Inductive Logic Programming: Theory and Methods

By Stephen H. Muggleton, Luc De Raedt. 1994

Abstract Inductive Logic Programming (ILP) is a new discipline which investigates the inductive construction of first-order clausal theories from examples and background knowledge. We survey the most important theories and methods of this new field. First, various problem specifications of ILP are formalized in semantic settings for ILP, yielding a “model-theory” for ILP. Second, a generic ILP algorithm is presented. Third, the inference rules and corresponding operators used in ILP are presented, resulting in a “proof-theory” for ILP. Fourth, since inductive inference does not produce statements which are assured to follow from what is given, inductive inferences require an alternative form of justification. This can take the form of either probabilistic support or logical constraints on the hypothesis language. Information compression techniques used within ILP are presented within a unifying Bayesian approach to confirmation and corroboration of hypotheses. Also, different ways to constrain the hypothesis language or specify the declarative bias are presented. Fifth, some advanced topics in ILP are addressed. These include aspects of computational learning theory as applied to ILP, and the issue of predicate invention. Finally, we survey some applications and implementations of ILP. ILP applications fall under two different categories: first, scientific discovery and knowledge acquisition, and second, programming assistants.

Subfields: Knowledge Graphs & Representation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Knowledge Graphs & Representation

Citations in the corpus

Integrating Learning and Reasoning with Deep Logic Models (2019, background): “On the other hand, symbolic reasoning [7, 16, 23], which is typically based on logical and probabilistic inference, allows to perform high-level reasoning (possibly under uncertainty) without having to deal with thousands of learning hyper-parameters.”
Interpretable classifiers using rules and Bayesian analysis: Building a better stroke prediction model (2015, methodology): “Consider the challenge of constructing a predictive model that discretizes the input space in the same way as decision trees [Breiman et al. (1984), Quinlan (1993)], decision lists [Rivest (1987)] or associative classifiers [Liu, Hsu and Ma (1998)].”
Learning Certifiably Optimal Rule Lists (2017, background): “For realistic cases, it is possible to solve fairly large cases of this problem to optimality, with the careful use of algorithms, data structures, and implementation techniques.”
Learning First-Order Symbolic Representations for Planning from the Structure of the State Space (2019, methodology): “Inductive logic programming methods [31] have been used for learning general policies but from symbolic encodings too [23, 28, 14].”
Learning customized and optimized lists of rules with mathematical programming (2018, background): “, [47]) that locate logical rules but do not optimize how they are going to be used together in a predictive model, or other types of techniques that learn models for individual rules but do not learn how they should be combined (e.”
Relational Deep Reinforcement Learning (2018, background): “The core idea behind RRL is to combine reinforcement learning with relational learning or Inductive Logic Programming [16] by representing states, actions and policies using a first order (or relational) language [8, 9, 17, 18].”
An optimization approach to learning falling rule lists (2017, methodology): “It relates closely to associative classiﬁcation (e.g. the RIPPER k algorithm [7] and the CBA algorithm [17]; see [25] for a comprehensive review) and inductive logic programming [20].”
Scalable Bayesian Rule Lists (2016, neutral): “Inductive logic programming (Muggleton & De Raedt, 1994), greedy top-down decision list algorithms (Rivest, 1987; Sokolova, Marchand, Japkowicz, & Shawe-Taylor, 2003; Anthony, 2005; Marchand & Sokolova, 2005; Rudin, Letham, & Madigan, 2013; Goessling & Kang, 2015), associative classiﬁcation…”
Learning Optimized Or's of And's (2015, methodology): “We build OA models as a diagnostic screening tool for obstructive sleep apnea, that achieves high accuracy with a substantial gain in interpretability over other methods.”
Or's of And's for Interpretable Classification, with Application to Context-Aware Recommender Systems (2015, background): “Inductive logic programming [33] is similar, in that it mines (potentially complicated) patterns and takes the simple union of these patterns as the pattern set, rather than optimizing the pattern set directly.”
NLProlog: Reasoning with Weak Unification for Question Answering in Natural Language (2019, methodology): “Speciﬁcally, we train NLP ROLOG with backpropagation using a learning from entailment setting (Muggleton and Raedt, 1994), in which the model is trained to decide whether a Prolog program R entails the truth of a candidate triple c ∈ C , where C is the set of candidate triples.”
Neural Program Synthesis with Priority Queue Training (2018, methodology): “Our method shares the same goal with traditional techniques in program synthesis and inductive programming (Summers, 1977; Biermann, 1978; Muggleton & de Raedt, 1994; Angulin, 1987).”
Frequent substructure-based approaches for classifying chemical compounds (2003, background): “They applied an inductive logic programming (ILP) system [62], Golem [63], to study the behavior of 44 trimethoprin analogues and their observed inhibition of Escherichia coli dihydrofolate reductase and reported a considerable improvement in classification accuracy over the traditional QSAR-based m”
Learning Action Strategies for Planning Domains (1999, methodology): “Our work is also related to Reinforcement Learning [19] and to Inductive Logic Programming [30].”
Binary Rule Generation via Hamming Clustering (2002, methodology): “To this class of techniques belong methods developed in the area of Inductive Logic Programming (ILP) [22], [23], [24], [25]; they aim at constructing a logic program associated with a set of rules expressed in first-order logic.”
Learning Relational Navigation Policies (2006, background): “This can be realized using the learning from interpretations settings well studied in the field of inductive logic programming [16] where relational programs are induced from interpretation-class pairs.”