Machine Learning: Between Accuracy and Interpretability
Predictive accuracy is the usual measure of success of Machine Learning (ML) applications. However, experience from many ML applications in difficult, domains indicates the importance of interpretability of induced descriptions. Often in such domains, predictive accuracy is hardly of interest to the user. Instead, the users’ interest now lies in the interpretion of the induced descriptions and not, in their use for prediction. In such cases, ML is essentially used as a tool for exploring the domain, to generate new, potentially useful ideas about the domain, and thus improve the user’s understanding of the domain. The important questions are how to make domain-specific background knowledge usable by the learning system, and how to interpret the results in the light of this background expertise. These questions are discussed and illustrated by relevant example applications of ML, including: medical diagnosis, ecological modelling, and interpreting discrete event simulations. The observations in these applications show that predictive accuracy, the usual measure of success in ML, should be accompanied by a. criterion of interpretability of induced descriptions. The formalisation of interpretability is however a completely new challenge for ML.
KeywordsMachine Learn Predictive Accuracy Regression Tree Ecological Modelling Discrete Event Simulation
Unable to display preview. Download preview PDF.
- Bratko, I., Mozetic, I., Lavrac, N. (1989) KARDIO: A Study in Deep and Qualitative Knowledge for Expert Systems,MIT Press.Google Scholar
- Dzeroski, S., Grbovié, J., Walley, W. (1997) Machine learning applications in biological classification of river water quality. In: Machine Learning and Data Mining: Methods and Applications (eds. R.S. Michalski, I. Bratko, M. Kubat) Wiley (in press).Google Scholar
- Grobelnik, M. (1992) Markus: an optimized model inference system. Logic Approaches to Machine Learning Workshop, Vienna, August 1992.Google Scholar
- Karalie., A. (1992) Employing Linear Regression in Regression Tree Leaves. Proc. 10th European Conference on Artificial Intelligence,pp. 440–441, Vienna, 1992.Google Scholar
- Kompare, B. (1995) Artificial Intelligence Techniques in Ecological Modelling. Ph.D. Thesis, Kopenhagen: Royal School of Pharmacy.Google Scholar
- Kononenko, I., Bratko, I. (1991) Information based evaluation criterion for classifier’s performance. Machine Learning Journal, Vol. 6, pp. 283–312.Google Scholar
- Krizman, V. (1994) Handling noisy data in automated modelling of dynamical systems. M.Sc. Thesis, Faculty of Computer and Info. Sc., University of Ljubljana (in Slovenian).Google Scholar
- Mladenié, D., Bratko, I., Paul, R.J., Grobelnik, M., Using machine learning techniques to interpret results from discrete event simulation. Proc. ECML-94 Conf., Catania, Italy, april 1994.Google Scholar
- Muggleton, S. (1995) Inverse entailment and Progol. New Generation Computing, Vol. 13, Nos. 3, 4 (Special issue on Inductive Logic Programming), pp. 245–286.Google Scholar
- Pirnat, V., Kononenko, I., Janc, T., Bratko, I. (1989), Medical analysis of automatically induced diagnostic rules. Third Int. Conf. AI in Medicine, London.Google Scholar
- Srinivasan, A., Muggleton, S.H., King, R.D., Sternberg, M.J.E. (1994) Mutagene-sis: ILP experiments in a non-determinate biological domain. In Proc. Fourth Int. Workshop on Inductive Logic Programming ILP-94,Bad Honnef/Bonn.Google Scholar
- Stankovski, V., Debeljak, M., Bratko, I., Adamic, M. (1996) Modelling the population dynamics of red deer (cervus elaphus L.) with regard to forest development. Ecological Modelling Conf., Copenhagen; to appear in Ecological Modelling Journal.Google Scholar