Improving Generalization Performance in Co-Evolutionary Learning
Recently, the generalization framework in co-evolutionary learning has been theoretically formulated and demonstrated in the context of game-playing. Generalization performance of a strategy (solution) is estimated using a collection of random test strategies (test cases) by taking the average game...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
2012
|
| Subjects: | |
| Online Access: | http://shdl.mmu.edu.my/3386/ http://shdl.mmu.edu.my/3386/1/10.pdf |
| _version_ | 1848790313722707968 |
|---|---|
| author | Chong, Siang Yew Tino, Peter Ku, Day Chyi Yao, Xin |
| author_facet | Chong, Siang Yew Tino, Peter Ku, Day Chyi Yao, Xin |
| author_sort | Chong, Siang Yew |
| building | MMU Institutional Repository |
| collection | Online Access |
| description | Recently, the generalization framework in co-evolutionary learning has been theoretically formulated and demonstrated in the context of game-playing. Generalization performance of a strategy (solution) is estimated using a collection of random test strategies (test cases) by taking the average game outcomes, with confidence bounds provided by Chebyshev's theorem. Chebyshev's bounds have the advantage that they hold for any distribution of game outcomes. However, such a distribution-free framework leads to unnecessarily loose confidence bounds. In this paper, we have taken advantage of the near-Gaussian nature of average game outcomes and provided tighter bounds based on parametric testing. This enables us to use small samples of test strategies to guide and improve the co-evolutionary search. We demonstrate our approach in a series of empirical studies involving the iterated prisoner's dilemma (IPD) and the more complex Othello game in a competitive co-evolutionary learning setting. The new approach is shown to improve on the classical co-evolutionary learning in that we obtain increasingly higher generalization performance using relatively small samples of test strategies. This is achieved without large performance fluctuations typical of the classical approach. The new approach also leads to faster co-evolutionary search where we can strictly control the condition (sample sizes) under which the speedup is achieved (not at the cost of weakening precision in the estimates). |
| first_indexed | 2025-11-14T18:10:38Z |
| format | Article |
| id | mmu-3386 |
| institution | Multimedia University |
| institution_category | Local University |
| language | English |
| last_indexed | 2025-11-14T18:10:38Z |
| publishDate | 2012 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | mmu-33862014-03-05T03:48:05Z http://shdl.mmu.edu.my/3386/ Improving Generalization Performance in Co-Evolutionary Learning Chong, Siang Yew Tino, Peter Ku, Day Chyi Yao, Xin QA75.5-76.95 Electronic computers. Computer science Recently, the generalization framework in co-evolutionary learning has been theoretically formulated and demonstrated in the context of game-playing. Generalization performance of a strategy (solution) is estimated using a collection of random test strategies (test cases) by taking the average game outcomes, with confidence bounds provided by Chebyshev's theorem. Chebyshev's bounds have the advantage that they hold for any distribution of game outcomes. However, such a distribution-free framework leads to unnecessarily loose confidence bounds. In this paper, we have taken advantage of the near-Gaussian nature of average game outcomes and provided tighter bounds based on parametric testing. This enables us to use small samples of test strategies to guide and improve the co-evolutionary search. We demonstrate our approach in a series of empirical studies involving the iterated prisoner's dilemma (IPD) and the more complex Othello game in a competitive co-evolutionary learning setting. The new approach is shown to improve on the classical co-evolutionary learning in that we obtain increasingly higher generalization performance using relatively small samples of test strategies. This is achieved without large performance fluctuations typical of the classical approach. The new approach also leads to faster co-evolutionary search where we can strictly control the condition (sample sizes) under which the speedup is achieved (not at the cost of weakening precision in the estimates). 2012-02 Article PeerReviewed text en http://shdl.mmu.edu.my/3386/1/10.pdf Chong, Siang Yew and Tino, Peter and Ku, Day Chyi and Yao, Xin (2012) Improving Generalization Performance in Co-Evolutionary Learning. IEEE Transactions on Evolutionary Computation, 16 (1). pp. 70-85. ISSN 1089-778X http://dx.doi.org/10.1109/TEVC.2010.2051673 doi:10.1109/TEVC.2010.2051673 doi:10.1109/TEVC.2010.2051673 |
| spellingShingle | QA75.5-76.95 Electronic computers. Computer science Chong, Siang Yew Tino, Peter Ku, Day Chyi Yao, Xin Improving Generalization Performance in Co-Evolutionary Learning |
| title | Improving Generalization Performance in Co-Evolutionary Learning |
| title_full | Improving Generalization Performance in Co-Evolutionary Learning |
| title_fullStr | Improving Generalization Performance in Co-Evolutionary Learning |
| title_full_unstemmed | Improving Generalization Performance in Co-Evolutionary Learning |
| title_short | Improving Generalization Performance in Co-Evolutionary Learning |
| title_sort | improving generalization performance in co-evolutionary learning |
| topic | QA75.5-76.95 Electronic computers. Computer science |
| url | http://shdl.mmu.edu.my/3386/ http://shdl.mmu.edu.my/3386/ http://shdl.mmu.edu.my/3386/ http://shdl.mmu.edu.my/3386/1/10.pdf |