Problem with size of a 0-by-1 empty matrix.

hi dear friends. For computing the jaccard similarity index of a graph's nodes I need to compute a pairwise intersection among their neighborhood matrix and find the number of common neighbors. for example (2, [1,4], [], 2) are neighbors of nodes 1 to 4 and node 3 is alone. The problem is that when I compute the intersection, the size of results for the alone node is [1x0 double] or [0x1 double] like this:

(Int)

[ 2] [1x0 double] [1x0 double] [ 2]

[1x0 double] [1x2 double] [0x1 double] [1x0 double]

[1x0 double] [0x1 double] [0x1 double] [1x0 double]

[ 2] [1x0 double] [1x0 double] [ 2]

and the size of [0x1 double] vector will be computed as 1 while the correct answer is 0 like this!

the size matrix:

1 0 0 1

0 2 1 0

0 1 1 0

1 0 0 1

I have the same problem with union, how can I fix this problem??

related part of my code is this:

[ii, jj] = ndgrid(1:Size) %Size=number of nodes

H=NeighborsS(L1,:); %%% I have a multiplex network and L1 is the layernumber

result= arrayfun(@(n) intersect(H{ii(n)},H{jj(n)}), 1:Size^2, 'uni', 0);

Int(:,:,L1) = reshape(result,Size,Size);

IntNum=cellfun('size',Inti,2);

Answer 1

[Walter Roberson]

empty = cellfun(@isempty, result);
result(empty) = {zeros(0,0)};         %same size of emptiness for each empty result
Int(:,:,L1) = reshape(result,Size,Size);

IntNum = cellfun('size',Inti,2);
 

[sam kamiab]

tnx alot. it works:)

Answer 2

[Steven Lord]

You don't actually care about the size of the neighbors vector, you care about the number of neighbors. Use @numel instead of 'size' in your cellfun call.