Skip to content

Latest commit

 

History

History
16 lines (13 loc) · 475 Bytes

README.md

File metadata and controls

16 lines (13 loc) · 475 Bytes
Raw

Bandit algorithms.

For a Reinforcement Learning class, I worked on a few popular algorithms for the bandit problem. Among them were :

  • Epsilon-greedy bandit
  • BESA
  • Softmax
  • UCB1
  • Thompson sampling
  • KL-UCB

Bandits are implemented in agent.py

How to use ?

For the purpose of the class, each agent was tested on a specific configuration : 1000 rounds for 2000 agents in parallel : python main.py --niter 1000 --batch 2000 Use python main.py -h to know more.