I want to write a function to generate a uniformly distributed random variable and a standard normally distributed random variable (see below c++ code). when i call the unirand function, I always get the same value. Why is that?Thanks!#include <iostream>#include <cstdlib>#include <ctime> #include <cmath> using namespace std; double unirand(int number =1){ srand((unsigned)time(0)); double random; random = (rand()%100+1); random = random /100; return random; };double standardnormal(){ //Box-Muller-transform// double z0, z1; const double pi = 3.14; double u1,u2; u1=unirand(); u2=unirand(); cout << u1 << " = "<<u2<< " <- u1 and u2 have the same value. Why?"<< endl; z0 = sqrt(-2*log(u1))*cos(2*pi*u2); //z1 = sqrt(-2*log(u1))*sin(2*pi*u2); return z0; };int main(){ standardnormal(); system("pause"); };

Because you are setting the random number seed to time(0) every time the function is called:http://www.cplusplus.com/reference/clib ... lib/srand/

Maybe it's random...

QuoteOriginally posted by: HansiBecause you are setting the random number seed to time(0) every time the function is called:http://www.cplusplus.com/reference/clib ... nd/alright, i need to execute "srand((unsigned)time(0));" only at the begining before generating the random numbers (e.g. including it in the int main() ) . although, i dont quite understand why this is the case. if you use srand((unsigned)time(0)); later in the program the system time is differnt and, hence, the seed should be different from when it was first called. thanks anyways

It's basically related to not enough time having passed, e.g. add #include <windows.h> and set it to:Sleep(1000);srand((unsigned)time(NULL));But it's best practice to just set the seed once for simple things like this unless you need to draw the same numbers again for some reason.If you want to use a more robust RNG then the default rand() you might want to check out the Boost RNG:http://www.boost.org/doc/libs/1_43_0/do ... andom.html

Use #include <pthread.h> is on a posix system.

QuoteOriginally posted by: HansiIt's basically related to not enough time having passed, e.g. add #include <windows.h> and set it to:Sleep(1000);srand((unsigned)time(NULL));But it's best practice to just set the seed once for simple things like this unless you need to draw the same numbers again for some reason.If you want to use a more robust RNG then the default rand() you might want to check out the Boost RNG:http://www.boost.org/doc/libs/1_43_0/do ... om.htmlYou guys seem to be getting on like a house on fire, but if I could just inform of my blog on Boost.Random There's a ready-to-run example. Mersenne Twister.rand() is scary..

QuoteOriginally posted by: CuchulainnIf you want to use a more robust RNG then the default rand() you might want to check out the Boost RNG:http://www.boost.org/doc/libs/1_43_0/do ... om.htmlwow, this Boost example makes Java look terse. i didn't know you can make C++ so verbose

QuoteOriginally posted by: jawabeanQuoteOriginally posted by: CuchulainnIf you want to use a more robust RNG then the default rand() you might want to check out the Boost RNG:http://www.boost.org/doc/libs/1_43_0/do ... om.htmlwow, this Boost example makes Java look terse. i didn't know you can make C++ so verboseAn OO option is to create two class hierachies using polymorphic functions and inheritance. Performance is 10 times slower. But another way is CRTP pattern if you insist. BTW that was Hansi's post. BTW SHOW US your Java RNG

QuoteOriginally posted by: Cuchulainn BTW SHOW US your Java RNGstandard one goes like this:double r = Math.random(); if you want more control then:Random rng = new Random(seed);r = rng.nextDouble();r = rng.nextGaussian();If you don't like a standard one, there are many options, e.g. with apache.commons.math:RandomGenerator rng = new MersenneTwister();rng.setSeed(seed);r = rng.nextDouble();

QuoteOriginally posted by: jawabeanQuoteOriginally posted by: Cuchulainn BTW SHOW US your Java RNGstandard one goes like this:double r = Math.random(); if you want more control then:Random rng = new Random(seed);r = rng.nextDouble();r = rng.nextGaussian();If you don't like a standard one, there are many options, e.g. with apache.commons.math:RandomGenerator rng = new MersenneTwister();rng.setSeed(seed);r = rng.nextDouble();Looks nice.

QuoteOriginally posted by: CuchulainnLooks nice.Java standard Random is very slow. i use it only for prototyping. Java interfaces are usually clean and simple, but in this case i'm missing getSeed(). it would be nice to be able to get a seed in the middle of the sequence, so to use it later to reproduce the rest of the sequence.

QuoteOriginally posted by: jawabeanQuoteOriginally posted by: CuchulainnLooks nice.Java standard Random is very slow. i use it only for prototyping. Java interfaces are usually clean and simple, but in this case i'm missing getSeed(). it would be nice to be able to get a seed in the middle of the sequence, so to use it later to reproduce the rest of the sequence.What's the reason? the polympoprhic behaviour or floats. or both?

QuoteOriginally posted by: CuchulainnWhat's the reason? the polympoprhic behaviour or floats. or both?i think it's because they try to make it very safe. they use a class called AtomicLong, which calls a native method. all it does is updates the value of the seed. i couldn't rationalize this implementation though