CODEWITHSUNDEEP
arraysmatlabbsxfun
Umut Tabak
Umut Tabak

Reputation: 1942

arrayfun to bsxfun possible

I know that bsxfun(which works fast!) and arrayfun(as far as I could understand, uses loops internally which is expected to be slow) are intended for different uses, at least, at the most basic level.

Having said this, I am trying

  1. to sum up all the numbers in a given array, say y, before a certain index
  2. add the number at the specific location(which is the number at the above index location) to the above sum.

I could perform this with the below piece of example code easily:

% index array
x = [ 1:6 ]; % value array
y = [ 3 3 4 4 1 1 ];
% arrayfun version
o2 = arrayfun(@(a) ...
              sum(y(1:(a-1)))+...
              y(a), ...
              x)

But it seems to be slow on large inputs.

I was wondering what would be a good way to convert this to a version that works with bsxfun, if possible.

P.S. the numbers in y do not repeat as given above, this was just an example, it could also be [3 4 3 1 4 ...]

Upvotes: 0

Views: 108

Answers (2)

Baback Kh
Baback Kh

Reputation: 1

if you have a supported GPU device, you can define your variables as gpuArray type since arrayfun, bsxfun and pagefun are compatible with GPUs. GPU computing is supposed to be faster for large data.

Upvotes: 0

verbatross
verbatross

Reputation: 607

Is x always of the form 1 : n? Assuming the answer is yes, then you can get the same result with the much faster code:

o2 = cumsum(y);

Side note: you don't need the brackets in the definition of x.

Upvotes: 1

Related Questions