Fastest way to have a matrix with intersected elements only between matrix and array

Hello,

Lets say I have the following matrix:

0.1 0.2 0.3

0.3 0.5 0.7

0.8 0.1 0.9

And the following array:

0.1 0.3 0.5

Id like know the fastest way to get to the following result:

0.1 0 0.3

0.3 0.5 0.7

0 0 0

So far Ive achieved it with the following, which creates a mask and multiplys it:

 idxs = ismembc( domainWeight, dWeightVector );
 dWeightVector = bsxfun(@times, idxs, domainWeight);

Is there a fastest way to intersect a matrix with an array and return a matrix, with its original size, but only populated with the elements that exist on both original array and matrix?

Thank you

Answer 1

[Azzi Abdelmalek]

A=[0.1 0.2 0.3
0.3 0.5 0.7
0.8 0.1 0.9]
B=[0.1 0.5 0.3]
idx=ismember(A,B)
A(~idx)=0

[Thiago Motta]

Although the last line in fact does give me some speedup since it gets rid of element-wise multiplication, ismember is still the bottleneck of this code.

Im trying to rewrite ismembc2.cpp (generating a mex file) in order to get unique entries and sort them

Answer 2

[dpb]

Close

>> a.*ismember(a,v)
ans =
  0.1000         0    0.3000
  0.3000    0.5000         0
       0    0.1000         0
>>

NB: It appears your answer isn't correct; 0.7 isn't included in v so 2nd row, 3rd column should be 0 but the 0.1 in row3, column 2, does match the 0.1 in the vector. Unless it were to be corresponding positions in the vector and the array row, but then the 0.3 wouldn't appear because while it's in both, it isn't in same column so that doesn't seem to be the rule, either.

[Thiago Motta]

That actually happens because the array isnt sorted, which is one of the reasons why ismembc is faster than ismember, for it doesnt do some sanity checks. (My bad, should have sorted that out first)