Description of slgaussrnd

Home > sltoolbox > stat > slgaussrnd.m

slgaussrnd

PURPOSE

SLGAUSSRND Generates random samples from Gaussian models

SYNOPSIS

function X = slgaussrnd(GS, nums)

DESCRIPTION

SLGAUSSRND Generates random samples from Gaussian models

 $ Syntax $
   - X = slgaussrnd(GS, nums)

 $ Arguments $
   - GS:       The Gaussian model struct
   - nums:     the number of samples from the models

 $ Description $
   - X = slgaussrnd(GS, nums) randomly draws samples from Gaussian models.
     Suppose there are k constituent models. Then nums can be of
     the following forms:
       - length-k vector, representing the number of samples of the
         models respectively.

 $ History $
   - Created by Dahua Lin, on Aug 24, 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
• slmulvec SLMULVEC multiplies a vector to columns or rows of a matrix
• slsymeig SLSYMEIG Compute the eigenvalues and eigenvectors for symmetric matrix
• slgausstype SLGAUSSTYPE Judges the type of a Gaussian model struct
• slpartition SLPARTITION Partition a range into blocks in a specified manner

This function is called by:

SUBFUNCTIONS

• function X = gensamples_univar(vmean, var, d, n)
• function X = gensamples_diagvar(vmean, vars, d, n)
• function X = gensamples_covar(vmean, C, d, n)

SOURCE CODE

0001 function X = slgaussrnd(GS, nums)
0002 %SLGAUSSRND Generates random samples from Gaussian models
0003 %
0004 % $ Syntax $
0005 %   - X = slgaussrnd(GS, nums)
0006 %
0007 % $ Arguments $
0008 %   - GS:       The Gaussian model struct
0009 %   - nums:     the number of samples from the models
0010 %
0011 % $ Description $
0012 %   - X = slgaussrnd(GS, nums) randomly draws samples from Gaussian models.
0013 %     Suppose there are k constituent models. Then nums can be of
0014 %     the following forms:
0015 %       - length-k vector, representing the number of samples of the
0016 %         models respectively.
0017 %
0018 % $ History $
0019 %   - Created by Dahua Lin, on Aug 24, 2006
0020 %   - Modified by Dahua Lin, on Sep 10, 2006
0021 %       - Replace sladd by sladdvec to increase efficiency
0022 %
0023 
0024 %% parse and verify input arguments
0025 
0026 tyinfo = slgausstype(GS);
0027 d = GS.dim;
0028 k = GS.nmodels;
0029 
0030 if length(nums) ~= k
0031     error('sltoolbox:invalidarg', ...
0032         'The nums should be a length-k vector');
0033 end
0034 
0035 n = sum(nums);
0036 ps = slpartition(n, 'blksizes', nums);
0037 
0038 %% generate samples
0039 
0040 X = zeros(d, n);
0041 
0042 switch tyinfo.varform
0043     case 'univar'
0044         if tyinfo.sharevar
0045             for i = 1 : k
0046                 si = ps.sinds(i); ei = ps.einds(i);
0047                 X(:, si:ei) = gensamples_univar(GS.means(:,i), GS.vars, d, nums(i));
0048             end
0049         else
0050             for i = 1 : k
0051                 si = ps.sinds(i); ei = ps.einds(i);
0052                 X(:, si:ei) = gensamples_univar(GS.means(:,i), GS.vars(i), d, nums(i));
0053             end                
0054         end
0055         
0056     case 'diagvar'
0057         if tyinfo.sharevar
0058             for i = 1 : k
0059                 si = ps.sinds(i); ei = ps.einds(i);
0060                 X(:, si:ei) = gensamples_diagvar(GS.means(:,i), GS.vars, d, nums(i));
0061             end
0062         else
0063             for i = 1 : k
0064                 si = ps.sinds(i); ei = ps.einds(i);
0065                 X(:, si:ei) = gensamples_diagvar(GS.means(:,i), GS.vars(:,i), d, nums(i));
0066             end                
0067         end
0068         
0069     case 'covar'
0070         if tyinfo.sharevar
0071             for i = 1 : k
0072                 si = ps.sinds(i); ei = ps.einds(i);
0073                 X(:, si:ei) = gensamples_covar(GS.means(:,i), GS.covs, d, nums(i));
0074             end
0075         else
0076             for i = 1 : k
0077                 si = ps.sinds(i); ei = ps.einds(i);
0078                 X(:, si:ei) = gensamples_covar(GS.means(:,i), GS.covs(:,:,i), d, nums(i));
0079             end                
0080         end
0081         
0082     otherwise
0083         error('sltoolbox:invalidarg', ...
0084             'Invalid variance form in GS: %s', tyinfo.varform);
0085         
0086 end
0087 
0088 
0089 %% Core generation routines
0090 
0091 function X = gensamples_univar(vmean, var, d, n)
0092 
0093 X = randn(d, n) * sqrt(max(var, 0));
0094 X = sladdvec(X, vmean, 1);
0095 
0096 function X = gensamples_diagvar(vmean, vars, d, n)
0097 
0098 X = randn(d, n);
0099 X = slmulvec(X, sqrt(max(vars, 0)), 1);
0100 X = sladdvec(X, vmean, 1);
0101 
0102 function X = gensamples_covar(vmean, C, d, n)
0103 
0104 X = randn(d, n);
0105 [D, V] = slsymeig(C);
0106 T = V * diag(sqrt(max(D, 0)));
0107 clear D V;
0108 X = T * X;
0109 clear T;
0110 X = sladdvec(X, vmean, 1);
0111 
0112 
0113 
0114 
0115 
0116 
0117 
0118 
0119 
0120 
0121

---

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