Learning Structured Declarative Rule Sets -- A Challenge for Deep Discrete Learning

Johannes Fürnkranz; Eyke Hüllermeier; Eneldo Loza Mencía; Michael Rapp

Learning Structured Declarative Rule Sets -- A Challenge for Deep Discrete Learning

Johannes Fürnkranz, Eyke Hüllermeier, Eneldo Loza Mencía, Michael Rapp

Institute for Application-oriented Knowledge Processing

Research output: Chapter in Book/Report/Conference proceeding › Conference proceedings

Abstract

Arguably the key reason for the success of deep neural networks is their ability to autonomously form non-linear combinations of the input features, which can be used in subsequent layers of the network. The analogon to this capability in inductive rule learning is to learn astructured rule base, where the inputs are combined to learn new aux-iliary concepts, which can then be used as inputs by subsequent rules. Yet, research on rule learning algorithms that have such capabilities is still in their infancy, which is — we would argue — one of the key impediments to substantial progress in this field. In this position paper, we want to draw attention to this unsolved problem, with a particular focuson previous work in predicate invention and multi-label rule learning

Original language	English
Title of host publication	Proceedings of the 2nd Workshop on Deep Continuous-Discrete Machine Learning (DeCoDeML)
Editors	Kristian Kersting and Stefan Kramer and Zahra Ahmadi
Number of pages	6
Publication status	Published - 2020

Fields of science

102019 Machine learning
102033 Data mining

JKU Focus areas

Digital Transformation

Cite this

@inproceedings{6d9b0a1199b446bf9d9809b656428a55,

title = "Learning Structured Declarative Rule Sets -- A Challenge for Deep Discrete Learning",

abstract = "Arguably the key reason for the success of deep neural networks is their ability to autonomously form non-linear combinations of the input features, which can be used in subsequent layers of the network. The analogon to this capability in inductive rule learning is to learn astructured rule base, where the inputs are combined to learn new aux-iliary concepts, which can then be used as inputs by subsequent rules. Yet, research on rule learning algorithms that have such capabilities is still in their infancy, which is — we would argue — one of the key impediments to substantial progress in this field. In this position paper, we want to draw attention to this unsolved problem, with a particular focuson previous work in predicate invention and multi-label rule learning",

author = "Johannes F{\"u}rnkranz and Eyke H{\"u}llermeier and {Loza Menc{\'i}a}, Eneldo and Michael Rapp",

year = "2020",

language = "English",

editor = "{Kristian Kersting and Stefan Kramer and Zahra Ahmadi}",

booktitle = "Proceedings of the 2nd Workshop on Deep Continuous-Discrete Machine Learning (DeCoDeML)",

}

Learning Structured Declarative Rule Sets -- A Challenge for Deep Discrete Learning. / Fürnkranz, Johannes; Hüllermeier, Eyke; Loza Mencía, Eneldo et al.
Proceedings of the 2nd Workshop on Deep Continuous-Discrete Machine Learning (DeCoDeML). ed. / Kristian Kersting and Stefan Kramer and Zahra Ahmadi. 2020.

Research output: Chapter in Book/Report/Conference proceeding › Conference proceedings

TY - GEN

T1 - Learning Structured Declarative Rule Sets -- A Challenge for Deep Discrete Learning

AU - Fürnkranz, Johannes

AU - Hüllermeier, Eyke

AU - Loza Mencía, Eneldo

AU - Rapp, Michael

PY - 2020

Y1 - 2020

N2 - Arguably the key reason for the success of deep neural networks is their ability to autonomously form non-linear combinations of the input features, which can be used in subsequent layers of the network. The analogon to this capability in inductive rule learning is to learn astructured rule base, where the inputs are combined to learn new aux-iliary concepts, which can then be used as inputs by subsequent rules. Yet, research on rule learning algorithms that have such capabilities is still in their infancy, which is — we would argue — one of the key impediments to substantial progress in this field. In this position paper, we want to draw attention to this unsolved problem, with a particular focuson previous work in predicate invention and multi-label rule learning

AB - Arguably the key reason for the success of deep neural networks is their ability to autonomously form non-linear combinations of the input features, which can be used in subsequent layers of the network. The analogon to this capability in inductive rule learning is to learn astructured rule base, where the inputs are combined to learn new aux-iliary concepts, which can then be used as inputs by subsequent rules. Yet, research on rule learning algorithms that have such capabilities is still in their infancy, which is — we would argue — one of the key impediments to substantial progress in this field. In this position paper, we want to draw attention to this unsolved problem, with a particular focuson previous work in predicate invention and multi-label rule learning

M3 - Conference proceedings

BT - Proceedings of the 2nd Workshop on Deep Continuous-Discrete Machine Learning (DeCoDeML)

A2 - Kristian Kersting and Stefan Kramer and Zahra Ahmadi, null

ER -