Here is my problem:
I have a nxn
matrix in matlab. I want to delete all the zeros of this matrix and put the rows of it in vectors. For n=4
, let say I have the following matrix:
A = [ 1 1 0 0
1 2 0 0
1 0 0 0
1 2 1 0 ];
How to get the following:
v1 = [ 1 1 ];
v2 = [ 1 2 ];
v3 = [ 1 ];
v4 = [ 1 2 1 ];
I did the following:
for i = 1:size(A, 1)
tmp = A(i, :);
tmp(A(i, :)==0)=[];
v{i} = tmp;
end
Slightly faster than Divakar's answer:
nzv = arrayfun(@(n) nonzeros(A(n,:)), 1:size(A,1), 'uniformoutput', false);
Benchmarking
Small matrix
A = randi([0 3],100,200);
repetitions = 1000;
tic
for count = 1:repetitions
nzv =cellfun(@(x) nonzeros(x),mat2cell(A,ones(1,size(A,1)),size(A,2)),'uni',0);
end
toc
tic
for count = 1:repetitions
nzv = arrayfun(@(n) nonzeros(A(n,:)), 1:size(A,1), 'uniformoutput', false);
end
toc
Elapsed time is 3.017757 seconds.
Elapsed time is 2.025967 seconds.
Large matrix
A = randi([0 3],1000,2000);
repetitions = 100;
Elapsed time is 11.483947 seconds.
Elapsed time is 5.563153 seconds.
Convert to a cell array such that you have a cell for each row and then use nonzeros
for each cell, that deletes zeros and finally store them into separate variables.
Code
nzv =cellfun(@(x) nonzeros(x),mat2cell(A,ones(1,size(A,1)),size(A,2)),'uni',0)
[v1,v2,v3,v4] = nzv{:}