C++ NAN.
I WILL FIND YOUR SOURCE.
AND I WILL KILL YOU, DESTROY YOU.
I WILL DIE TRYING.
YOU DISGUST ME.
I HATE YOU.

FUCK YOU NAN

I HATE YOU

@DGApps Its not the concept of NaN that I hate, its that when you are iterating thousands of times, which subsequently have functions which themselves have hundreds of iterations, and with no error and no warnings given, it becomes impossible to locate the source of the NaN.

@DGApps Yes I 100000% agree

You might, I am just guessing here, actually hate your bugy algorithm. Why would you hate NaN. That poor fella is not even a number.

@anicka-burova I'm designing this neural net which involves a lot of mathematics. The problem arises due to the fact that its impossible to pinpoint where the first NaN appears, and since every piece of data is dependant on the other, it fucks up everything. A simple error would be really good.

@harambae put debug asserts, that is what I did, when was hunting for it. It did work. Just simple assert(a == a) will do the magic. I am sure you know, all numbers are equal to self, but not NaN.

@harambae doing simple error requires runtime checks. And you don't want that in your neural code, which is performance critical. Using asserts, you will get to your bug. NaN is always a bug in the algorithm.

@anicka-burova Yes, infact I was just implementing this. Hopefully this will show the errors.

Just don't use exceptions! It is a bug, so it needs bug fix and not a shush :)