Samples elements of X so result uses at most maxMegs megabytes of memory. If X is m+1 dimensional, say of size [d1 x d2 x...x dm x n], each [d1 x d2 x...x dm] element is treated as one observation, and X is treated as having n such observations. The subsampling then occurs over the last dimension n. Different types of arrays require different amounts of memory. Each double requries 8 bytes of memory, hence an array with 1.024 million elements of type double requires 8MB memory. Each uint8 requires 1 byte, so the same size array would require 1MB. Note that when saved to .mat files arrays may take up more or less memory (due to compression, etc.) Different from Matlab randsample ! Note, to see how much memory a variable x is using in memory, use: s=whos('x'); mb=s.bytes/2^20 USAGE [X,keeplocs] = subsampleMatrix( X, maxMegs ) INPUTS X - [d1 x ... x dm x n], treated as n [d1 x ... x dm] elements maxMegs - maximum number of megs Xsam is allowed to take up OUTPUTS Xsam - [d1 x ... x dm x n'] (n'<=n) Xsam=X(:,..,:,keeplocs); keeplocs - vector of indicies kept from X; EXAMPLE % Xsam should have size: 1024xround(1024/10) X = uint8(ones(2^10,2^10)); Xsam = subsampleMatrix( X, 1/10 ); % Xsam should have size: 100x10x~(1000/8) X = rand(100,10,1000); Xsam = subsampleMatrix( X, 1 ); See Also Piotr's Image&Video Toolbox Version 2.0 Copyright 2008 Piotr Dollar. [pdollar-at-caltech.edu] Please email me if you find bugs, or have suggestions or questions! Licensed under the Lesser GPL [see external/lgpl.txt]