Description of slpartitionpca_construct

Home > sltoolbox > subspace_ex > slpartitionpca_construct.m

slpartitionpca_construct

PURPOSE

SLPARTITIONPCA_CONSTRUCT Constructs the array from features

SYNOPSIS

function A = slpartitionpca_construct(S, modeldir, feas)

DESCRIPTION

SLPARTITIONPCA_CONSTRUCT Constructs the array from features 

 $ Syntax $
   - A = slpartitionpca_construct(S, modeldir, feas)

 $ Arguments $
   - S:            the partition-based PCA model struct or model core file
   - modeldir:     the directory path of the projection files
   - feas:         the partition PCA features
   - A:            the constructed arrays in original space

 $ Description $
   - A = slpartitionpca_construct(S, modeldir, feas) constructs the 
     array units in the original space from the extracted partiton PCA
     features. Here, S can be a struct loaded from core file or the core 
     filename. modeldir is the directory of the corefile and the related 
     projection matrix files.

 $ Remarks $
   - When the combined model is learned, the dimension of features can be
     lower than the dimension of feature subspace of the PCA model. In 
     this case, only the first k principal components are used.

 $ History $
   - Created by Dahua Lin, on Jul 30th, 2006
   - Modified by Dahua Lin, on Sep 10, 2006
       - replace sladd by sladdvec to increase efficiency

CROSS-REFERENCE INFORMATION

This function calls:

• sladdvec SLADDVEC adds a vector to columns or rows of a matrix
• slreadarray SLREADARRAY Reads an array from an array file
• raise_lackinput RAISE_LACKINPUT Raises an error indicating lack of input argument
• sladdpath SLADDPATH Adds dirpath to precede the filenames
• slrange2indcells SLRANGE2INDCELLS Converts a range array to indices cell array

This function is called by:

SOURCE CODE

0001 function A = slpartitionpca_construct(S, modeldir, feas)
0002 %SLPARTITIONPCA_CONSTRUCT Constructs the array from features
0003 %
0004 % $ Syntax $
0005 %   - A = slpartitionpca_construct(S, modeldir, feas)
0006 %
0007 % $ Arguments $
0008 %   - S:            the partition-based PCA model struct or model core file
0009 %   - modeldir:     the directory path of the projection files
0010 %   - feas:         the partition PCA features
0011 %   - A:            the constructed arrays in original space
0012 %
0013 % $ Description $
0014 %   - A = slpartitionpca_construct(S, modeldir, feas) constructs the
0015 %     array units in the original space from the extracted partiton PCA
0016 %     features. Here, S can be a struct loaded from core file or the core
0017 %     filename. modeldir is the directory of the corefile and the related
0018 %     projection matrix files.
0019 %
0020 % $ Remarks $
0021 %   - When the combined model is learned, the dimension of features can be
0022 %     lower than the dimension of feature subspace of the PCA model. In
0023 %     this case, only the first k principal components are used.
0024 %
0025 % $ History $
0026 %   - Created by Dahua Lin, on Jul 30th, 2006
0027 %   - Modified by Dahua Lin, on Sep 10, 2006
0028 %       - replace sladd by sladdvec to increase efficiency
0029 %
0030 
0031 %% parse and verify input arguments
0032 
0033 if nargin < 3
0034     raise_lackinput('slpartitionpca_apply', 3);
0035 end
0036 
0037 if ischar(S)
0038     S = load(S);
0039 elseif ~isstruct(S)
0040     error('sltoolbox:invalidarg', ...
0041         'The S should be the filename of the core file or the core struct');
0042 end
0043 
0044 [k, n] = size(feas);
0045 if k > S.diminfo.feadim
0046     error('sltoolbox:sizmismatch', ...
0047         'k is larger than the dimension of feature space');
0048 elseif k < S.diminfo.feadim && isempty(S.combprojfile)
0049     error('sltoolbox:sizmismatch', ...
0050         'When the combined model is not learned, k should be exactly the same as the dimension of feature space');
0051 end
0052 
0053 %% Construct Intermediate features
0054 
0055 if ~isempty(S.combprojfile)
0056    
0057     combprojpath = sladdpath(S.combprojfile, modeldir);
0058     P = slreadarray(combprojpath);
0059     if k < S.diminfo.feadim
0060         P = P(:, 1:k);
0061     end
0062     intfeas = P * feas;
0063     clear P;
0064     
0065 else
0066     intfeas = feas;
0067     
0068 end
0069     
0070 %% Construct Array Units
0071 
0072 A = zeros([size(S.meanarr), n]);
0073 NBlks = numel(S.blocks);
0074 projpaths = sladdpath(S.projfiles, modeldir);
0075 
0076 dc = 0;
0077 for ib = 1 : NBlks
0078     
0079     curblock = S.blocks{ib};
0080     rgncell = slrange2indcells(curblock);
0081     
0082     cursiz = curblock(2,:) - curblock(1,:) + 1;
0083     curdim = prod(cursiz);
0084     cursubdim = S.diminfo.subdims(ib);
0085     
0086     curfeasec = intfeas(dc+1:dc+cursubdim, :);        
0087     curproj = slreadarray(projpaths{ib});
0088     
0089     localmean = S.meanarr(rgncell{:});
0090     localmean = reshape(localmean, [curdim, 1]);
0091     
0092     localarr = curproj * curfeasec;
0093     clear curproj;
0094     
0095     localarr = sladdvec(localarr, localmean, 1);
0096     localarr = reshape(localarr, [cursiz, n]);
0097     
0098     A(rgncell{:}, :) = localarr;    
0099     
0100     clear localarr localmean;
0101     
0102     dc = dc + cursubdim;
0103     
0104 end
0105 
0106 
0107 
0108 
0109     
0110

---

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