Element by element (xor) operation on cells.

I have two variables, Test and Train.

Test = cell of dimension 10 x 1 and

Train = cell of dimension 20 x 1

All elements are logical.

I want an operation similar to matrix multiplication, except that I want xor, instead of multiplication.

All of the 100 elements in Test should be xor(ed) with each of the 2000 elements in Train. All the while, I was doing this using for loop, but I want to avoid for loops now. I was trying to figure out performing this using bsxfun, but couldn't get it as desired. What is the most efficient way to do this?

[Walter Roberson]

Are you saying that Test is a cell array in which each element is a single scalar logical value? If so then why are you storing the values in a cell array instead of a vector?

Could you confirm that the output you want is 100 x 2000?

[James Tursa]

Please show a small example of inputs and desired output. E.g., in a regular matrix multiply,

c11 = a11*b11 + a12*b21 + a13*b31 + ...

What would be the equivalent "xor matrix multiply" operation that you want?

c11 = xor(a11,b11) or xor(a12,b21) or xor(a13,b31) or ...

Or something else perhaps using "and"?

Or will this be more like an outer product as Walter suggests?

[Meghana Dinesh]

For simplicity, I have reduced the dimensionality and attached example .mat files.

And yes, the output should have 10 rows and 20 columns.

output(i,j) = ( i th element from Test ) xor ( jth element from Train )

[James Tursa]

Loops is probably the way to go here. You could potentially pull things out via cell2mat, do the calculations, and then use cell2mat to get your result, but why bother with all of that data copying? Using loops is straightforward and easy to read.

[Meghana Dinesh]

Right, thanks!

Since MATLAB is used for matrix operations, I thought there would be a more efficient way to perform this task. In reality, the dimensions are in the range of 27,000 so it takes a lot of time to compute the output.

Answer 1

[Akira Agata]

I think cellfun would be suitable to this, like:

load('Train&Test.mat');

Result = cell(numel(Test), numel(Train));

for kk = 1:numel(Train)
    Result(:,kk) = cellfun(@xor, repmat(Train(kk),numel(Test),1), Test,...
      'UniformOutput', false);
end