


This function normalises the kernel matrix such that each entry is divided by the product of the std deviations, i.e. K_new(x,y) = K(x,y) / sqrt(var(x)*var(y)) __________________________________________________________________________ Copyright (C) 2011 Machine Learning & Neuroimaging Laboratory


0001 function C = prt_normalise_kernel(K) 0002 0003 % This function normalises the kernel matrix such that each entry is 0004 % divided by the product of the std deviations, i.e. 0005 % K_new(x,y) = K(x,y) / sqrt(var(x)*var(y)) 0006 %__________________________________________________________________________ 0007 % Copyright (C) 2011 Machine Learning & Neuroimaging Laboratory 0008 0009 % Written by A. Marquand 0010 % Id: $ 0011 0012 d = diag(K); 0013 K0 = sqrt(repmat(d,[1,size(K,1)]).* repmat(d',[size(K,1),1])); 0014 C = K./K0; 0015 0016 end