M. Buro. Statistical feature combination for the evaluation of game positions. JAIR, 3:373--382, 1995 . Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
The Games Computers (and People) Play - Schaeffer (2000) (1 citation) (Correct)
....each entry in each pattern (46) for each phase of the game (11) are determined by linear regression. There are over 1.5 million table entries that need to be determined. The data was trained using 11 million scored positions obtained from self play games and practice games against another program [87]. The evaluation function is completely table driven. Given a position, all 46 patterns are matched against the position, with a successful match returning the associated weight. These weights are summed to get the overall evaluation which approximates the nal disc di erential. Opening books are ....
M. Buro. Statistical feature combination for the evaluation of game positions. Journal of Articial Intelligence Research, 3:373-382, 1995.
KEYANO Unplugged - The Construction of an Othello Program - Brockington (1997) (1 citation) (Correct)
....value (such as 1) to represent a winning position, while a minimal value (such as Gamma1) represents a lost position. However, these maximal and minimal values are not absolute. For example, Logistello s former evaluation function attempts to determine the probability of winning from a position [5]. This implies that the evaluation of a position varies between 0 and 1 3 . Logistello s new evaluation function and the evaluation function within Keyano attempt to approximate the final outcome of the game, returning a value between Gamma64 and 64. 3.1 Features of the Evaluation Function ....
....Bill used a quadratic discriminant function [10] in an attempt to improve upon the linear evaluation function. Buro later showed data that Fisher s linear discriminant and the quadratic discriminant function are both weaker than a linear evaluation function determined by logistic regression [5]. Despite the advantages of these approaches, Keyano has always used a linear regression to determine the constants to be used in the evaluation function. The reason behind this is that the author believes that the expected disc count at the end of the game is a natural metric for success, rather ....
M. Buro. Statistical Feature Combination for the Evaluation of Game Positions. Journal of Artificial Intelligence Research, 3:373--382, 1995.
Improving Heuristic Mini-Max Search by Supervised Learning - Buro Self-citation (Buro) (Correct)
....in August 1997. Over the years, Logistello s evaluation function changed considerably: from a classic form featuring only a hand full manually weighted features, over a version that estimated configuration values using the naive Bayes approach and weighted whole patterns by logistic regression [4], to its current form utilizing approximately 100,000 binary features in conjunction with over 1.2 million automatically tuned parameters. In each step the evaluation accuracy and speed was increased significantly. Table 1 shows experimental evidence for the considerable accuracy gain obtained ....
M. Buro. Statistical feature combination for the evaluation of game positions. JAIR, 3:373--382, 1995 .
Experiments with Multi-ProbCut and a New High-Quality Evaluation.. - Buro (1997) (4 citations) Self-citation (Buro) (Correct)
No context found.
M. Buro. Statistical Feature Combination for the Evaluation of Game Positions, JAIR 3, 373--382.
Improving Heuristic Mini-Max Search by Supervised Learning - Buro (2000) Self-citation (Buro) (Correct)
....in August 1997. Over the years, Logistello s evaluation function changed considerably: from a classic form featuring only a handfull manually weighted features, over a version that estimated configuration values using the naive Bayes approach and weighted whole patterns by logistic regression [5], to its current form utilizing approximately 100,000 binary features in conjunction with over 1.2 million automatically tuned parameters. In each step the evaluation accuracy and speed was increased significantly. Table 1 shows experimental evidence for the considerable accuracy gain obtained ....
M. Buro. Statistical feature combination for the evaluation of game positions. JAIR, 3:373--382, 1995 2 . 14
Experiments with Multi-ProbCut and a New High-Quality Evaluation.. - Buro (1997) (4 citations) Self-citation (Buro) (Correct)
No context found.
M. Buro. Statistical Feature Combination for the Evaluation of Game Positions, JAIR 3, 373--382.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC