Abstract
Despite the undisputed strength of today’s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.
| Original language | English (US) |
|---|---|
| Title of host publication | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
| Publisher | Springer Verlag |
| Pages | 426-433 |
| Number of pages | 8 |
| Volume | 1821 |
| ISBN (Print) | 3540676899, 9783540450498 |
| DOIs | |
| State | Published - 2000 |
| Externally published | Yes |
| Event | 13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000 - New Orleans, United States Duration: Jun 19 2000 → Jun 22 2000 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 1821 |
| ISSN (Print) | 03029743 |
| ISSN (Electronic) | 16113349 |
Other
| Other | 13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000 |
|---|---|
| Country | United States |
| City | New Orleans |
| Period | 6/19/00 → 6/22/00 |
Fingerprint
Keywords
- Computer chess
- Concept learning
- Pattern recognition
ASJC Scopus subject areas
- Computer Science(all)
- Theoretical Computer Science
Cite this
Learning middle-game patterns in chess : A case study. / Kubat, Miroslav; Žižka, Jan.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1821 Springer Verlag, 2000. p. 426-433 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1821).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Learning middle-game patterns in chess
T2 - A case study
AU - Kubat, Miroslav
AU - Žižka, Jan
PY - 2000
Y1 - 2000
N2 - Despite the undisputed strength of today’s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.
AB - Despite the undisputed strength of today’s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.
KW - Computer chess
KW - Concept learning
KW - Pattern recognition
UR - http://www.scopus.com/inward/record.url?scp=84957899807&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84957899807&partnerID=8YFLogxK
U2 - 10.1007/3-540-45049-1_52
DO - 10.1007/3-540-45049-1_52
M3 - Conference contribution
AN - SCOPUS:84957899807
SN - 3540676899
SN - 9783540450498
VL - 1821
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 426
EP - 433
BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PB - Springer Verlag
ER -