Thread: Help with random numbers

Hello!

I would be grateful if someone could explain why this code produces identical random numbers (vectors). I am trying to generate a series of uniform and (flat)-dirichlet distributed random vectors of size n. The dirichlet random variables are obtained by first generating exponential random variables and then computing a vector of shares in the sum of n exponential random variables. But the vectors are identical each time. Is this a seed issue?

Thank you for your help
serge

Code:

////////////////////////////////////////////////////////////////////////////////////////////////////
#include <algorithm>
#include <functional>
#include <iostream>
#include <random>
#include <string>
////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////////////////////////////////////////////////////////////
int main ( int argc , char* argv[ ] ) {
	// DIMENSION ( default 3 )
	int n = 3;
	if ( argc > 1 ) n = std::stoi ( argv[ 1 ] );

	// ITERATIONS ( default 100 )
	long total_iterations = 100ll;
	if ( argc > 2 ) total_iterations = std::stol ( argv[ 2 ] );

	// ROUND OUTPUT TO DECIMAL PLACES ( default 3 )
	int display_precision = 3;
	if ( argc > 3 ) display_precision = std::stoi ( argv[ 3 ] );

	// SET FORMAT FOR DECIMAL OUTPUT
	std::cout.unsetf ( std::ios::floatfield );
	std::cout.precision ( display_precision ); // number of decimal places
	std::cout.setf ( std::ios::fixed , std::ios::floatfield ); // display fixed length

	// PSEUDO-RANDOM NUMBER GENERATOR
	std::random_device rnd_device;
	std::mt19937 mersenne_engine( rnd_device( ) );
	std::uniform_real_distribution<double> uniform_dist( 0.0 , 1.0 );
	std::exponential_distribution<double> exponential_dist( 1.0 );
	auto uniform_generator = std::bind( uniform_dist , mersenne_engine );
	auto exponential_generator = std::bind( exponential_dist , mersenne_engine );

	// PSEUDO-RANDOM VECTORS
	std::vector<double> uniform_vec( n , 0.0f );
	std::vector<double> exponential_vec( n , 0.0f );
	double sum_exponential_vec = 0.0f;
	std::vector<double> dirichlet_vec( n , 1.0f );

	for ( int i = 0; i < total_iterations; ++i ) {
		generate( begin( uniform_vec ) , end( uniform_vec ) , uniform_generator );
		generate( begin( exponential_vec ) , end( exponential_vec ) , exponential_generator );
		sum_exponential_vec = std::accumulate( exponential_vec.begin( ) , exponential_vec.end( ) , 0.0f );
		transform( exponential_vec.begin( ) , exponential_vec.end( ) , dirichlet_vec.begin( ), std::bind2nd( std::divides<double>( ) , sum_exponential_vec ) );

		// PRINT OUTPUT
		std::cout << "inputs: " << n << " " << total_iterations << " " << display_precision << std::endl;
		std::cout << "uniform: ";
		for ( auto j: uniform_vec ) std::cout << j << ' ';
		std::cout << std::endl;
		std::cout << "exponential: ";
		for ( auto j: exponential_vec ) std::cout << j << ' ';
		std::cout << "sum: " << sum_exponential_vec << std::endl;
		std::cout << "flat dirichlet: ";
		for ( auto j: dirichlet_vec ) std::cout << j << ' ';
		std::cout << std::endl;
		std::cout << std::endl;
	}

	return 0;
}

random is not random:
if that code is only ran once on each start up of program then yes it will always return the same, because random is not random. if it is in a continuing run in the program and calling for a random number, or numbers they will change, but if you run that again you may/will start to see the same 'random' number patterns because random is not random. It relies on a mathematical algorithm to produce a 'random' number.street word says, this is why urandom and seed ( which I am not too familiar with) produces a larger pool of random numbers.