Pretend I have two cell arrays A and B, each element in those cells is N*M matrix, example :A={ [2 3;4 5] [1 5;7 8]}
and B={ [1 2;4 5] [7 9;10 1]}
both are cells each element is 2*2 matrix.
假设我有两个单元阵列A和B,这些单元格中的每个元素都是N * M矩阵,例如:A = {[2 3; 4 5] [1 5; 7 8]}和B = {[1 2; 4 5] [7 9; 10 1]}两者都是细胞,每个元素是2 * 2矩阵。
Now I can subtract those cell arrays element-wise like this:
现在我可以像这样以元素方式减去那些单元格数组:
C=cellfun(@minus,A,B,'UniformOutput',false);
this will result in C={[1 1;0 0] [-6 -4;-3 7]}
.
这将导致C = {[1 1; 0 0] [-6 -4; -3 7]}。
Now is that the fastest way ? or is there a faster approach ?
现在是最快的方式吗?还是有更快的方法?
Consider cells with large number of matrices each matrix is small.
考虑具有大量矩阵的单元,每个矩阵都很小。
2 个解决方案
#1
As already mentioned a lot depends on the data, but in your example the fastest way is probably a nested for loop:
如前所述,很大程度上取决于数据,但在您的示例中,最快的方法可能是嵌套for循环:
A={ [2 3;4 5] [1 5;7 8]};
B={ [1 2;4 5] [7 9;10 1]};
tic
C=cellfun(@minus,A,B,'UniformOutput',false);
toc
tic
s = size(A);
for ii=1:s(1)
for jj=1:s(2)
D{ii,jj} = A{ii,jj}-B{ii,jj};
end
end
toc
isequal ( C, D )
output:
Elapsed time is 0.001420 seconds.
Elapsed time is 0.000017 seconds.
ans =
1
#2
You could convert to a 3D array, subtact, and convert back:
您可以转换为3D数组,子关系,并转换回:
mat2cell(cat(3, A{:}) - cat(3, B{:}), size(A{1},1), size(A{1},2), [1 1]);
Of course it would be faster if you could avoid the conversions, that is, work with 3D arrays directly.
当然,如果你可以避免转换,那就更快,也就是直接使用3D数组。
#1
As already mentioned a lot depends on the data, but in your example the fastest way is probably a nested for loop:
如前所述,很大程度上取决于数据,但在您的示例中,最快的方法可能是嵌套for循环:
A={ [2 3;4 5] [1 5;7 8]};
B={ [1 2;4 5] [7 9;10 1]};
tic
C=cellfun(@minus,A,B,'UniformOutput',false);
toc
tic
s = size(A);
for ii=1:s(1)
for jj=1:s(2)
D{ii,jj} = A{ii,jj}-B{ii,jj};
end
end
toc
isequal ( C, D )
output:
Elapsed time is 0.001420 seconds.
Elapsed time is 0.000017 seconds.
ans =
1
#2
You could convert to a 3D array, subtact, and convert back:
您可以转换为3D数组,子关系,并转换回:
mat2cell(cat(3, A{:}) - cat(3, B{:}), size(A{1},1), size(A{1},2), [1 1]);
Of course it would be faster if you could avoid the conversions, that is, work with 3D arrays directly.
当然,如果你可以避免转换,那就更快,也就是直接使用3D数组。