% [ll, f, cl, P] = normal(data, alpha, mu, sigma)
%
% This function serves several purposes:
%  - evaluate PDF of normal mixture at each of the data points
%  - evaluate log-likelihood of the data under the model (for 
%    out-of-sample evaluation)
%  - return deterministic classification
%  - return ``weights" P (probabilistic classification)
%
% INPUT: 
%    data   : data points: n rows (points), d columns (dimensions)
%    alpha  : component weights (k by 1)
%    mu     : component means (k by d)
%    sigma  : component covariance matrices (d by d by k) or (d by d*k)
%
% OUTPUT:
%    ll     : log likelihood of the data under the model
%    f      : (n by 1) value of (total) PDF at each of the data points
%    cl     : (n by 1) deterministic classification of data points
%    P      : (n by k) probabilistic classification of data points
%
%                            by Igor Cadez (03/05/99)