source: branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/IPolicy.cs @ 11744

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using HeuristicLab.Problems.GrammaticalOptimization;

namespace HeuristicLab.Algorithms.Bandits {
  // this interface represents a policy for reinforcement learning
  public interface IPolicy<in TState, TAction> {
    TAction SelectAction(Random random, TState state, IEnumerable<TAction> actions);
    void UpdateReward(TState state, TAction action, double reward, TState newState); // reward received when after taking action in state and new state
    bool Done(TState state); // for deterministic MDP with deterministic rewards and goal to find a state with max reward
  }

  public interface IGrammarPolicy : IPolicy<ReadonlySequence, ReadonlySequence> {

  }
}