A Efficient Rule Extraction Method Using Hidden Unit Clarification in Trained Neural Network

인공 신경망에서 은닉 유닛 명확화를 이용한 효율적인 규칙추출 방법

  • 이헌주 (고려대학교 정보대학 컴퓨터학과) ;
  • 김현철 (고려대학교 정보대학 컴퓨터학과)
  • Received : 2018.01.11
  • Accepted : 2018.01.29
  • Published : 2018.01.31


Recently artificial neural networks have shown excellent performance in various fields. However, there is a problem that it is difficult for a person to understand what is the knowledge that artificial neural network trained. One of the methods to solve these problems is an algorithm for extracting rules from trained neural network. In this paper, we extracted rules from artificial neural networks using ordered-attribute search(OAS) algorithm, which is one of the methods of extracting rules, and analyzed result to improve extracted rules. As a result, we have found that the distribution of output values of the hidden layer unit affects the accuracy of rules extracted by using OAS algorithm, and it is suggested that efficient rules can be extracted by binarizing hidden layer output values using hidden unit clarification.

인공 신경망은 최근 다양한 분야에서 뛰어난 성능을 보여주고 있다. 하지만 인공 신경망이 학습한 지식이 정확히 어떤 내용인지를 사람이 파악하기 어렵다는 문제점이 존재하는데, 이를 해결하기 위한 방법 중 하나로 학습된 인공 신경망에서 규칙을 추출하는 방법들이 연구되고 있다. 본 연구에서는 학습된 인공 신경망으로부터 규칙을 추출하는 방법 중 하나인 ordered-attribute search(OAS) 알고리즘을 사용하여 인공 신경망으로부터 규칙을 추출해보고, 추출된 규칙을 개선하기 위해 규칙들을 분석하였다. 그 결과로 은닉 층의 출력값 분포가 OAS 알고리즘을 이용해 추출된 규칙의 정확도에 영향을 주는 것을 파악하였고, 은닉 유닛 명확화 기법을 통해 은닉 층 출력값을 이진화하여 효율적인 규칙을 추출할 수 있음을 제시하였다.



Supported by : 한국연구재단


  1. T. Hailesilassie. (2016). Rule extraction algorithm for deep neural networks: A review. International Journal of Computer Science and Information Security, 14, 7, p. 376.
  2. Andrews, R. Diederich, J. & Tickle, A. B. (1995). Survey and critique of techniques for extracting rules from trained artificial neural networks. Knowledge-Based System, 8(6), 373-389. https://doi.org/10.1016/0950-7051(96)81920-4
  3. Kim H. (2000). Computationally Efficient Heuristics for If-Then Rule Extraction from Feed-Forward Neural Networks. Lecture Notes in Computer Science, vol 1967, Springer, Berlin, Heidelberg.
  4. Ishikawa, M. (1996). Structural Learning with Forgetting. Neural Networks, vol.9, no.3, 509-521. https://doi.org/10.1016/0893-6080(96)83696-3
  5. L. M. Fu. (1994). Rule Generation from Neural Networks. Systems, Man and Cybernetics, 24(8), 1114-1124. https://doi.org/10.1109/21.299696
  6. L. M. Fu. (1991). Rule Learning by Searching on Adapted Nets. AAAI, 590-595.
  7. L. M. Fu. (1993). Knowledge-based connectionism for revising domain theories. IEEE Transactions on Systems, Man, and Cybernetics 23(1), 173-182. https://doi.org/10.1109/21.214775
  8. S. Thrun. (1995). Extracting rules from artificial neural networks with distributed representations. Advances in neural information processing systems, 505-512.
  9. I. A. a. J. G. Taha. (1999). Symbolic interpretation of artificial neural network. IEEE Transactions on knowledge and data engineering, 11(3), 448-463. https://doi.org/10.1109/69.774103
  10. G. P. C. A. a. F. S. G. Schmitz. (1999). ANN-DT: an algorithm for extraction of decision trees from artificial neural networks. IEEE Transactions on Neural Networks, 10(6), 1392-1401. https://doi.org/10.1109/72.809084
  11. G. G. a. j. W. S. Towell. (1993) Extracting refined rules from knowledge-based neural networks. Machine learning, 13(1), 71-101. https://doi.org/10.1007/BF00993103
  12. R. Setiono, W. K. Leow (2000). FERNN: An algorithm for fast extraction. Applied Intelligence, 12(1-2), 15-25. https://doi.org/10.1023/A:1008307919726