www.gusucode.com > emg_functions > emg_functions/classify_online/KLDpredict.m
%% Betthauser - 2016 -- Compute KLDiv algorithm accuracy from data. % INPUTS: Atrain - training data (numF x N) % trainlabels - training data labels (1 x N) % Atest - testing data (numF x P) % testlabels - testing data labels (1 x P) % % OUTPUT: [KLDacc, predict] - KLDiv algorithm accuracy, predictions function [KLDpredict, kldivs] = KLDpredict(trainMeansPDF, trainMax, testsample) numClasses = size(trainMeansPDF,2); testsample = testsample./trainMax; testsample = testsample / sum(testsample); %% KLD kldivs = ones(numClasses,1); minKL = 1e300; % init to impossibly high number for j = 1:numClasses kldivs(j) = KL_divergence(testsample,trainMeansPDF(:,j)); if kldivs(j) < minKL minKL = kldivs(j); KLDpredict = j; end end end