# Select random data from a matrix and replace it

17 views (last 30 days)
stelios loizidis on 14 May 2019
Commented: Jos (10584) on 16 May 2019
Hello,
I have the following problem: I have a matrix e.g
1 0 1 0 1
1 1 1 0 0
1 1 1 1 1
1 1 0 0 1
and I try to do this: When the sum of each column exceeds the threshold=4 some random "1" of the column that goes beyond the threshold become zero. How do I implement this in matlabThis can be done without using loops?I would be very grateful if someone helpend me

Andrei Bobrov on 14 May 2019
Edited: Andrei Bobrov on 15 May 2019
A = [1 0 1 0 1
1 1 1 0 0
0 0 0 0 0
1 1 1 1 1
1 1 0 0 1
0 0 0 0 0];
p = 4;
[ii,jj] = find(A);
jjj = accumarray(ii,jj,[size(A,1),1],@(x){x(randperm(numel(x),min(numel(x),p)))});
k = cellfun(@numel,jjj);
out = accumarray([repelem((1:numel(k))',k),cell2mat(jjj)],1,size(A));
stelios loizidis on 15 May 2019
It works!!!Thank you very much!!!!!!

Jos (10584) on 15 May 2019
Edited: Jos (10584) on 15 May 2019
Here is another, indexing, approach:
A = randi(2, 6, 8)-1 % random 0/1 array
M = 3 % max number of 1's per column
szA = size(A) ;
B = zeros(szA) ;
tf = A == 0 ;
[~, P] = sort(rand(szA)) ; % randperm for matrix along columns
P(tf) = 0 ;
[~, r] = maxk(P, M) ; % rows with M highest values in P (including 0's) per column
i = r + szA(1)*(0:szA(2)-1) ; % convert to linear indices
B(i) = 1 ; % B contains M 1's per column
B(tf) = 0 % reset those that were 0 in A -> only maximally M 1's per column remain
##### 2 CommentsShowHide 1 older comment
Jos (10584) on 16 May 2019
I don't get it. Why not simply run the code only once for a given matrix A?

Jan on 16 May 2019
Edited: Jan on 16 May 2019
A logical mask is much simpler than handling the indices:
A = randi([0,1], 8, 8); % Test data
p = 4;
mask = (cumsum(A, 1) > p) & A;
This replaces 1s in each column by a 0 with a probability of 50%, but leaves the first p ones untouched.
But if you want to change any 1s without keeping the first p 1s of each column:
mask = (sum(A, 1) > p) & A; % cumsum -> sum, auto-expand: >= R2016b
If you want to replace the 1s in the columns with more than p 1s with another probability:
Jos (10584) on 16 May 2019
From the OP: "When the sum of each column exceeds the threshold=4 some random "1" of the column that goes beyond the threshold become zero."
If I now read it again, just a single one in such a column should be set to 0 then? ...