Abstract
In this position paper, we discuss the reasons for the lack of success of rule learning, as witnessed by the quasi-absence of commercial applications, and what can be done to revive the domain and possibly to kick-start the kind of explosive development that statistical Machine Learning and Neural Networks have experienced over the past 15 years. The root cause of the problem is well-known, and it is not the rule learning algorithms themselves: if the representation language to which a rule learning algorithm has access – often only the representation model of the input data – is not appropriate to represent the decision rules, the algorithm has no way to generate a decision ruleset that generalizes well. Feature generation and other methods that have been proposed to augment the data representation language are useful, but we argue that the focus should be on discovering the conceptual model that underly the decision. We claim that this amounts to discovering the structure of decisions, that is, to learn decision models. We outline some potentially fruitful research directions, and how this topic is central to neuro-symbolic learning.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Typically, additional data such as the applicant home address, household etc. will be available, some of it completely unrelated to the decision at hand and making the learning problem more difficult. However, this simple example will be enough for our purpose.
- 2.
We used Scikit Learn DecisionTreeClassifier, which is an optimised version of the CART algorithm [7] and Scikit Learn MLPClassifier for the further experiment, reported below. In both cases, the performance is computed using the classifier “score” method on a test set.
- 3.
Each leave in a decision tree defines a rule, where the condition is the conjunction of the tests along the path and the leave specifies the corresponding decision.
- 4.
Other configurations produce similar results. See: https://github.com/cfmrsma/RuleML21.
- 5.
More precisely: the part of the input space that is relevant with respect to the decision at hand.
- 6.
That is, until measurements break the theory, and the theory must be changed, of course….
References
Haohan, W., Bhiksha R.: On the origin of deep learning. arXiv:1702.07800 (2017)
Hinton, G.E., et al.: A fast learning algorithm for deep belief nets. Neural Comput. 18, 1527–1554 (2006)
Furnkranz, J.: Separate-and-conquer rule learning. Artif. Intell. Rev. 13(1), 3–54 (1999)
Quinlan, J.R.: C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers (1993)
Mitchell, T.: Generalization as Search. Artif. Intell. 18, 203–226 (1982)
Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.J.: Classification and Regression Trees. 1st edn. Routledge (1984). https://doi.org/10.1201/9781315139470
Dash, S., Günlük, O., Wei, D.: Boolean decision rules via column generation. In: 32nd Conference on Neural Information Processing Systems (NeurIPS 2018) (2018)
Miller, T.: Explanation in artificial intelligence: insights from the social sciences. Artif. Intell. 267, 1–38 (2019)
Bengio, Y., et al.: Representation learning: a review and new perspectives. IEEE Trans. Pattern Anal. Mach. Intell. 35, 1798–1828 (2013)
Scholkopf, B., et al.: Toward causal representation learning. Proc. IEEE 109, 612–634 (2021)
Kramer, S.: A brief history of learning symbolic higher-level representations from data (and a curious look forward). In: IJCAI (2020)
Sondhi, P.: Feature construction methods: a survey (2009)
Garnelo, M., et al.: Towards deep symbolic reinforcement learning. ArXiv abs/1609.05518 (2016)
Ganter, B., Wille, R.: Formal Concept Analysis: Mathematical Foundations. Springer, Berlin (1999). https://doi.org/10.1007/978-3-642-59830-2
Fürnkranz, J., et al.: Learning structured declarative rule sets - a challenge for deep discrete learning. ArXiv abs/2012.04377 (2020)
Shindo, H., et al.: Differentiable inductive logic programming for structured examples. In: AAAI (2021)
Fish, A.: Melding process models and decision models. Modeling decision-making processes. https://dmcommunity.files.wordpress.com/2016/06/decisioncamp2016-alanfish.pdf
OMG, Decision Model and Notation. https://www.omg.org/spec/DMN
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 Springer Nature Switzerland AG
About this paper
Cite this paper
de Sainte Marie, C. (2021). Learning Decision Rules or Learning Decision Models?. In: Moschoyiannis, S., Peñaloza, R., Vanthienen, J., Soylu, A., Roman, D. (eds) Rules and Reasoning. RuleML+RR 2021. Lecture Notes in Computer Science(), vol 12851. Springer, Cham. https://doi.org/10.1007/978-3-030-91167-6_19
Download citation
DOI: https://doi.org/10.1007/978-3-030-91167-6_19
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-91166-9
Online ISBN: 978-3-030-91167-6
eBook Packages: Computer ScienceComputer Science (R0)