Description of slverifyroc

Home > sltoolbox > perfeval > slverifyroc.m

slverifyroc

PURPOSE

SLVERIFYROC Computes the verification ROC

SYNOPSIS

function [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op, npts)

DESCRIPTION

SLVERIFYROC Computes the verification ROC

 $ Syntax $
   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op)
   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op, npts)
 
 $ Arguments $
   - scores:       the score matrix for verification
   - labels1:      the labels of referred samples
   - labels2:      the labels of query samples
   - op:           the option stating the attributes of the scores
   - npts:         the number of threshold points (default = 200)
   - thrs:         the sampled thresholds
   - fars:         the false accept rates at the sampled thresholds
   - frrs:         the false reject rates at the sampled thresholds

 $ Description $
   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op) 
     Computes the verification ROC based on the pairwise scores. 
     If there are m samples in the first set, n samples in the second set, 
     then the scores matrix should be of size m x n, and the labels1 and 
     labels2 should be a vector of m and n elements respectively. 
     If labels2 is given empty, then it is supposed to be the same as 
     labels1, where we are performing self-pairwise-verification.
     Here op can be 'low' or 'high', if op is 'high' means that a higher
     score indicating higher similarity, vice versa for 'low' op. 
     For output, if n thresholds are sampled, then thrs, fars, and frrs
     for all n x 1 vectors, containing the sampled threshold values,
     and the corresponding false accept rates, and false reject rates.

   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op, npts)
     You can specify the number of threshold points to sampled by npts.

 $ History $
   - Created by Dahua Lin on Jun 10th, 2005
   - Modified by Dahua Lin on May 1st, 2006
     - Base on the sltoolbox v4
   - Modified by Dahua Lin on Aug 8th, 2006
     - Add one more argument npts to tune the density of sampling

CROSS-REFERENCE INFORMATION

This function calls:

• slroc SLROC Computes the ROC
• raise_lackinput RAISE_LACKINPUT Raises an error indicating lack of input argument

This function is called by:

SOURCE CODE

0001 function [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op, npts)
0002 %SLVERIFYROC Computes the verification ROC
0003 %
0004 % $ Syntax $
0005 %   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op)
0006 %   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op, npts)
0007 %
0008 % $ Arguments $
0009 %   - scores:       the score matrix for verification
0010 %   - labels1:      the labels of referred samples
0011 %   - labels2:      the labels of query samples
0012 %   - op:           the option stating the attributes of the scores
0013 %   - npts:         the number of threshold points (default = 200)
0014 %   - thrs:         the sampled thresholds
0015 %   - fars:         the false accept rates at the sampled thresholds
0016 %   - frrs:         the false reject rates at the sampled thresholds
0017 %
0018 % $ Description $
0019 %   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op)
0020 %     Computes the verification ROC based on the pairwise scores.
0021 %     If there are m samples in the first set, n samples in the second set,
0022 %     then the scores matrix should be of size m x n, and the labels1 and
0023 %     labels2 should be a vector of m and n elements respectively.
0024 %     If labels2 is given empty, then it is supposed to be the same as
0025 %     labels1, where we are performing self-pairwise-verification.
0026 %     Here op can be 'low' or 'high', if op is 'high' means that a higher
0027 %     score indicating higher similarity, vice versa for 'low' op.
0028 %     For output, if n thresholds are sampled, then thrs, fars, and frrs
0029 %     for all n x 1 vectors, containing the sampled threshold values,
0030 %     and the corresponding false accept rates, and false reject rates.
0031 %
0032 %   - [thrs, fars, frrs] = slverifyroc(scores, labels1, labels2, op, npts)
0033 %     You can specify the number of threshold points to sampled by npts.
0034 %
0035 % $ History $
0036 %   - Created by Dahua Lin on Jun 10th, 2005
0037 %   - Modified by Dahua Lin on May 1st, 2006
0038 %     - Base on the sltoolbox v4
0039 %   - Modified by Dahua Lin on Aug 8th, 2006
0040 %     - Add one more argument npts to tune the density of sampling
0041 %
0042 
0043 %% parse and verify input arguments
0044 if nargin < 4
0045     raise_lackinput('slverifyroc', 4);
0046 end
0047 if isempty(labels2)
0048     labels2 = labels1;
0049 end
0050 if ndims(scores) ~= 2
0051     error('sltoolbox:invaliddims', ...
0052         'The scores should be a 2D matrix');
0053 end
0054 [m, n] = size(scores);
0055 if length(labels1) ~= m || length(labels2) ~= n 
0056     error('The sizes of labels do not match that of scores');
0057 end
0058 
0059 if nargin < 5 || isempty(npts)
0060     npts = 200;
0061 end
0062 
0063 
0064 %% generate the signs
0065 labels1 = labels1(:);
0066 labels2 = labels2(:)';
0067 L1 = labels1(:, ones(1, n));
0068 L2 = labels2(ones(m, 1), :);
0069 signals = (L1 == L2);
0070 clear L1 L2;
0071 
0072 %% compute
0073 [thrs, fars, frrs] = slroc(scores, signals, npts, op);
0074 
0075 
0076 
0077

---

Generated on Wed 20-Sep-2006 12:43:11 by m2html © 2003