Bootstrapping from Game Tree Search

By Joel Veness, David Silver, William T. B. Uther, Alan D. Blair. 2009

67 citations recorded by the source corpus.

How researchers see this paper

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

In Reinforcement Learning

Citations in the corpus

Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm (2017, background): “Meep ( 32 ) evaluated positions by a linear evaluation function based on handcrafted features.”
Giraffe: Using Deep Reinforcement Learning to Play Chess (2015, background): “A more recent attempt is the Meep project by Veness et al. [16], whose evaluation function is a linear combination of hand-designed features, and where the weights are trained using reinforcement learning.”
The Hanabi Challenge: A New Frontier for AI Research (2019, methodology): “This leads to methods than can combine planning with RL, as first demonstrated by TD-Leaf [65] and TreeStrap [66].”
Learning to Search with MCTSnets (2018, methodology): “…algorithm apply reinforcement learning to ﬁnd an evaluation function that combines with minimax search to produce an accurate root evaluation (Veness et al., 2009); while comparison training (Tesauro, 1988) applies supervised learning to the same problem; these methods have been successful…”
Temporal-difference search in computer Go (2012, unknown): “Our framework draws extensively from successful examples of reinforcement learning in classic games such as chess (Veness et al. 2009), checkers (Schaeffer et al. 2001), and backgammon (Tesauro 1994).”
Beyond the One Step Greedy Approach in Reinforcement Learning (2018, background): “The celebrated MCTS al-gorithm, which instantiates several steps of lookahead improvement, encompasses additional historical impressive accomplishments dating back to the past century and previous decade (Tesauro & Galperin, 1997; Sheppard, 2002; Bouzy & Helmstetter, 2004; Veness et al., 2009).”
Large-Scale Optimization for Evaluation Functions with Minimax Search (2014, methodology): “However, strong human players have ratings of more than 3, 000 points at ICC, and this difference means these programs have not reached the top level of chess programs; that is, evaluation functions tuned by reinforcement learning have not yet reached the level of the best-handcrafted evaluation fun”
Analysis of Evaluation-Function Learning by Comparison of Sibling Nodes (2011, background): “In contrast, it is interesting that chess programs with evaluation functions tuned by machine learning have not outperformed programs with hand-tuned evaluation functions [2,18].”