COMPUTING “ELO RATINGS” OF MOVE PATTERNS IN THE GAME OF GO

By Rémi Coulom. 2007

359 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Clinical Psychology, Economics and Econometrics, Game & Decision Theory

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Clinical Psychology

In Economics and Econometrics

In Game & Decision Theory

Citations in the corpus

Mastering the game of Go without human knowledge (2017, background): “MCTS programs have previously achieved strong amateur level in Go 46,47 , but used substantial domain expertise: a fast rollout policy, based on handcrafted features 13,48 , that evaluates positions by running simulations until the end of the game; and a tree policy, also based on handcrafted…”
MoHex 2.0: A Pattern-Based MCTS Hex Player (2013, methodology): “Following Go players such as Erica, we apply the supervised learning algorithm minorization-maximization (MM) [12] to learn the win correlation of board patterns, and use the resulting pattern weights in both the leaf selection and simulation phases of MCTS.”
Monte Carlo Tree Search: a review of recent modifications and applications (2021, background): “(2018), Brügmann (1993), Coulom (2007), Drake and Uurtamo (2007) Hex Gao et al.”
The grand challenge of computer Go (2012, background): “The second approach is similar, but uses supervised learning to weight the patterns, so as to match the distribution with which human experts select those patterns [13].”
Mastering the game of Go with deep neural networks and tree search (2016, methodology): “Prior work has focused on handcrafted patterns 50 or learning rollout policies by supervised learning 13 , reinforcement learning 16 , simulation balancing 51,52 or online adaptation 30,53 ; however, it is known that rollout-based position evaluation is frequently inaccurate 54 .”
Transformers are Sample Efficient World Models (2022, background): “MuZero leverages MCTS as a policy improvement operator, by unrolling multiple hypothetical trajectories in the latent space of a world model.”
Monte-Carlo tree search and rapid action value estimation in computer Go (2011, neutral): “Monte-Carlo tree search [1] is a new paradigm for search, which has revolutionised computer Go [2, 3], and is rapidly replacing traditional search algorithms as the method of choice in challenging domains such as General Game Playing [4], Amazons [5], Lines of Action [6], multi-player card games [7,”
Temporal-difference search in computer Go (2012, neutral): “Previous work in simulation-based search has largely been restricted to Monte-Carlo methods (Tesauro and Galperin, 1996; Kocsis and Szepesvari, 2006; Gelly et al., 2006; Gelly and Silver, 2007; Coulom, 2007).”
A Survey of Monte Carlo Tree Search Methods (2012, methodology): “In Sparse UCT, a node may have several children corresponding to the same move, each child corresponding to a different stochastic outcome of that move.”
Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos (2022, background): “Imitation learning is simplest when demonstrations are labeled with corresponding actions Imitating labeled trajectories has seen success in aerial vehicles, 40,41 self-driving cars, 42,43 board games, 10,44 and video games.”
Progressive Strategies for Monte-Carlo Tree Search (2008, methodology): “The simplest rule, proposed by [Coulom (2007)], is to expand one node per simulation.”
Multi-armed bandits with episode context (2011, methodology): “In computer Go it is common to prepare a predictor in advance via an offline process, then freezing it for later use in bandit-based search (Gelly & Silver 2007; Coulom 2007a).”
Monte Carlo Tree Search in Hex (2010, background): “As with Go, it is beneficial to apply some knowledge during simulated games [9].”
On-the-Job Learning with Bayesian Decision Theory (2015, neutral): “Human n-query: The majority vote of n human crowd workers was used as a prediction.”
Pairwise ranking aggregation in a crowdsourced setting (2013, methodology): “This numerical problem can be addressed via the so-called virtual node regularization which has been used in ranking problems under different settings [4, 11].”
Large-Scale Optimization for Evaluation Functions with Minimax Search (2014, background): “For example, the function could be a linear combination of weighted features, i.e., e(p,w) = ∑ iwifi(p), where fi(p) is the i-th feature value of position p.”
Training Deep Convolutional Neural Networks to Play Go (2015, methodology): “We expect the target function to be highly complex, since it is fair to assume human experts think in complex, non-linear ways when they choose moves.”
Sample-based learning and search with permanent and transient memories (2008, methodology): “9× 9 Go programs based on the UCT algorithm(Kocsis&Szepesvari, 2006)havenow achieved masterlevel(Gelly&Silver, 2007;Coulom, 2007).”