MATLAB: Kohonen (example from help does not work)

Deep Learning Toolboxexamplekohonenneural networkweight

Hi all, please I need your help !

example (Kohonen) from help does not work

>> p = [.1 .8 .1 .9; .2 .9 .1 .8];

>> net = newc([0 1; 0 1],2);

>> wts = net.IW{1,1}

>> biases = net.b{1};

>> net.trainParam.epochs = 500;

>> net = train(net,p);

>> a = sim(net,p);

>> ac = vec2ind(a); >> ac

ac =

     1     1     1     1

>> wts

wts =

    0.5000    0.5000
    0.5000    0.5000

My weight did not change. in example

wts =

    0.1000    0.1467
    0.8474    0.8525

Maybe I'm doing something wrong?

Best Answer

7.10.0 (R.2010b)

Related Solutions

MATLAB: Newff Create a feed-forward backpropagation network. – Obsoleted in R2010b NNET 7.0.

There are two obsolete versions of newff. Using defaults for an I-H-O MLP,

net1 = newff(minmax(p),[ H O ]); %very obsolete

net2 = newff( p, t, H ); % obsolete

The latter version has the additional defaults of

 1. Removal of constant (zero variance) rows
 2. Mapminmax [-1,1] normalization of input and target 
 3. 70/15/15 train/val/test data division

It is not clear what versions of MATLAB and NNToolbox you have.

However, both obsolete versions should run. In addition, newfit and newpr are versions of the latter newff that are specialized for regression/curve-fitting and classification/pattern-recognition, respectively.

You can find the documentation for old and new functions in the MATLAB website. However, the commands help, doc, and type will yield most of the information you need to know for most functions on your machine.

Hope this helps.

Thank you for formally accepting my answer

Greg

MATLAB: How to initialize the neural network to a set of weights ??

Unlike the older nets (e.g., newfit, newpr, newff,...), you cannot assign weights to the newer networks (e.g., fitnet, patternnet, feedforwardnet,...) unless the net is configured.

There are two ways to configure the net before manually assigning your own initial weights. Both will assign initial weights that you can overwrite:

1. help/doc configure.

net = configure(net, x, t );

2. Train the net for 1 epoch

net.trainParam.epochs= 1.

net = train(net,x,t);