Description of slcumuscore

Home > sltoolbox > perfeval > slcumuscore.m

slcumuscore

PURPOSE

SLCUMUSCORE Computes the cumulative score on multi-class classification

SYNOPSIS

function cs = slcumuscore(scores, clabels, qlabels, op, maxr)

DESCRIPTION

SLCUMUSCORE Computes the cumulative score on multi-class classification

 $ Syntax $
   - cs = slcumuscore(scores, clabels, qlabels, op)
   - cs = slcumuscore(scores, clabels, qlabels, op, maxr)

 $ Arguments $
   - scores:           the scores to support the classification
   - clabels:          the labels of classes
   - qlabels:          the groundtruth of the labels of query samples
   - op:               the option of the score
   - maxr:             the maximum number of ranked classes
   - cs:               the matrix of cumulative scores

 $ Description $
   - cs = slcumuscore(scores, clabels, qlabels, op) Computes the 
     cumulative scores of score-based classification. Suppose there 
     are m classes and n query samples to be classified. Then scores
     should be an m x n matrix with the entry at the i-th row and j-th
     column representing the score of the j-th sample belonging to the
     the i-th class. clabels and qlabels should be length-m and length-n
     vectors respectively. op states the attributes of the scores, which
     takes either of the two values: 'high' or 'low'. If op is 'high', a 
     higher score indicates a better match; if op is 'low', a lower score
     indicates a better match. The cumulative score will be computed
     up to the number of all classes.

   - cs = slcumuscore(scores, clabels, qlabels, op, maxr) computes the
     cumulative scores of score-based classification up to the number 
     of classes specified by maxr.

 $ History $
   - Created by Dahua Lin on Jun 10th, 2005
   - Modified by Dahua Lin on May 1st, 2006
     - Base on the sltoolbox v4

CROSS-REFERENCE INFORMATION

This function calls:

• raise_lackinput RAISE_LACKINPUT Raises an error indicating lack of input argument
• slignorevars SLIGNOREVARS Ignores the input variables

This function is called by:

SOURCE CODE

0001 function cs = slcumuscore(scores, clabels, qlabels, op, maxr)
0002 %SLCUMUSCORE Computes the cumulative score on multi-class classification
0003 %
0004 % $ Syntax $
0005 %   - cs = slcumuscore(scores, clabels, qlabels, op)
0006 %   - cs = slcumuscore(scores, clabels, qlabels, op, maxr)
0007 %
0008 % $ Arguments $
0009 %   - scores:           the scores to support the classification
0010 %   - clabels:          the labels of classes
0011 %   - qlabels:          the groundtruth of the labels of query samples
0012 %   - op:               the option of the score
0013 %   - maxr:             the maximum number of ranked classes
0014 %   - cs:               the matrix of cumulative scores
0015 %
0016 % $ Description $
0017 %   - cs = slcumuscore(scores, clabels, qlabels, op) Computes the
0018 %     cumulative scores of score-based classification. Suppose there
0019 %     are m classes and n query samples to be classified. Then scores
0020 %     should be an m x n matrix with the entry at the i-th row and j-th
0021 %     column representing the score of the j-th sample belonging to the
0022 %     the i-th class. clabels and qlabels should be length-m and length-n
0023 %     vectors respectively. op states the attributes of the scores, which
0024 %     takes either of the two values: 'high' or 'low'. If op is 'high', a
0025 %     higher score indicates a better match; if op is 'low', a lower score
0026 %     indicates a better match. The cumulative score will be computed
0027 %     up to the number of all classes.
0028 %
0029 %   - cs = slcumuscore(scores, clabels, qlabels, op, maxr) computes the
0030 %     cumulative scores of score-based classification up to the number
0031 %     of classes specified by maxr.
0032 %
0033 % $ History $
0034 %   - Created by Dahua Lin on Jun 10th, 2005
0035 %   - Modified by Dahua Lin on May 1st, 2006
0036 %     - Base on the sltoolbox v4
0037 %
0038 
0039 %% parse and verify input arguments
0040 if nargin < 4
0041     raise_lackinput('slcumuscore', 4);
0042 end
0043 if ndims(scores) ~= 2
0044     error('sltoolbox:invaliddims', ...
0045         'The matrix scores should be a 2D matrix');
0046 end
0047 [nclasses, nsamples] = size(scores);
0048 qlabels = qlabels(:);
0049 clabels = clabels(:);
0050 if length(clabels) ~= nclasses || length(qlabels) ~= nsamples
0051     error('sltoolbox:sizmismatch', ...
0052         'the labels vectors do not match the size of scores');
0053 end
0054 
0055 if nargin < 5 || isempty(maxr)
0056     maxr = nclasses;
0057 end
0058 
0059 %% compute
0060 switch op
0061     case 'low'
0062         [ss, sp] = sort(scores, 1, 'ascend');
0063     case 'high'
0064         [ss, sp] = sort(scores, 1, 'descend');
0065     otherwise
0066         error('sltoolbox:invalidarg', ...
0067             'Invalid option %s for slcumuscore', op);
0068 end
0069 slignorevars(ss);
0070 
0071 if maxr < nclasses
0072     sp = sp(1:maxr, :);
0073 else
0074     maxr = nclasses;
0075 end
0076 decisions = clabels(sp);
0077 matches = (decisions == repmat(qlabels', [maxr, 1]));
0078 
0079 cs = cumsum(sum(matches, 2), 1) / nsamples;
0080 
0081

---

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