Last change
on this file since 13780 was
12893,
checked in by gkronber, 9 years ago
|
#2283: experiments on grammatical optimization algorithms (maxreward instead of avg reward, ...)
|
File size:
1.1 KB
|
Line | |
---|
1 | using System;
|
---|
2 | using System.Collections.Generic;
|
---|
3 | using System.Dynamic;
|
---|
4 | using System.Linq;
|
---|
5 | using System.Text;
|
---|
6 | using System.Threading.Tasks;
|
---|
7 | using HeuristicLab.Problems.GrammaticalOptimization;
|
---|
8 |
|
---|
9 | namespace 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 ISequentialDecisionPolicy<in TState> {
|
---|
14 | bool TrySelect(System.Random random, TState curState, IEnumerable<TState> afterStates, out int selectedStateIdx); // selectedState \in afterStates
|
---|
15 |
|
---|
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);
|
---|
18 |
|
---|
19 | void Reset(); // clears all internal state
|
---|
20 |
|
---|
21 | // for introspection
|
---|
22 | double GetValue(TState state);
|
---|
23 | int GetTries(TState state);
|
---|
24 | }
|
---|
25 | }
|
---|
Note: See
TracBrowser
for help on using the repository browser.