Error Using Feature Selector

Hey, I´m trying to use sequentialfs to find the more important features to feed my neural network. Here's the code:

in=normc(input);

out2=transpose(output);
out=out2(:,1);

xtrain=in(1:616,:);
xtest=in(617:end,:);
ytrain=out(1:616,:);
ytest=out(617:end,:);

ypred = classify(xtest, xtrain, ytrain);

fun=@(xtrain,ytrain,xtest,ytest) sum(ytest~=classify(xtest,xtrain,ytrain));

inmodel=sequentialfs(fun,in,out);

--------------------------------------------------------------------------------------------------------

The out variable is a 880*1 matrix with the predicted output and the in is a 880*36 matrix with the features.

When I run the program the following errors appear:

Error using classify (line 228)
The pooled covariance matrix of TRAINING must be positive definite.

Error in Example (line 8)
ypred = classify(xtest, xtrain, ytrain);

------------------------------------------------------------------------------------------------------

I checked and the problem is on the matrixes because if i use other two with random numbers it works.

I have no idea how to solve this. Can you help me?

Thanks in advance,

Luís Dias

[Image Analyst]

We don't have your data, so it's hard to help you.

input() is a built-in function. I don't think it's good practice to use input() as the argument to another function. You didn't make the mistake of calling some array your have "input" did you?

[Greg Heath]

The default for CLASSIFY is 'linear'.

Therefore you might have an easier time using either

 PLSREGRESS or STEPWISEFIT.

Although you could add squares and cross-products as extra variables to pick from, I don't recommend it.

Hope this helps.

Greg