public class MonteCarloSearch extends MCTSPredictor
Created by piers on 20/12/16.
agentID, agents
DEFAULT_ITERATIONS, DEFAULT_ROLLOUT_DEPTH, DEFAULT_TREE_DEPTH_MUL, logger, NO_LIMIT, random, rolloutDepth, roundLength, treeDepthMul
Constructor and Description |
---|
MonteCarloSearch(Agent policy)
Constructs a new MC Search player with the given policy to use
|
Modifier and Type | Method and Description |
---|---|
protected int |
calculateTreeDepthLimit(GameState state) |
Action |
doMove(int agentID,
GameState state)
Standardised interface for game playing agents.
|
protected Action |
selectActionForRollout(GameState state,
int agentID) |
String |
toString() |
buildPMCTSND, doSuperMove, parseAgents, select, selectActionForExpand
buildMCTSND, expand, rollout
public MonteCarloSearch(Agent policy)
policy
- The policy to use instead of Random rolloutspublic Action doMove(int agentID, GameState state)
Agent
The agent gets a copy of the game state and it's agent ID and should return a move.
doMove
in interface Agent
doMove
in class MCTSPredictor
agentID
- the ID of this agentstate
- the current state of the gameprotected Action selectActionForRollout(GameState state, int agentID)
selectActionForRollout
in class MCTSPredictor
protected int calculateTreeDepthLimit(GameState state)
calculateTreeDepthLimit
in class MCTS
public String toString()
toString
in class MCTSPredictor
Copyright © 2017 FOSS Galaxy. All rights reserved.