Implementations of UCB1, Bayesian UCB, Epsilon Greedy, and EXP3 bandit algorithms on the Movielens-20m dataset. Algorithms are evaluated offline using replay.
To reproduce:
git clone https://github.com/jldbc/bandits
cd bandits/bandits
bash run.sh
Impementation details and results
Final results:
