Exploring Online Algorithms for Causal Bandits
This repository contains the code for our CS747 course project on online learning algorithms for causal bandits. You can find our project proposal here
This project explores the Causal Bandit problems and various learning algorithms to solve such problem formulation for various graph topologies while minimizing the cummulative regret.
Also we explore a new model-based approcah to the problem which is using epsilon-greedy strategy with maintaining a similar Bayesian Graph as the true graph and updating as per exploration and in exploiting phase, acting greedily to the graph.
Choose the graphs and algorithm you wanna test in plot.py file.