Mercurial > hg > ltpda
view m-toolbox/classes/+utils/@math/loglikelihood_td.m @ 11:9174aadb93a5 database-connection-manager
Add LTPDA Repository utility functions into utils.repository
author | Daniele Nicolodi <nicolodi@science.unitn.it> |
---|---|
date | Mon, 05 Dec 2011 16:20:06 +0100 |
parents | f0afece42f48 |
children |
line wrap: on
line source
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % % Compute log-likelihood in time domain assuming a multivariate gaussian % distribution % % INPUT % % - res, a vector of AOs containing residuals % - noise, a vector of AOs containing noise % - cutbefore, followed by the data samples to cut at the starting of the % data series % - cutafter, followed by the data samples to cut at the ending of the % data series % % L Ferraioli 10-10-2010 % % $Id: loglikelihood_td.m,v 1.1 2011/03/15 16:19:20 miquel Exp $ % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function loglk = loglikelihood_td(res,noise,varargin) cutbefore = []; cutafter = []; if ~isempty(varargin) for j=1:length(varargin) if strcmp(varargin{j},'cutbefore') cutbefore = varargin{j+1}; end if strcmp(varargin{j},'cutafter') cutafter = varargin{j+1}; end end end nres = numel(res); loglk = 0; for ii=1:nres yres = res(ii).y; ynoise = noise(ii).y; % willing to work with rows if size(yres,2)<size(yres,1) yres = yres.'; end if size(ynoise,2)<size(ynoise,1) ynoise = ynoise.'; end if ~isempty(cutbefore) yres(1:cutbefore) = []; ynoise(1:cutbefore) = []; end if ~isempty(cutafter) yres(end-cutafter:end) = []; ynoise(end-cutafter:end) = []; end % R = utils.math.Rcovmat(yres); % % xx = R\yres.'; % Ndim = numel(yres); % % loglk = loglk + abs(xx'*xx)./Ndim; cmat = utils.math.xCovmat(ynoise); [L,p] = chol(cmat,'lower'); if p==0 xx = L\yres.'; Ndim = numel(yres); else q = p-1; yyres = yres(1:q); xx = L\yyres.'; Ndim = numel(yyres); end loglk = loglk + abs(xx'*xx)./Ndim; end %%% get cros terms if nres>1 for ii=1:nres-1 for jj=ii+1:nres yres1 = res(ii).y; yres2 = res(jj).y; ynoise1 = noise(ii).y; ynoise2 = noise(jj).y; % willing to work with rows if size(yres1,2)<size(yres1,1) yres1 = yres1.'; end if size(yres2,2)<size(yres2,1) yres2 = yres2.'; end if size(ynoise1,2)<size(ynoise1,1) ynoise1 = ynoise1.'; end if size(ynoise2,2)<size(ynoise2,1) ynoise2 = ynoise2.'; end if ~isempty(cutbefore) yres1(1:cutbefore) = []; yres2(1:cutbefore) = []; ynoise1(1:cutbefore) = []; ynoise2(1:cutbefore) = []; end if ~isempty(cutafter) yres1(end-cutafter:end) = []; yres2(end-cutafter:end) = []; ynoise1(end-cutafter:end) = []; ynoise2(end-cutafter:end) = []; end % Rx = utils.math.Rcovmat(yres1); % Ry = utils.math.Rcovmat(yres2); % % xx = Rx\yres1.'; % yy = Ry\yres2.'; % % Ndim = numel(yres1); % % loglk = loglk + 2.*abs(xx'*yy)./Ndim; cmat = utils.math.xCovmat(ynoise1,ynoise2); [L,p] = chol(cmat,'lower'); if p==0 xx = L\yres1.'; Ndim = numel(yres1); yy = L\yres2.'; else q = p-1; yyres1 = yres1(1:q); yyres2 = yres2(1:q); xx = L\yyres1.'; Ndim = numel(yyres1); yy = L\yyres2.'; end loglk = loglk + 2.*abs(xx'*yy)./Ndim; end end end end