Free cookie consent management tool by TermsFeed Policy Generator

source: branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/IPolicy.cs @ 11770

Last change on this file since 11770 was 11770, checked in by gkronber, 9 years ago

#2283: worked on generic sequential search alg with bandit policy as parameter

File size: 1.1 KB
1using System;
2using System.Collections.Generic;
3using System.Dynamic;
4using System.Linq;
5using System.Text;
6using System.Threading.Tasks;
7using HeuristicLab.Problems.GrammaticalOptimization;
9namespace HeuristicLab.Algorithms.Bandits {
10  // this interface represents a policy for episodic reinforcement learning (with afterstates)
11  // here we assume that a reward is only recieved at the end of the episode and the update is done only after an episode is complete
12  // we also assume that the policy can fail to select one of the followStates
13  public interface IPolicy<TState> {
14    bool TrySelect(Random random, TState curState, IEnumerable<TState> afterStates, out TState selectedState); // selectedState \in afterStates
16    // state-trajectory are the states of the episode, at the end we recieved the reward (only for the terminal state)
17    void UpdateReward(IEnumerable<TState> stateTrajectory, double reward);
19    void Reset(); // clears all internal state
21    // for introspection
22    double GetValue(TState state);
23    int GetTries(TState state);
24  }
Note: See TracBrowser for help on using the repository browser.