Mercurial > hg > camir-aes2014
annotate toolboxes/FullBNT-1.0.7/KPMstats/gaussian_prob.m @ 0:e9a9cd732c1e tip
first hg version after svn
| author | wolffd |
|---|---|
| date | Tue, 10 Feb 2015 15:05:51 +0000 |
| parents | |
| children |
| rev | line source |
|---|---|
| wolffd@0 | 1 function p = gaussian_prob(x, m, C, use_log) |
| wolffd@0 | 2 % GAUSSIAN_PROB Evaluate a multivariate Gaussian density. |
| wolffd@0 | 3 % p = gaussian_prob(X, m, C) |
| wolffd@0 | 4 % p(i) = N(X(:,i), m, C) where C = covariance matrix and each COLUMN of x is a datavector |
| wolffd@0 | 5 |
| wolffd@0 | 6 % p = gaussian_prob(X, m, C, 1) returns log N(X(:,i), m, C) (to prevents underflow). |
| wolffd@0 | 7 % |
| wolffd@0 | 8 % If X has size dxN, then p has size Nx1, where N = number of examples |
| wolffd@0 | 9 |
| wolffd@0 | 10 if nargin < 4, use_log = 0; end |
| wolffd@0 | 11 |
| wolffd@0 | 12 if length(m)==1 % scalar |
| wolffd@0 | 13 x = x(:)'; |
| wolffd@0 | 14 end |
| wolffd@0 | 15 [d N] = size(x); |
| wolffd@0 | 16 %assert(length(m)==d); % slow |
| wolffd@0 | 17 m = m(:); |
| wolffd@0 | 18 M = m*ones(1,N); % replicate the mean across columns |
| wolffd@0 | 19 denom = (2*pi)^(d/2)*sqrt(abs(det(C))); |
| wolffd@0 | 20 mahal = sum(((x-M)'*inv(C)).*(x-M)',2); % Chris Bregler's trick |
| wolffd@0 | 21 if any(mahal<0) |
| wolffd@0 | 22 warning('mahal < 0 => C is not psd') |
| wolffd@0 | 23 end |
| wolffd@0 | 24 if use_log |
| wolffd@0 | 25 p = -0.5*mahal - log(denom); |
| wolffd@0 | 26 else |
| wolffd@0 | 27 p = exp(-0.5*mahal) / (denom+eps); |
| wolffd@0 | 28 end |