weight2.txt

简单易懂的elaman离线学习程序

本程序为采用简化的BPTT算法离线学习的Elman网络：
每个节拍离线训练样本个数：394
校验个数：300
每个节拍最大训练次数：2000
输入层神经元个数：3
隐层层神经元个数：6
状态层神经元个数：6
输出层神经元个数：1
前后学习速率分别是：0.4	0.4	0.8
冲量系数:0.18
记忆系数:0.1

调整前权值：
-0.109871	0.0960005	0.0659154	
-0.128759	0.0272423	0.0415891	
-0.0933088	-0.0284509	0.092613	
0.144076	-0.0889691	-0.0997269	
-0.0502869	-0.0627659	0.0550752	
-0.0534455	0.0285699	0.0888134	

0.112334	0.109743	0.121517	-0.0993789	-0.00922422	-0.124419	
0.0921003	-0.130581	-0.112261	0.0391629	0.143948	-0.132046	
0.0844371	-0.091203	-0.0536195	-0.144635	0.051944	-0.124575	
-0.0361782	-0.0233879	-0.00927915	-0.0181967	0.0857738	0.0600375	
-0.0247246	-0.0241844	0.104222	-0.127441	0.126058	-0.117571	
0.135434	0.141934	-0.0196982	-0.0214652	0.132907	0.00525071	

-0.116143	0.105065	-0.111565	-0.106676	-0.124273	-0.0991867	

调整前阈值：
-0.131854	0.110082	0.0872662	-0.107628	0.080372	-0.0835398	

-0.0720771	

调整后权值：
-1.04558	-0.145783	-0.0146511	
0.266766	-0.721259	-0.999099	
-1.1535	0.0458837	0.390835	
0.119627	-0.514171	-0.728295	
-1.11424	-0.316658	0.019569	
0.0114668	-0.519072	-0.639131	


0.153914	-1.89298	1.39207	-0.883493	0.327729	-1.24736	
-0.976728	2.28949	-3.12009	0.435588	-1.20558	0.693396	
0.276941	-3.17241	1.98881	-1.33358	0.692026	-1.80774	
-0.85959	0.53024	-1.60745	-0.243351	-0.795628	-0.0528212	
0.318084	-2.31933	1.97447	-0.88482	0.83299	-1.29329	
-0.767807	1.05441	-1.91437	-0.158261	-0.878357	0.0610858	


-1.20059	2.05551	-2.01402	1.02707	-1.59348	1.25253	

调整后阈值：
0.774772	0.423083	0.40096	0.683228	0.688944	0.67843	

-0.302143