how to choose matrix columns randomly?

hi all. i have two related matrices (A and B). i want to choose randomly 80% of one matrix's columns (e.g A) and corresponding values from second matrix(B)for example:

suppose that two matrices are like this:

A=[1 2 3 4 5 6 7 8 9 10];
B=[11 12 13 14 15 16 17 18 19 20];

desired matrices are for example like:

C=[1 3 4 5 7 8 9 10]
D=[11 13 14 15 17 18 19 20]

would you help me on this? thank you

[Stephen23]

How many columns does A have? What is 80% of one column? What is 80% of two columns? How do you define "80% of one matrix's columns" ?

[maryam]

matrices are 11*200. so i need two 11*160 matrices. i want to choose these 160 columns randomly.

[the cyclist]

Do you want the 160 to be unique, or possibly repeating? For example, might you choose the 5th column twice?

Also, do you have the Statistics Toolbox? A couple commands there make this a bit cleaner.

[maryam]

no, columns should be unique. yes, i have statistics toolbox,but i haven't work with it yet!

Answer 1

[Star Strider]

Use the randperm function.

This might do what you want:

cols = randperm(200, 160);

(Note: Untested code, but it should work.)

[Star Strider]

With respect to your edited Question, my edited Answer becomes:

A=[1 2 3 4 5 6 7 8 9 10];
B=[11 12 13 14 15 16 17 18 19 20];

cols = randperm(size(A,2), round(0.8*size(A,2)));

producing (for example):

cols = 

       4     2     9     8     3     7     6     5

and:

C = A(cols);
D = B(cols);

producing:

C =

       4     2     9     8     3     7     6     5

D =

      14    12    19    18    13    17    16    15

[maryam]

thank you for your answer

[Star Strider]

My pleasure!

Answer 2

[the cyclist]

As Stephen's questions suggest, you need some more specification.

Possible elements of the solution are

• randi, randsample, or other commands to select some random indices. (Say that vector of random indices is called idx )
• Something like either B(:,idx) or B(:,A(idx)) to choose from B directly with the indices, or using values from A that are selected by the indices

[the cyclist]

idx = randsample(200,160);
C = A(idx);
D = B(idx)'

Answer 3

[Jos (10584)]

Exactly 80 vs 20? Or in approximation? And how many columns? I also assume the number of rows in A and B are the same.

A suggestion is that you first draw N columns from A and M columns from B, where N/M is 80/20

Ntotal = 10 ;
N = ceil(.8 * Ntotal) % fixed 80 vs 20
% or
% N = sum(rand(100,1)<0.8)) % roughly

M = Ntotal - N 

% select N columns from A and M columns from B
cidxA = randi([1 size(A,2)],1,N)
cidxB = randi([1 size(B,2)],1,M)

% stick them together
OUT = [A(:,cidxA) B(:,cidxB)]

% randomise the order
r = randperm(size(OUT,2))
OUT = OUT(:,r)

[maryam]

Dear Jos, i edited my question. it wasn't clear. 80 vs 20 should occur in approximation and in both two matrices.