m
-----------?---------??----------?-----??----
????????????????12146????????????????????1
評論
共有 條評論