liu.seSearch for publications in DiVA
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Learning Generalized Policies without Supervision Using GNNs
Linköping University, Department of Computer and Information Science, Artificial Intelligence and Integrated Computer Systems. Linköping University, Faculty of Science & Engineering. (Representation, Learning and Planning Lab)ORCID iD: 0000-0002-4092-8175
Universitat Pompeu Fabra, Spain.ORCID iD: 0000-0001-8212-7408
Linköping University, Department of Computer and Information Science, Artificial Intelligence and Integrated Computer Systems. Linköping University, Faculty of Science & Engineering. Universitat Pompeu Fabra, Spain; Instituci ́o Catalana de Recerca i Estudis Avanc ̧ats (ICREA), Barcelona, Spain. (Representation, Learning and Planning Lab)ORCID iD: 0000-0001-9851-8219
2022 (English)In: Proceedings of the 19th International Conference on Principles of Knowledge Representation and Reasoning: Special Session on KR and Machine Learning, IJACAI Organization , 2022, p. 474-483Conference paper, Published paper (Refereed)
Abstract [en]

We consider the problem of learning generalized policies forclassical planning domains using graph neural networks fromsmall instances represented in lifted STRIPS. The problemhas been considered before but the proposed neural architec-tures are complex and the results are often mixed. In thiswork, we use a simple and general GNN architecture andaim at obtaining crisp experimental results and a deeper un-derstanding: either the policy greedy in the learned valuefunction achieves close to 100% generalization over instanceslarger than those used in training, or the failure must be under-stood, and possibly fixed, logically. For this, we exploit therelation established between the expressive power of GNNsand the C2 fragment of first-order logic (namely, FOL with2 variables and counting quantifiers). We find for examplethat domains with general policies that require more expres-sive features can be solved with GNNs once the states are ex-tended with suitable ”derived atoms” encoding role composi-tions and transitive closures that do not fit into C2. The workfollows an existing approach based on GNNs for learning op-timal general policies in a supervised fashion, but the learnedpolicies are no longer required to be optimal (which expandsthe scope, as many planning domains do not have general op-timal policies) and are learned without supervision. Interest-ingly, value-based reinforcement learning methods that aimto produce optimal policies, do not always yield policies thatgeneralize, as the goals of optimality and generality are inconflict in domains where optimal planning is NP-hard.

Place, publisher, year, edition, pages
IJACAI Organization , 2022. p. 474-483
Series
Proceedings of the International Conference on Principles of Knowledge Representation and Reasoning (KR), ISSN 2334-1033
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:liu:diva-188843DOI: 10.24963/kr.2022/49Scopus ID: 2-s2.0-85141885575ISBN: 9781956792010 (print)OAI: oai:DiVA.org:liu-188843DiVA, id: diva2:1699395
Conference
19th International Conference on Principles of Knowledge Representation and Reasoning, Haifa, Israel, July 31 - August 5, 2022
Available from: 2022-09-27 Created: 2022-09-27 Last updated: 2024-08-22Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Ståhlberg, SimonGeffner, Hector

Search in DiVA

By author/editor
Ståhlberg, SimonBonet, BlaiGeffner, Hector
By organisation
Artificial Intelligence and Integrated Computer SystemsFaculty of Science & Engineering
Computer Sciences

Search outside of DiVA

GoogleGoogle Scholar

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 260 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf