A Python library for all popular multiarmed bandit algorithms.
The following algorithms are currently being explored before we set a roadmap for the first release.
- Epsilon Bandit:
- Epsilon-greedy strategy (done)
- Epsilon-first strategy (in progress)
- Epsilon-decreasing strategy
- Epsilon-adaptive strategy
- Bayesian Bandit
- Thompson Sampling
- Contextual Bandit
- Linear Classifier
- LinUCB (Upper Confidence Bound) algorithm:
- LinRel (Linear Associative Reinforcement Learning) algorithm:
- Non-linear Classifier
- UCBogram algorithm
- NeuralBandit algorithm
- KernelUCB algorithm
- Bandit Forest algorithm
- Constrained
- UCB-ALP algorithm
- Greedy
- Contextual-Epsilon-greedy strategy
- Linear Classifier
- Adverserial Bandit
- Dueling Bandit
- Collaborative Bandit
- COFIBA