DOI QR코드

DOI QR Code

Search Space Analysis of R-CORE Method for Bayesian Network Structure Learning and Its Effectiveness on Structural Quality

R-CORE를 통한 베이지안 망 구조 학습의 탐색 공간 분석

  • 정성원 (한국과학기술원 바이오및뇌공학과) ;
  • 이도헌 (한국과학기술원 바이오및뇌공학과) ;
  • 이광형 (한국과학기술원 바이오및뇌공학과, AITrc)
  • Published : 2008.08.25

Abstract

We analyze the search space considered by the previously proposed R-CORE method for learning Bayesian network structures of large scale. Experimental analysis on the search space of the method is also shown. The R-CORE method reduces the search space considered for Bayesian network structures by recursively clustering the random variables and restricting the orders between clusters. We show the R-CORE method has a similar search space with the previous method in the worst case but has a much less search space in the average case. By considering much less search space in the average case, the R-CORE method shows less tendency of overfitting in learning Bayesian network structures compared to the previous method.

본 논문에서는 대규모 베이지안 망 구조 학습을 위해 제안되었던 R-CORE 방법의 탐색 공간의 크기에 대한 개략적인 분석과 실제 문제에 적용하였을 경우의 효과에 대한 실험적 결과를 제시한다. R-CORE 방법은 베이지안 망 구조 학습의 탐색 공간을 축소하기 위해 제안된 확률변수들의 재귀적 군집화와 오더 제한 방법이다. 알려진 벤치마크 베이지안 망을 이용한 분석을 통해, 제안되었던 R-CORE 방법이 worst case에는 기존의 방법과 유사한 탐색 공간을 가지나 평균적으로 기존방법보다 훨씬 적은 탐색 공간만을 고려한다는 것을 보인다. 또한 평균적으로 훨씬 적은 탐색 공간만을 고려하는 결과, 구조 탐색에서 기존 방법에 비해 상대적으로 적은 overfitting이 일어남을 실험적으로 보인다.

Keywords

References

  1. R. E. Neapolitan, Learning Bayesian Networks, Pearson Prentice Hall, New Jersey, 2004
  2. D. Heckerman, D. Gerger and D. M. Chickering, "Learning Bayesian Networks: The Combination of Knowledge and Statistical Data," Machine Learning, Vol. 20, pp. 197-243, 1995
  3. P. Grunwald, "A Tutorial Introduction to the Minimum Description Length Principle," Advances in Minimum Description Length: Theory and Applications, MIT Press, 2004
  4. J. Suzuki, "Learning Bayesian Belief Networks Based on the Minimum Description Length Principle: Basic Properties," IEICE Transactions on Fundamentals, Vol. E82-A, No. 9, pp. 1-9, 1999
  5. R. W. Robinson, "Counting Labeled Acyclic Digraphs," New Directions in the Theory of Graphs, pp. 239-273, Academic Press, New York, 1973
  6. R. Etxeberria, P. Larranaga and J. M. Picaza, "Analysis of the Behaviour of Genetic Algorithms when Learning Bayesian Network Structure from Data," Pattern Recognition Letters, Vol. 18, pp. 1269-1273, 1997 https://doi.org/10.1016/S0167-8655(97)00106-2
  7. N. Friedman, I. Nachman and D. Peer, "Learning Bayesian Network Structure from Massive Datasets: The "Sparse Candidate" Algorithm," Proceedings of the Fifteenth Conference on Unceratainty in Artificial Intelligence, pp. 206-215, 1999
  8. K. Hwang, J. Lee, S. Chung and B. Zhang, "Construction of Large-Scale Bayesian Networks by Local to Global Search," PRICAI 2002, pp. 375-384, 2002
  9. L. E. Brown, I. Tsamardinos and C. F. Aliferis, "A Novel Algorithm for Scalable and Accurate Bayesian Network Learning," MEDINFO, 2004
  10. S. Jung, K. Lee and D. Lee, "Enabling Large-Scale Bayesian Network Learning by Preserving Intercluster Directionality," IEICE Transactions on Information and Systems, Vol. E90-D, No. 7, 2007, to appear