package prbnmcn-ucb1

  1. Overview
  2. Docs
package prbnmcn-ucb1
Legend:
Page
Library
Module
Module type
Parameter
Class
Class type
Source

Module Ucb1Source

Sourcemodule type Arm_sig = sig ... end

The UCB1 module is parameterised by a finite set of actions presented as an array of abstract "arms", each arm corresponding to an action.

Phantom types used to tag the state of the bandit.

Sourcetype awaiting_reward

awaiting_rewards tags bandits from which an arm was selected and are awaiting the reward associated to this arm.

Sourcetype ready_to_move

ready_to_move tags bandits that are ready to perform another action.

Sourcetype arm_statistics = {
  1. number_of_activations : float;
    (*

    Number of times this arm was activated.

    *)
  2. cumulative_reward : float;
    (*

    Total reward gathered by this arm.

    *)
  3. empirical_reward : float;
    (*

    Average reward gathered by this arm.

    *)
}

Statistics of a given arm.

Sourcemodule type S = sig ... end
Sourcemodule Make (Arm : Arm_sig) : S with type arm = Arm.t