weight.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：294
校验个数：300
每个节拍最大训练次数：5000
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：
-0.121212	-0.088908	0.184828	
0.159803	-0.12756	-0.00669118	
0.0894116	-0.0302362	-0.061243	
0.103679	-0.0631046	0.0722449	
0.154462	0.0132221	0.160459	
0.1788	0.101894	0.0737556	

0.158094	-0.0148396	-0.225829	-0.185057	0.0636082	0.171323	
0.137494	0.0700171	-0.0676672	0.15643	-0.0712989	-0.104869	
0.0452208	-0.000694296	0.0165944	-0.201476	-0.215911	0.0849712	
-0.0954848	0.0447172	-0.218612	-0.0120014	0.202971	0.0496155	
-0.202544	0.153272	0.149823	0.132046	0.0283746	-0.215316	
0.00664541	0.137341	-0.202223	0.0825449	-0.0908002	0.198637	

-0.164594	0.220122	-8.39259e-005	0.151471	0.0958205	0.20024	

调整前阈值：
-0.0789895	-0.132641	-0.179807	-0.0177694	-0.00415815	0.0219352	

-0.20558	

调整后权值：
0.0493498	-0.908835	0.139279	
1.35686	-1.40949	-0.799476	
-1.33827	0.167077	0.0156122	
0.323169	-1.09573	-0.715643	
-0.94437	-1.23043	-0.935895	
1.76682	-1.38254	-0.780161	


3.30988	-2.42481	0.285893	-1.45849	-0.741197	-2.68583	
-0.478209	2.28482	-2.47983	1.1066	-0.505994	1.72675	
-1.31643	-2.08728	1.9983	-1.45728	0.0443456	-1.82495	
-0.866905	0.828239	-1.56358	0.140699	-0.33274	0.703738	
-0.711272	-0.149036	0.638453	-0.0931734	0.202535	-0.555483	
-0.585831	3.55898	-2.93687	1.73756	-0.268587	3.16322	


0.753175	1.00512	-1.64454	0.288834	-1.08365	1.26053	

调整后阈值：
1.202	1.80757	0.641994	1.49075	0.98448	2.04828	

-0.415224