CODEWITHSUNDEEP
pythonc++11gaussian
John Difool
John Difool

Reputation: 5702

Gaussian distribution with mean and sigma in C++11

I am trying to get gaussian distribution with mean and sigma in C++11. I have been successful at converting Python to C++ but I have a doubt about the way I am initializing the random generator. Do I need to call random_device() and mt19937() inside the call to get a distribution or can I just call them once statically and re-use those all the time? What is the cost of leaving the code as it is?

# Python

# random.gauss(mu, sigma)
# Gaussian distribution. mu is the mean, and sigma is the standard deviation.

import random

result = random.gauss(mu, sigma)


// C++11

#include <random>

std::random_device rd;
std::mt19937 e2(rd());

float res = std::normal_distribution<float>(m, s)(e2);

Upvotes: 0

Views: 549

Answers (1)

xuhdev
xuhdev

Reputation: 9333

There are two parts of the algorithm:

  • uniform random number generator,
  • and convert the uniform random number to a random number according to Gaussian distribution.

In your case, e2 is your uniform random number generator given the seed rd, std::normal_distribution<float>(m, s) generates an object which does the 2nd part of the algorithm.

The best way to do it is:

// call for the first time (initialization)
std::random_device rd;
std::mt19937 e2(rd());
std::normal_distribution<float> dist(m, s);
// bind the distribution generator and uniform generator
auto gen_gaussian = std::bind(dist, e2);

// call when you need to generate a random number
float gaussian = gen_gaussian();

If you don't care about which uniform random number generator to use, you can use std::default_random_engine instead of std:mt19937.

Upvotes: 2

Related Questions