This is a demonstration of the work done by D.J. Best and N.I. Fisher (found here)
This is a finite domain analogue to the Gaussian distribution. The von Mises distribution lies in the interval [-π, π], so to normalize it, you would have to take your output n and do the the following to it: n_norm = (n + π)/2π.
You can use this distribution to generate a random number over an interval, with a bias towards the center of that interval. This is quite handy sometimes and more practical than using a Gaussian distribution that could generate any real number.
A random sampling of the distribution can be found by running python3 main.py