Automation of Logic
Deutsch

Inductive Logic Programming: a Symbolic Approach to Machine Learning

Winter 2018/2019 block seminar: 7 ECTS
Open slots: 0/8
If you would like to attend, you can contact the lecturer to be put on a waiting list and come to the first meeting. If registered students do not attend or decide to quit, you can join instead of them.

Content

Through this seminar, you will discover the field of Inductive Logic Programming (ILP), where techniques are developed to make computers learn logic programs from examples and some background knowledge.

Topics

After an introductory meeting where the basics of ILP will be presented, several papers will be proposed to the participants to choose from. They cover the following topics:

  • foundations of ILP,
  • ILP and predicate invention,
  • using Answer Set Programming (ASP) for ILP,
  • learning efficient programs with ILP,
  • applications of ILP to planning.

Sample Paper

The following paper illustrates the kind of paper that will be proposed to the participants:

Legras, Swann, Céline Rouveirol, and Véronique Ventos.
"The Game of Bridge: A Challenge for ILP."
International Conference on Inductive Logic Programming. Springer, Cham, 2018.

A pdf version of this paper is available until the beginning of the seminar here.

List of Papers

  1. J. Ross Quinlan: Learning First-Order Definitions of Functions. CoRR cs.AI/9610102 (1996).

  2. Andrew Cropper, Stephen H. Muggleton: Learning Higher-Order Logic Programs through Abstraction and Invention. IJCAI 2016: 1418-1424.

  3. Cropper, Andrew, and Stephen H. Muggleton. "Learning efficient logic programs." Machine Learning (2018): 1-21.

  4. Aws Albarghouthi, Paraschos Koutris, Mayur Naik, Calvin Smith: Constraint-Based Synthesis of Datalog Programs. CP 2017: 689-706

  5. David Martínez, Guillem Alenyà, Carme Torras, Tony Ribeiro, Katsumi Inoue: Learning Relational Dynamics of Stochastic Domains for Planning. ICAPS 2016: 235-243

  6. Tobias Kaminski, Thomas Eiter, Katsumi Inoue: Exploiting Answer Set Programming with External Sources for Meta-Interpretive Learning. TPLP 18(3-4): 571-588 (2018)

  7. Céline Hocquette, Stephen Muggleton: How Much Can Experimental Cost Be Reduced in Active Learning of Agent Strategies? ILP 2018: 38-53

  8. Michael Siebers, Ute Schmid: Was the Year 2000 a Leap Year? Step-Wise Narrowing Theories with Metagol. ILP 2018: 141-156

  9. Johannes Rabold, Michael Siebers, Ute Schmid: Explaining Black-Box Classifiers with ILP - Empowering LIME with Aleph to Approximate Non-linear Decisions with Relational Rules. ILP 2018: 105-117

Organisation

lecturer: Sophie Tourret

Principle

During this seminar,  you will be asked to read a ILP paper, present it, attend the talks of the other participants and discuss with them on this topic. You will be offered two 30 minutes sessions with the lecturer during the preparatory phase: one over the technical details of your assigned paper and one over your presentation.

Prerequisite

Participants should have attended or be attending one of the following lectures (or an equivalent in another university):

  • Artificial Intelligence
  • Automated Reasoning I
  • Introduction to Computational Logic

Knowledge of propositional and first-order logic is expected.

Registration

This seminar is open to 8 students. To register, send an email to the lecturer stating which of the required lecture(s) you attended (or how you learnt about logic).

Grading

Grading will be based on the following criteria:

  • attendance to the two meetings (mandatory to get a passing grade)
  • understanding of the topic of the paper
  • quality of the talk
  • quality of the talk material
  • participation to the discussions
  • your critical feedback over your own work

Schedule

  • 03/09-31/10: registration
  • 31/10 14:00-16:00, building E1 5 (mpi-sws) room 630: introductory meeting
  • 05/11: final selection of papers
  • 22/01: seminar day
  • 30/01: deadline for sending your self-assessment to the lecturer