This repository contains a C++ implementation of the Multi-Armed Bandit described in "Reinforcement Learning - An introduction" by Sutton and Barto.

The following exploration policies have been implemented:

• Epsilon-greedy
• Boltzmann (softmax)
• UCB
• Gradient Bandit Algorithm with softmax action preferences

To compile Gaussian Bandit code use:

make -f Makefile.gaus

run the executable with ./a.out and insert the requested parameters.

In the same way, to compile Bernoulli Bandit code use:

make -f Makefile.bern