problem train in anfis network

1 次查看(过去 30 天)
siti nur
siti nur 2011-4-8
i am doing face recogntion using anfis. i already train the network but its doesnt follow the target.I already put 1000 epoch and 20 mf to train the network but still not to target. my code is:
input=[trainData(:)];
data=[input target];
init_fis = genfis1(data);
anfis_net1=anfis(data,20, [1000,0,0.01,1.1,1.1],[1,1,1,1]);
anfis_eval1=evalfis(input,anfis_net1);
comp1=[target anfis_eval1]
input and target that i got are:
comp1=
1.0000 0.2379
0 -0.1536
0 0.3950
0 0.2237
0 0.3400
0 0.2628
1.0000 0.3671
0 0.5821
0 0.3832
0 0.4688
0 0.3510
0 0.2667
1.0000 0.2301
0 -0.1526
0 0.3949
0 0.2195
0 0.5111
0 0.2725
1.0000 0.3208
0 0.6746
0 0.4190
0 0.6348
0 0.5834
0 0.2862
1.0000 0.2605
0 0.4826
0 0.4048
0 0.2038
0 0.4601
0 0.2187
There are 840 input and just some of it is nearer with the target. But when I put the target 1.000, all predicted value is equal to the target. Parameters that can be change are mf and epoch only. I already try until 3000 epoch but still not the same with the target. Can somebody suggest to me what should I do.

回答(1 个)

Ahmed
Ahmed 2012-1-18
My dear I think you can try the following:
1. Triangular MF;
2. Zero order Sugeno model instead of first order Sugeno;
3. initial step-size;
4. step-size decrease rate;
5. step-size increase rate;
Hope this will help you

类别

Help CenterFile Exchange 中查找有关 Fuzzy Logic Toolbox 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by