Theory reconstruction: a representation learning view on predicate invention

TL;DR

This paper introduces a theory reconstruction approach that extends autoencoder-based representation learning to relational settings, aiming to unify predicate invention and theory revision in a formal framework.

Contribution

It proposes a novel formalism that adapts autoencoder techniques for predicate invention within relational learning, fostering interdisciplinary dialogue.

Findings

Extends autoencoder models to relational predicate invention

Provides a formal framework for theory reconstruction

Aims to unify relational and deep learning approaches

Abstract

With this positional paper we present a representation learning view on predicate invention. The intention of this proposal is to bridge the relational and deep learning communities on the problem of predicate invention. We propose a theory reconstruction approach, a formalism that extends autoencoder approach to representation learning to the relational settings. Our intention is to start a discussion to define a unifying framework for predicate invention and theory revision.