gmmcTrain

Train a GMM classifier

Contents

Syntax

Description

[cPrm, logLike] = gmmcTrain(DS) returns the parameters of a GMM classifier based on the training of the give dataset DS.

• DS: Design dataset
• opt: GMMC options
  • opt.arch.gaussianNum: A column vector indicating no. of Gaussians for each class
  • opt.arch.covType: Type of covariance matrix
  • tmmcOpt.train: Parameters for training each GMM, which can be obtained via gmmcTrain('defaultOpt').
• cPrm: Parameters for GMM classifier
  • cPrm.gmm(i): Parameters for class i, which is modeled as a GMM
    • cPrm.gmm(i).gmmPrm(j).mu: a mean vector of dim x 1 for Gaussian component j
    • cPrm.gmm(i).gmmPrm(j).sigma: a covariance matrix for Gaussian component j
    • cPrm.gmm(i).gmmPrm(j).w: a weighting factor for Gaussian component j
  • cPrm.prior: Vector of priors, or simply the vector holding no. of entries in each class
  • (To obtain the class sizes, you can use "dsClassSize".
• logLike: Vector of log likelihood during training

Example

DS=prData('iris');
DS.input=DS.input(3:4, :);
trainSet.input=DS.input(:, 1:2:end); trainSet.output=DS.output(:, 1:2:end);
 testSet.input=DS.input(:, 2:2:end);  testSet.output=DS.output(:, 2:2:end);
opt=gmmcTrain('defaultOpt');
cPrm=gmmcTrain(trainSet, opt);
cOutput=gmmcEval(trainSet, cPrm);
recogRate1=sum(trainSet.output==cOutput)/length(trainSet.output);
fprintf('Inside-test recog. rate = %g%%\n', recogRate1*100);
cOutput=gmmcEval(testSet, cPrm);
recogRate2=sum(testSet.output==cOutput)/length(testSet.output);
fprintf('Outside-test recog. rate = %g%%\n', recogRate2*100);
TS.hitIndex=find(testSet.output==cOutput);
gmmcPlot(testSet, cPrm, 'decBoundary');

Inside-test recog. rate = 97.3333%
Outside-test recog. rate = 96%

See Also

gmmcEval, gmmEval, gmmTrain, gmmMixNumEstimate.

Published with MATLAB® R2018a