Gaussian mixture model

[EDIT: 20110610 10:31 CDT - reformat - WDR]

Hello,

I have two sets of training data each with size 800x3. I'd like to fit them to 3 Gaussians and find their mean, covariance and their weight. I'm trying to implement the third demo given in: http://www.mathworks.com/matlabcentral/fileexchange/19630-gaussian-mixture-model-gmm-gaussian-mixture-regression-gmr

but I'm getting the following warning and the program does not converge:

Warning: Matrix is singular, close to singular or badly scaled.
         Results may be inaccurate. RCOND = NaN.
> In gaussPDF at 21
  In EM at 94
Warning: Matrix is singular, close to singular or badly scaled.
         Results may be inaccurate. RCOND = NaN.
> In gaussPDF at 21
  In EM at 72
>>

Can you please tell me what to do?

Answer 1

[HSN]

Hi Bahareh,

The reason of this warning is zero probability in 'GMR\gaussPDF.m'. You can simply add the following line to the end of this function in order to replace the zeros with minimum real number:

prob(prob==0) = realmin;