Why do I keep getting Nan when I use bsxfun?

Hi, I have two vectors YearAlloc.LoadAlloc of size 1*688 and Gen.Capacity 1*688. The size of YearAlloc.LoadAlloc keeps on changing; i.e the 688 goes on increasing. I want to divide the two vectors to get a vector with one column. I tried

bsxfun(@rdivide,YearAlloc.LoadAlloc.',Gen.Capacity)

It creates Nan where YearAlloc.LoadAlloc is 0. I don't want the Nan. Can you please help me out.

Answer 1

[Walter Roberson]

0/0 is defined as NaN, but non_zero_value / 0 is defined as sign(non_zero_value)*inf

You are not exactly getting Nan where the LoadAlloc is 0: you are getting it for the combinations where both LoadAlloc and Capacity are 0.

You cannot prevent the NaN except by not dividing 0 by 0. For example,

bsxfun(@rdivide,YearAlloc.LoadAlloc.', max(realmin, Gen.Capacity))

... provided that the Capacity is never negative.

The appropriate way to handle this is to use isnan() afterwards to detect the nans and to assign in whatever value you feel is appropriate for division of 0 by 0, whether you feel the result should be -inf or +inf or 0 or 1 (all of which can be justified.)

[Aditi Bhalerao]

Thank you very much for the response. I used isnan() as below but it did not solve the problem.

for i = 1 : length(YearAlloc.AllocPerCapacity)
TF(i) = isnan(YearAlloc.AllocPerCapacity(i));
    if TF(i) == 1
        YearAlloc.AllocPerCapacity(i) = 0 ;
    end
end

[Walter Roberson]

YearAlloc.AllocPerCapacity( isnan(YearAlloc.AllocPerCapacity) ) = 0;

No loop needed.

[Aditi Bhalerao]

It worked. Thank you very much for the help.