GMMB_FRACTHRESH Threshold PDF values according to density quantile. MASK = GMMB_FRACTHRESH(pdfmat, histS, thr) pdfmat = N x K matrix of PDF values at N points in K different PDFs (the output of gmmb_pdf) histS = the histS structure (1 x K cell array) created with the bayesS structure that was used to compute PDFs. thr = scalar in the range [0, 1], the density quantile MASK = N x K logical matrix See also GMMB_PDF, GMMB_HIST, GMMB_GENERATEHIST. The recommended way to create histS is with gmmb_generatehist. The output is a logical N x K matrix that tells whether point N is an outlier in distribution K, i.e., it does not belong to the thr-quantile of distribution K. References: [1] Paalanen, P., Kamarainen, J.-K., Ilonen, J., Kälviäinen, H., Feature Representation and Discrimination Based on Gaussian Mixture Model Probability Densities - Practices and Algorithms, Research Report 95, Lappeenranta University of Technology, Department of Information Technology, 2005. Author(s): Pekka Paalanen <pekka.paalanen@lut.fi> Copyright: Bayesian Classifier with Gaussian Mixture Model Pdf functionality is Copyright (C) 2003, 2004 by Pekka Paalanen and Joni-Kristian Kamarainen. $Name: $ $Revision: 1.2 $ $Date: 2005/04/14 10:33:34 $
0001 %GMMB_FRACTHRESH Threshold PDF values according to density quantile. 0002 % 0003 % MASK = GMMB_FRACTHRESH(pdfmat, histS, thr) 0004 % 0005 % pdfmat = N x K matrix of PDF values at N points 0006 % in K different PDFs (the output of gmmb_pdf) 0007 % histS = the histS structure (1 x K cell array) created with 0008 % the bayesS structure that was used to compute PDFs. 0009 % thr = scalar in the range [0, 1], the density quantile 0010 % 0011 % MASK = N x K logical matrix 0012 % 0013 % See also GMMB_PDF, GMMB_HIST, GMMB_GENERATEHIST. 0014 % 0015 % The recommended way to create histS is with gmmb_generatehist. 0016 % The output is a logical N x K matrix that tells whether point N 0017 % is an outlier in distribution K, i.e., it does not belong to the 0018 % thr-quantile of distribution K. 0019 % 0020 % References: 0021 % [1] Paalanen, P., Kamarainen, J.-K., Ilonen, J., Kälviäinen, H., 0022 % Feature Representation and Discrimination Based on Gaussian Mixture Model 0023 % Probability Densities - Practices and Algorithms, Research Report 95, 0024 % Lappeenranta University of Technology, Department of Information 0025 % Technology, 2005. 0026 % 0027 % Author(s): 0028 % Pekka Paalanen <pekka.paalanen@lut.fi> 0029 % 0030 % Copyright: 0031 % 0032 % Bayesian Classifier with Gaussian Mixture Model Pdf 0033 % functionality is Copyright (C) 2003, 2004 by Pekka Paalanen and 0034 % Joni-Kristian Kamarainen. 0035 % 0036 % $Name: $ $Revision: 1.2 $ $Date: 2005/04/14 10:33:34 $ 0037 % 0038 0039 function mask = gmmb_fracthresh(pdfmat, histS, thr); 0040 0041 N = size(pdfmat, 1); 0042 K = size(pdfmat, 2); 0043 0044 thresh = gmmb_frac2lhood(histS, thr*ones(1,K)); 0045 mask = (pdfmat < repmat(thresh, N, 1));