Adding expert knowledge and exploration in monte-carlo tree search
read more
Citations
A Survey of Monte Carlo Tree Search Methods
Fuego—An Open-Source Framework for Board Games and Go Engine Based on Monte Carlo Tree Search
Multi-armed bandits with episode context
Ensemble Determinization in Monte Carlo Tree Search for the Imperfect Information Card Game Magic: The Gathering
Current Frontiers in Computer Go
References
Bandit based monte-carlo planning
Efficient selectivity and backup operators in Monte-Carlo tree search
Progressive Strategies for Monte-Carlo Tree Search
Combining online and offline knowledge in UCT
Computing “elo ratings” of move patterns in the game of go
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the key point in improving the MC engine?
Reducing the probability of simulations in which a group which should clearly live dies (or vice versa) improves the overall performance of the algorithm.
Q3. What is a Monte Carlo Tree Search?
MCTS consists in building a tree, in which nodes are situations of the considered environnement and branches are the actions that can be taken by the agent.
Q4. What is the idea of the modification of Monte-Carlo?
The idea is to increase the number of locations at which Monte-Carlo simulations can find pattern-matching in order to diversify the Monte-Carlo simulations.
Q5. What is the main idea of MCTS?
After each simulation from the current position (the root) until the end of the game, the win and loss statistics are updated in every node concerned by the simulation, and a new node corresponding to the first new situation of the simulation is created.
Q6. What are the two improvements that were made to the Monte-Carlo simulator?
The authors present below two new improvements, both of them centered on an increased diversity when the computational power increases; in both cases, the improvement is negative or negligible for small computational power and becomes highly significant when the computational power increases.
Q7. Who accepted to play and discuss test games against MoGo?
Many thanks to the french federation of Go and to Recitsproque for having organized an official game against a high level human; many thanks also to the several players from the French and Taiwanese Federations of Go who accepted to play and discuss test games against MoGo.
Q8. What are the rules used in the implementation of 19x19 Go?
The following rules are used in their implementation in 19x19, and improve the results:– Territory line (i.e. line number 3), Line of death (i.e. first line), Peep-connect (ie. connect two strings when the opponent threatens to cut), Hane (a move which “reaches around” one or more of the opponent’s stones), Threat, Connect, Wall, Bad Kogeima(same pattern as a knight’s move in chess), Empty triangle (three stones making a triangle without any surrounding opponent’s stone).
Q9. What is the main idea of Monte-Carlo Tree Search?
Other related algorithms have been proposed as in [5], essentially using a decreasing impact of a heuristic (pattern-dependent) bias as the number of simulations increases.
Q10. What was the effect of a modification of the Monte-Carlo simulator on the global?
The following parameters had to be modified, when this model was included in H:– time scales for the convergence of the weight of online statistics to 1 (see Eq. 1) are increased;– the number of simulations of a move at a given node before the subsequent nodes is created is increased (because the computational cost of a creation is higher).– the optimal coefficients of expert rules are modified;– importantly, the results were greatly improved by adding the constant C (see Eq. 1).
Q11. What is the purpose of the term nakade?
The authors will use the term nakade to design a situation in which a surrounded group has a single large internal, enclosed space in which the player won’t be able to establish two eyes if the opponent plays correctly.
Q12. What are the main objectives of the Monte-Carlo simulator?
plenty of experiments around increasing the level of the Monte-Carlo simulator as a stand-alone player have given negative results - diversity and playing strength are too conflicting objectives.