debug Mode when train

Can you help me????

I have code like this

    [imagefile] = imread([direktori,namafile],'bmp');
    filegambar = im2bw(imagefile,0.5);
    imagefile = bwmorph(~filegambar,'thin','Inf');
    P= imagefile(:,1:200)';
    T= imagefile(:,1:200)';
    [pn,meanp,stdp,tn,meant,stdt]=prestd(P,T)
    net=newff(minmax(pn),[4,3,1], {'tansig' 'logsig' 'purelin'},'traingdm');
    net.IW{1,1}
    net.b{1}
    net.LW{2,1}
    net.b{2}
    net.LW{3,2}
    net.b{3}
    net.trainParam.lr = 0.1
    net.trainParam.goal = 0.00001
    net.trainParam.epochs = 500
    net.trainParam.show = 500
    net=train(net,P)

but when i write this code "net=train(net,P)" , the result is nothing, even out of debug mode,

Can show where the error??

and Can you give me a solution for??

I really need your help

Answer 1

[Greg Heath]

    [imagefile] = imread([direktori,namafile],'bmp');
    filegambar = im2bw(imagefile,0.5);
    imagefile = bwmorph(~filegambar,'thin','Inf');
    P= imagefile(:,1:200)';
    T= imagefile(:,1:200)';

WHAT IS THIS NET SUPPOSED TO DO?

WHATEVER IT IS P AND T SHOULD NOT BE EQUAL.

PLEASE EXPLAIN P,T AND THEIR DIMENSIONS.

    [pn,meanp,stdp,tn,meant,stdt]=prestd(P,T)
    net=newff(minmax(pn),[4,3,1], {'tansig' 'logsig' 'purelin'},'traingdm');

YOU ONLY NEED ONE HIDDEN LAYER.

CHOOSE THE NUMBER OF HIDDEN NODES BY TRIAL AND ERROR.

SEARCH THE NEWSGROUP USING

HEATH NEQ NW

FOR A 1-D OUTPUT SIZE(T) = [ 1 N ]

    net.IW{1,1}
    net.b{1}
    net.LW{2,1}
    net.b{2}
    net.LW{3,2}
    net.b{3}
    net.trainParam.lr = 0.1
    net.trainParam.goal = 0.00001
    net.trainParam.epochs = 500
    net.trainParam.show = 500

IT IS USUALLY BEST TO BEGIN WITH DEFAULTS AND THEN CORRECT AS NEEDED.

    net=train(net,P)

NO. YOU HAVE STANDARDIZED THE INPUTS. THEREFORE USE pn.

but when i write this code "net=train(net,P)" , the result is nothing, even out of debug mode,

HARD TO COMMENT ON THAT UNTIL YOU MAKE THE ABOVE CORRECTIONS.

P.S. i DID NOT CHECK THE FIRST 3 COMMANDS. SO, DO SO BEFORE RERUNNING,

HOPE THIS HELPS.

GREG

[lie akira]

thanks a lot, very helpful.... i'm new in pattern recognition...

i'm use backpropagation for that, but i'm so confused..