File Exchange

image thumbnail


version 1.0 (2.32 KB) by

randomly swap elements of a matrix



View License

For vectors, Y = RANDSWAP(X) randomly swaps the elements of X. For N-D matrices, RANDSWAP(X) randomly swaps the elements along the first non-singleton dimension of X.

RANDSWAP(X,DIM) randomly swaps the elements along the dimension DIM. For instance, RANDSWAP(X,1) randomly interchanges the rows of X.

Y = RANDSWAP(X,'partial') swaps the elements for each of the non-singleton dimensions of X separately. Rows are interchanged first, then columns, then planes, etc. In this case, elements that belong to the same row, column, ... stay together.

Y = RANDSWAP(X,'full') randomly swaps all the elements in X.

[Y, I, J] = RANDSWAP(...) return index matrices I and J so that Y = X(I) and X = Y(J).

X can be a numeric or a cell array.

Version 2.0 (nov 2007)

Comments and Ratings (14)


Sugato (view profile)


h612 (view profile)

Is there a way to store the new index of the new matrix? Im using for training and testing- and I'd like to arrange labels according to the new arrangement.


Thank you



it's great
thank you



thanks a lot, worked good with my cell array

Jos (10584)

Jos (10584) (view profile)

@Andrey ... or use randperm directly, of course

ri = randperm(size(a,1))
ra = a(ri,ri)

Jos (10584)

Jos (10584) (view profile)

@Andrey, you could accomplish this using the 2nd output of randswap:

a = reshape(1:25,5,5)
[ra,i] = randswap(a,1) % swap rows
ra = ra.' ; % transpose (columns become rows)
ra = ra(i) ; % swap rows again
ra = ra.' ; transpose again

For QAP, I need to scramble rows and columns of the adjacency matrix in exactly the same way. It would be nice if there was an option to apply the same permutation to both dimensions for scrambling square matrices.


grega (view profile)

Thank you for this great function!

Would it be possible to update the function with an argument specifying distribution also?

I suppose randomization is done according to uniform distribution. It would be great if it works also with other, e.g. normal, binomial, poisson, inverse of normal etc.

Thank you!

Kevin Grey

Great! This is exactly what I wanted. For network analysis of Quadratic Assignment Procedure (QAP), this step is pre-required.

Jos (the author)

Thanks Yidong for pointing out the errors. I am very sorry and will update the file shortly. Jos

Yidong Li

Unfortunately, it doesn't work for me. First thing confused me:
the line 68, all the "dim" should be "dimmode", is that right?
Second thing: Even I modified them, I still couldn't get the correct answer to "randswap(X,1)". The result seems like "randswap(X)".
Can you other guys check it as well?

MATLAB Release
MATLAB 6.5 (R13)

Inspired by: SHAKE

Inspired: randone1

Download apps, toolboxes, and other File Exchange content using Add-On Explorer in MATLAB.

» Watch video

Win prizes and improve your MATLAB skills

Play today