Crate ratel_bandit

source ·
Expand description

§Ratel

A library for running multi-armed bandit simulations.

The simulation has two components: a Bandit with some number of arms which each has a payout according to some distribution, and an Agent that uses some strategy to learn the optimal bandit arm. A Game then consists of an Agent learning about a Bandit by pulling arms according to its preferred strategy.

Structs§

  • A bandit whose arms distribute rewards according to the binomial distributions.
  • Agent that follows the Epsilon-Greedy Algorithm.
  • A bandit whose arms distribute rewards according to the exponential distributions.
  • Structure to make the Agent interact with the Bandit.
  • A bandit whose arms distribute rewards according to the gamma distributions.
  • A bandit whose arms distribute rewards according to the Gaussian distributions.
  • Agent that follows the Greedy Algorithm.
  • Stepper with a harmonically decreasing step size.
  • A bandit whose arms distribute rewards according to the Log Normal distributions.
  • Agent that follows the Optimistic Algorithm.
  • Counter for cases where individual records must be maintained.

Traits§

  • A trait for common members of the Agents.
  • A trait for common members of the Bandits
  • A trait for common methods required by all counters.
  • Trait containing all methods required by all steppers.

Functions§

  • Prints the amount of time that has elapsed since a timer was started.