Description of sllabeledsum

Home > sltoolbox > core > sllabeledsum.m

sllabeledsum

PURPOSE

SLLABELEDSUM Sums the numbers according to labels

SYNOPSIS

function S = sllabeledsum(X, labels, labelset, w)

DESCRIPTION

SLLABELEDSUM Sums the numbers according to labels

 $ Syntax $
   - S = sllabledsum(X, labels, labelset)
   - S = sllabledsum(X, labels, labelset, w)

 $ Arguments $
   - X:            The matrix of numbers (m x n)
   - labels:       The labels of columns in X (1 x n)
   - labelset:     The set of labels used (length-c vector)
   - w:            The column weights (1 x n) (default = [])
   - S:            The sum (m x c)

 $ Description $
   - S = sllabledsum(X, labels, labelset) sums the columns in X that 
     corresponding to the same label. If labelset has c labels, then 
     S would have c columns, the i-th column sums the columns in X 
     associative the labelset(i).

   - S = sllabledsum(X, labels, labelset, w) performs a weighted sum.

 $ History $
   - Created by Dahua Lin, on Aug 31, 2006
   - Modified by Dahua Lin, on Sep 10, 2006
       - replace slmul by slmulvec to increase efficiency

CROSS-REFERENCE INFORMATION

This function calls:

• slmulvec SLMULVEC multiplies a vector to columns or rows of a matrix
• raise_lackinput RAISE_LACKINPUT Raises an error indicating lack of input argument
• sllabelinds SLLABELINDS Extract indices corresponding to specified labels

This function is called by:

• slkmeansex SLKMEANSEX Performs Generalized K-means
• slcountvote SLCOUNTRULE Counts the votings to make histogram

SOURCE CODE

0001 function S = sllabeledsum(X, labels, labelset, w)
0002 %SLLABELEDSUM Sums the numbers according to labels
0003 %
0004 % $ Syntax $
0005 %   - S = sllabledsum(X, labels, labelset)
0006 %   - S = sllabledsum(X, labels, labelset, w)
0007 %
0008 % $ Arguments $
0009 %   - X:            The matrix of numbers (m x n)
0010 %   - labels:       The labels of columns in X (1 x n)
0011 %   - labelset:     The set of labels used (length-c vector)
0012 %   - w:            The column weights (1 x n) (default = [])
0013 %   - S:            The sum (m x c)
0014 %
0015 % $ Description $
0016 %   - S = sllabledsum(X, labels, labelset) sums the columns in X that
0017 %     corresponding to the same label. If labelset has c labels, then
0018 %     S would have c columns, the i-th column sums the columns in X
0019 %     associative the labelset(i).
0020 %
0021 %   - S = sllabledsum(X, labels, labelset, w) performs a weighted sum.
0022 %
0023 % $ History $
0024 %   - Created by Dahua Lin, on Aug 31, 2006
0025 %   - Modified by Dahua Lin, on Sep 10, 2006
0026 %       - replace slmul by slmulvec to increase efficiency
0027 %
0028 
0029 %% parse and verify input
0030 
0031 if nargin < 3
0032     raise_lackinput('sllabeledsum', 3);
0033 end
0034 
0035 if ndims(X) ~= 2 || ~isnumeric(X)
0036     error('sltoolbox:invalidarg', 'X should be a numeric 2D matrix');
0037 end
0038 [m, n] = size(X);
0039 
0040 if ~isequal(size(labels), [1 n])
0041     error('sltoolbox:sizmismatch', ...
0042         'The size of labels does not match the number of X-columns');
0043 end
0044 
0045 if ~isvector(labelset)
0046     error('sltoolbox:invalidarg', 'labelset should be a vector');
0047 end
0048 c = length(labelset);
0049 
0050 if nargin < 4
0051     w = [];
0052 else
0053     if ~isempty(w)
0054         if ~isequal(size(w), [1 n])
0055             error('sltoolbox:sizmismatch', ...
0056                 'The size of w does not match the number of X-columns');
0057         end
0058     end
0059 end
0060 
0061 %% compute
0062 
0063 Inds = sllabelinds(labels, labelset);
0064 S = zeros(m, c);
0065 
0066 
0067 for i = 1 : c
0068     curinds = Inds{i};
0069     if ~isempty(curinds)
0070         curX = X(:, curinds);        
0071         if ~isempty(w)
0072             curw = w(curinds);
0073             curX = slmulvec(curX, curw, 2);
0074         end        
0075         S(:,i) = sum(curX, 2);
0076     end
0077 end
0078 
0079 
0080 
0081     
0082 
0083 
0084

---

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