Multi-layer relation networks for relational reasoning

Marius Jahrens; Thomas Martinetz

doi:10.1145/3309772.3309782

Multi-layer relation networks for relational reasoning

Marius Jahrens, Thomas Martinetz

Institute of Neuro- and Bioinformatics

3 Citations (Scopus)

Abstract

Relational Networks (RN) as introduced by Santoro et al. in 2017 have demonstrated strong relational reasoning capabilities with a rather shallow architecture. Its single-layer design, however, only considers pairs of information objects, making it unsuitable for problems requiring reasoning across a higher number of facts. To overcome this limitation, we propose a multi-layer relation network architecture which enables successive refinements of relational information through multiple layers. We show that the increased depth allows for more complex relational reasoning by applying it to the bAbI 20 QA dataset, solving all 20 tasks with joint training and surpassing the state-of-the-art results.

Original language	English
Title of host publication	APPIS '19: Proceedings of the 2nd International Conference on Applications of Intelligent Systems
Editors	Nicolai Petkov, Nicola Strisciuglio, Carlos M. Travieso
Number of pages	5
Publisher	ACM
Publication date	07.01.2019
Pages	1–5
Article number	10
ISBN (Print)	978-1-4503-6085-2
DOIs	https://doi.org/10.1145/3309772.3309782
Publication status	Published - 07.01.2019
Event	2nd International Conference on Applications of Intelligent Systems - Museo Elder of Science and Technology, Las Palmas de Gran Canaria, Spain Duration: 07.01.2019 → 09.01.2019 Conference number: 149535

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure

Research Areas and Centers

Research Area: Intelligent Systems
Centers: Center for Artificial Intelligence Luebeck (ZKIL)

DFG Research Classification Scheme

4.43-05 Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing

Access to Document

10.1145/3309772.3309782

Cite this

@inproceedings{7dafd4d22f35420a986014b21bc13555,

title = "Multi-layer relation networks for relational reasoning",

abstract = "Relational Networks (RN) as introduced by Santoro et al. in 2017 have demonstrated strong relational reasoning capabilities with a rather shallow architecture. Its single-layer design, however, only considers pairs of information objects, making it unsuitable for problems requiring reasoning across a higher number of facts. To overcome this limitation, we propose a multi-layer relation network architecture which enables successive refinements of relational information through multiple layers. We show that the increased depth allows for more complex relational reasoning by applying it to the bAbI 20 QA dataset, solving all 20 tasks with joint training and surpassing the state-of-the-art results.",

author = "Marius Jahrens and Thomas Martinetz",

note = "Publisher Copyright: {\textcopyright} 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM. Copyright: Copyright 2019 Elsevier B.V., All rights reserved.; 2nd International Conference on Applications of Intelligent Systems, APPIS 2019 ; Conference date: 07-01-2019 Through 09-01-2019",

year = "2019",

month = jan,

day = "7",

doi = "10.1145/3309772.3309782",

language = "English",

isbn = "978-1-4503-6085-2",

series = "ACM International Conference Proceeding Series",

publisher = "ACM",

pages = "1–5",

editor = "Petkov, \{Nicolai \} and Strisciuglio, \{Nicola \} and \{M. Travieso\}, \{Carlos \}",

booktitle = "APPIS '19: Proceedings of the 2nd International Conference on Applications of Intelligent Systems",

}

Jahrens, M & Martinetz, T 2019, Multi-layer relation networks for relational reasoning. in N Petkov, N Strisciuglio & C M. Travieso (eds), APPIS '19: Proceedings of the 2nd International Conference on Applications of Intelligent Systems., 10, ACM International Conference Proceeding Series, ACM, pp. 1–5, 2nd International Conference on Applications of Intelligent Systems, Las Palmas de Gran Canaria, Spain, 07.01.19. https://doi.org/10.1145/3309772.3309782

Multi-layer relation networks for relational reasoning. / Jahrens, Marius; Martinetz, Thomas.
APPIS '19: Proceedings of the 2nd International Conference on Applications of Intelligent Systems. ed. / Nicolai Petkov; Nicola Strisciuglio; Carlos M. Travieso. ACM, 2019. p. 1–5 10 (ACM International Conference Proceeding Series).

Research output: Chapters in Books/Reports/Conference Proceedings › Conference contribution › peer-review

TY - GEN

T1 - Multi-layer relation networks for relational reasoning

AU - Jahrens, Marius

AU - Martinetz, Thomas

N1 - Conference code: 149535

PY - 2019/1/7

Y1 - 2019/1/7

N2 - Relational Networks (RN) as introduced by Santoro et al. in 2017 have demonstrated strong relational reasoning capabilities with a rather shallow architecture. Its single-layer design, however, only considers pairs of information objects, making it unsuitable for problems requiring reasoning across a higher number of facts. To overcome this limitation, we propose a multi-layer relation network architecture which enables successive refinements of relational information through multiple layers. We show that the increased depth allows for more complex relational reasoning by applying it to the bAbI 20 QA dataset, solving all 20 tasks with joint training and surpassing the state-of-the-art results.

AB - Relational Networks (RN) as introduced by Santoro et al. in 2017 have demonstrated strong relational reasoning capabilities with a rather shallow architecture. Its single-layer design, however, only considers pairs of information objects, making it unsuitable for problems requiring reasoning across a higher number of facts. To overcome this limitation, we propose a multi-layer relation network architecture which enables successive refinements of relational information through multiple layers. We show that the increased depth allows for more complex relational reasoning by applying it to the bAbI 20 QA dataset, solving all 20 tasks with joint training and surpassing the state-of-the-art results.

UR - https://www.scopus.com/pages/publications/85070537760

U2 - 10.1145/3309772.3309782

DO - 10.1145/3309772.3309782

M3 - Conference contribution

SN - 978-1-4503-6085-2

T3 - ACM International Conference Proceeding Series

SP - 1

EP - 5

BT - APPIS '19: Proceedings of the 2nd International Conference on Applications of Intelligent Systems

A2 - Petkov, Nicolai

A2 - Strisciuglio, Nicola

A2 - M. Travieso, Carlos

PB - ACM

T2 - 2nd International Conference on Applications of Intelligent Systems

Y2 - 7 January 2019 through 9 January 2019

ER -

Multi-layer relation networks for relational reasoning

Abstract

UN SDGs

Research Areas and Centers

DFG Research Classification Scheme

Access to Document

Other files and links

Fingerprint

Cite this