applying log to matrix converts from real to complex: why?

Hello everyone,

I'm experiencing a very strange behaviour from matlab: I have a 132x4 matrix of real numbers and I need to take their logs. Of course, if I take each number's log individually, it works fine. If I instead use the function log on the whole matrix transf = log(matrix); numbers are converted to their logs from row 1 to 64, but from row 65 to the end they are converted to complex numbers (NaNs as well).

What can this be caused by?

Thanks in advance

Answer 1

[Azzi Abdelmalek]

maybe you have negativ numbers

 log(-1.5)

it's a complex logarithm

http://en.wikipedia.org/wiki/Complex_logarithm

[mattia]

ok very stupid of me, just noticed there were some negative zeros, thanks for the suggestion!

[Wayne King]

but that does not explain your claim of NaNs or that there is a difference between taking the log of individual elements and the log of the matrix.

Answer 2

[Wayne King]

It makes perfect sense to me that log() can produce complex numbers, but I'm confused by the NaNs. Can you give an example of an entry in your matrix such that log(that_entry) = NaN?

Also, can you create a small matrix example such that the log of the individual elements is different than when you apply log() to the matrix.

 a = randn(4,4);
 log(a)

Whether they are in a matrix or not should not matter.

[Matt Fig]

log(nan) is the only thing I can think of here.

[mattia]

yes that is what i meant. since in position 64,1 there was a negative number, i guess it treated all subsequent values as complex, so NaNs became Nan+00000000i