MOGC

Trainable classifier based on Mixture of Gaussians

    W = MOGC(A,N)
    W = A*MOGC([],N,R,S);
    W = A*MOGC(N,R,S);

   INPUT
     A Dataset
     N Number of mixtures (optional; default 2)
     R,S Regularisation parameters, 0 <= R,S <= 1, see QDC
   OUTPUT

Description

For each class j in A a density estimate is made by GAUSSM, using N(j)  mixture components. Using the class prior probabilities they are combined  into a single classifier W. If N is a scalar, this number is applied to  all classes. The relative size of the components is stored in W.DATA.PRIOR.

For small class sizes or for large values of N it may be difficult or  impossible to find the desired number of components. This may result in  relatively long computing times.

Example(s)

 PREX_DENSITY

See also

datasets, mappings, qdc, plotm, testc,