Help random number generation

[smilingfrog]

Hello, first post.
I have a problem that I want to try to solve with a genetic algorithm.
I decided to try to learn to do this using c++, with no significant recent experience.

I am running Ubuntu 10.10, with gcc version 4.4.3, and codeblocks 10.05svn7671-1.

To help learn the language for the genetic algorithm, I googled genetic algorithm to get to this code http://www.ai-junkie.com/files/Basic_Genetic_Algorithm.zip.

When I compile the code, it doesn't work as the site claims.
The first errors I have corrected relate to multiple definitions of int i.

My first question regards this code:

//returns a float between 0 & 1
#define RANDOM_NUM		((float)rand()/(RAND_MAX+1))

...

string	GetRandomBits(int length)
{
	string bits;
	for (int i=0; i<length; i++)
	{
		if (RANDOM_NUM > 0.5f)
			bits += "1";
		else
			bits += "0";
	}
	return bits;
}

This seems to generate random negative numbers, and therefore the returned string is only 0s. Why is this?

Thanks.
G

[ResidentBiscuit]

Hm, why are you using define macro for this? Anyways,
Try changing it to
((float)rand()/(float)(RAND_MAX+1))
RAND_MAX is an int macro with the maximum value of rand(), so you need to cast it to a float as well.

EDIT:
I'd also recommend seeding this, otherwise you'll get the same series of numbers each time

[Peter87]

The problem is that RAND_MAX == INT_MAX with your compiler so when you add 1 you get an overflow so what you have is (float)rand()/(INT_MIN).

Casting before adding avoids the overflow rand()/((float)RAND_MAX + 1) but the precision of float is so bad that adding one doesn't make a difference in this case so you can just write (float)rand()/RAND_MAX

[eypros]

#define RANDOM_NUM ((float)rand()/(RAND_MAX+1))

It's not a really good practice to define functions this way but I move on.

I think the problem is that RAND_MAX is defined as the bigger number rand() can return and by adding 1 to it it turns to negative (at least in my implementation). So basically every float is negative and of course lesser than 0.5 so bits always adds 0s. Delete +1 in the above code and it should produce 1s also.

[smilingfrog]

Thank you for the replies. Again, this is not my code, but something I found on the net to help me learn.

@eypros: What would be best practice for this? My understanding is the #define tag is a preprocessor comment, so it isn't really a function. One can't create a variable named RANDOM_NUM, or the number will always be the same.

Why would a compiler accept RAND_MAX + 1 and not make it a negative value?

[ResidentBiscuit]

Define macros are merely a text to text conversion. It's similar to using a const. Better practice would be just to either make it function, or write it out of you only use it once