DOI QR코드

DOI QR Code

Extracting Rules from Neural Networks with Continuous Attributes

연속형 속성을 갖는 인공 신경망의 규칙 추출

  • Received : 2017.05.10
  • Accepted : 2017.10.25
  • Published : 2018.01.15

Abstract

Over the decades, neural networks have been successfully used in numerous applications from speech recognition to image classification. However, these neural networks cannot explain their results and one needs to know how and why a specific conclusion was drawn. Most studies focus on extracting binary rules from neural networks, which is often impractical to do, since data sets used for machine learning applications contain continuous values. To fill the gap, this paper presents an algorithm to extract logic rules from a trained neural network for data with continuous attributes. It uses hyperplane-based linear classifiers to extract rules with numeric values from trained weights between input and hidden layers and then combines these classifiers with binary rules learned from hidden and output layers to form non-linear classification rules. Experiments with different datasets show that the proposed approach can accurately extract logical rules for data with nonlinear continuous attributes.

지난 수십 년 동안 인공 신경망은 음성 인식에서 이미지 분류에 이르기까지 수많은 분야에서 성공적으로 사용되었다. 그러나 인공 신경망은 특정 결론이 어떻게 도출되었는지 알 필요가 있음에도 불구하고 이러한 결과를 설명할 수 있는 능력이 부족하다. 대부분의 연구는 신경망에서 이진 규칙을 추출하는데 초점을 맞추고 있지만, 기계 학습 응용 프로그램에 사용되는 데이터는 연속된 값이 포함되어 있기 때문에 실용적이지 않은 경우가 있다. 이러한 격차를 줄이기 위해 본 논문에서는 연속된 값이 포함된 데이터로부터 학습된 신경망에서 논리 규칙을 추출하는 알고리즘을 제안한다. 초평면 기반 선형 분류기를 사용하여 입력 및 은닉 층 사이에서 학습된 가중치로부터 규칙을 추출하고, 비선형 분류 규칙을 생성하기 위해 은닉 층과 출력 층에서 학습된 이진 규칙과 분류기를 결합한다. 비선형 연속값으로 구성된 여러 데이터셋을 대상으로 진행한 실험에서 제안하는 방법이 논리적 규칙을 정확하게 추출할 수 있음을 보였다.

Keywords

Acknowledgement

Supported by : National Research Foundation of Korea (NRF)

References

  1. Artur d'Avila Garcez, Tarek R. Besold, Luc de Raedt, Peter Foldiak, Pascal Hitzler, Thomas Icard, Kai-Uwe Kuhnberger, Luis C. Lamb, Risto Miikkulainen, Daniel L. Silver, "Neural-Symbolic Learning and Reasoning: Contributions and Challenges," AAAI, 2015.
  2. Geoffrey G. Towell, Jude W. Shavlik, "The Extraction of Refined Rules from Knowledge-Based Neural Networks," Machine Learning Resource Group Working Paper, 1991.
  3. Geoffrey G. Towell, Jude W. Shavlik, "Knowledge-Based Artificial Neural Networks," Artificial Intelligence, 1994.
  4. G. Fung, S. Sandilya, and R.B. Rao, "Rule extraction from linear support vector machines," Proc. of the 11th ACM SIGKDD international Conference on Knowledge Discovery in Data Mining, pp. 32-40, 2005.
  5. A.S. d'Avila Garcez, K. Broda, D.M. Gabbay. "Symbolic knowledge extraction from trained neural networks: A sound approach," Artificial Intelligence, 2001.
  6. N. Barakat, J. Diederich, Eclectic "Rule-extraction from support vector machines," International Journal of Computational Intelligence (2005), 59-62.
  7. Tran, S.N. d'Avila Garcez, A, "Knowledge extraction from deep belief networks for images," IJCAI-2013 Workshop on Neural-Symbolic Learning and Reasoning (2013).
  8. J. A. Taha and J. Ghosh, "Symbolic interpretation of artificial neural networks," IEEE Trans, Data Eng, Vol. 168, pp. 1009-1018, 2006.
  9. U. Markowska-Kaczmar and M.Chumieja, "Discovering the mysteries of neural networks," International Journal of Hybrid Intelligent Systems, Vol. 3-4, pp. 153-163, 2004.
  10. L. Breiman, J. Friedman, R. Olshen, and C. Stone, "Classification and regression trees," Wadsworth and Brooks, Monterey, CA, 1994.
  11. J. R. Quinlan, "C4.5 programs for machine learning," Morgan Kaufmann, 1993.
  12. R. Setiono, B. Baesens, and C. Mues, "Recursive Neural Network Rule Extraction for Data With Mixed Attributes," IEEE Trans, Neural Network, Vol. 19, 2008.
  13. Haixiang Guo, Yanan Li, Xiaio Liu, Yijing Li, "A New Neural Data Analysis Approach Using Ensemble Neural Network Rule Extraction," Artificial Neural Networks and Machine Learning, ICANN 2012, pp. 515-522, 2012.
  14. M. Abadi, "TensorFlow: Large-scale machine learning on heterogeneous systems," 2015. Software available from tensorflow.org.