Sample from Normal Distrabution

My first instinct would be to use boost::random. The source code for their normal distribution is also freely available if you're interested in the underlying algorithm and not just application.

Here you go... you just need to supply a random number between 0 and 1, the function then returns a normally distributed number. Multiply by your standard deviation and add the mean.

http://home.online.no/~pjacklam/notes/invnorm/

twisted! yeah that's just what i wanted thanks

A value from the normal distribtion is unbounded, so choosing a number between 0 and 1 can't be right. If want to generate the value from a uniform distribution then you take the sum of repeated samples, then divide by root n(n being the number of samples) and you'll be left with a normal distribution of some description, depending on the initial bounds of the uniform distribution.

I could be wrong about root n, but I don't think I am. This is all an application of the central limit theorem.

http://mathworld.wolfram.com/CentralLimitTheorem.html

There will ofcourse be far better ways of generating it, so a library is probably the thing to use.

Easy:

Set a and r to be two random numbers between 0 and 1.
Then

x = sqrt(-2*log(r)) * sin(2*pi*a)

is normally distributed with variance 1 and mean value 0.

Lutz