gaussian-bandits

Experimentation with Gaussian multi-armed bandits and action selection functions.

Status: Personal / university coursework archive — not actively maintained. Dependencies and tooling may be outdated.

How to run

Use Python 3 and Jupyter (pip install jupyter numpy matplotlib or similar). Open the main notebook in this repository and run cells top to bottom.

Description

The notebook contains a simple definition of a bandit and arm classes, as well as different functions for selecting the next action (arm to pull). The implementation is based on the referenced book by S. Sutton and A. Barto (see below).

Note: The functionality of the notebook can easily be extended by defining different types of arms, reward distributions and action selection functions and compare them.

References

[1] Richard S. Sutton and Andrew G. Barto. 2018. Reinforcement Learning: An Introduction. A Bradford Book, Cambridge, MA, USA.

[2] Lattimore, T., & Szepesvári, C. (2020). Bandit Algorithms. Cambridge: Cambridge University Press. doi:10.1017/9781108571401