使用不同的输入将repmat应用于Matlab中矩阵的每一行

Question

I want to stack each row of a matrix A in Matlab a different number of times according to what is reported in the vector count . To do that I use repmat in the following way:

counts=[524282; 524286; 524283; 524290];
A=randn(4,19); 
f=@() cell2mat(arrayfun(@(x) repmat(A(x,:),counts(x),1), 1:size(counts,1), 'UniformOutput', 0)');
timeit(f)

The code takes approx 0.45 sec.

Would you be able to suggest anything faster?

Answer 1

You could use repelem to repeat each element in the initial matrix a particular number of times

result = reshape(repelem(A(:), repmat(counts(:), size(A, 2), 1)), [], size(A, 2));

Explanation

A = [1, 2, 3; 4, 5, 6; 7, 8, 9];
counts = [1, 3, 2];

repelem (for vector inputs), repeats the ith element of the first input the number of times specified in the corresponding element of the second input.

repelem([1, 2, 3], [3, 4, 1])
%     1     1     1     2     2     2     2     3

Since repelem only supports vector inputs, we flatten A into a column vector using A(:) . We also then need to craft the second input (the number of times to repeat each element of A ) such that we repeat counts for each column of A .

nTimes = repmat(counts(:), size(A, 2), 1)

We then use repelem to perform the repetition

repeated = repelem(A(:), nTimes);

We then reshape the result to have the correct number of columns

reshape(repeated, [], size(A, 2))

Answer 2

A solution base on diff and cumsum :

co = size(A,2);
A_diff=diff([zeros(1,co);A]);
idx=cumsum([1;counts(1:end-1)]);
result=zeros(sum(counts),co);
result(idx,:)=A_diff;
result = cumsum(result)

The idea is using cumsum we can reconstruct a vector form the first difference of it.

a=[5 3 2 4];
d=  diff([0 a]);
out = cumsum(d)

So out is equal to a .

We can make a matrix all of zeros of the desired size:

result=zeros(sum(count),co);

we should find indices of rows of each section

idx=cumsum([1;count(1:end-1)]);

set the first row of each section to first difference of the matrix A

A_diff=diff([zeros(1,co);A]);
result(idx,:)=A_diff;

so

A = randi(4,4,7);
count=[3 ;4 ;2 ;5];
A =

   2   1   4   2   1   3   1
   2   3   2   4   4   3   2
   1   3   1   4   1   2   1
   2   3   2   2   3   1   2

A_diff =

   2   1   4   2   1   3   1
   0   2  -2   2   3   0   1
  -1   0  -1   0  -3  -1  -1
   1   0   1  -2   2  -1   1

result =

   2   1   4   2   1   3   1
   0   0   0   0   0   0   0
   0   0   0   0   0   0   0
   0   2  -2   2   3   0   1
   0   0   0   0   0   0   0
   0   0   0   0   0   0   0
   0   0   0   0   0   0   0
  -1   0  -1   0  -3  -1  -1
   0   0   0   0   0   0   0
   1   0   1  -2   2  -1   1
   0   0   0   0   0   0   0
   0   0   0   0   0   0   0
   0   0   0   0   0   0   0
   0   0   0   0   0   0   0

apply cumsum to it to get the desired result

result = cumsum(result)
result =

   2   1   4   2   1   3   1
   2   1   4   2   1   3   1
   2   1   4   2   1   3   1
   2   3   2   4   4   3   2
   2   3   2   4   4   3   2
   2   3   2   4   4   3   2
   2   3   2   4   4   3   2
   1   3   1   4   1   2   1
   1   3   1   4   1   2   1
   2   3   2   2   3   1   2
   2   3   2   2   3   1   2
   2   3   2   2   3   1   2
   2   3   2   2   3   1   2
   2   3   2   2   3   1   2