The Relational Bottleneck as an Inductive Bias for Efficient Abstraction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Webb, Taylor W., Frankland, Steven M., Altabaa, Awni, Segert, Simon, Krishnamurthy, Kamesh, Campbell, Declan, Russin, Jacob, Giallanza, Tyler, Dulberg, Zack, O'Reilly, Randall, Lafferty, John, Cohen, Jonathan D.
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914778085064704
author Webb, Taylor W.
Frankland, Steven M.
Altabaa, Awni
Segert, Simon
Krishnamurthy, Kamesh
Campbell, Declan
Russin, Jacob
Giallanza, Tyler
Dulberg, Zack
O'Reilly, Randall
Lafferty, John
Cohen, Jonathan D.
author_facet Webb, Taylor W.
Frankland, Steven M.
Altabaa, Awni
Segert, Simon
Krishnamurthy, Kamesh
Campbell, Declan
Russin, Jacob
Giallanza, Tyler
Dulberg, Zack
O'Reilly, Randall
Lafferty, John
Cohen, Jonathan D.
contents A central challenge for cognitive science is to explain how abstract concepts are acquired from limited experience. This has often been framed in terms of a dichotomy between connectionist and symbolic cognitive models. Here, we highlight a recently emerging line of work that suggests a novel reconciliation of these approaches, by exploiting an inductive bias that we term the relational bottleneck. In that approach, neural networks are constrained via their architecture to focus on relations between perceptual inputs, rather than the attributes of individual inputs. We review a family of models that employ this approach to induce abstractions in a data-efficient manner, emphasizing their potential as candidate models for the acquisition of abstract concepts in the human mind and brain.
format Preprint
id arxiv_https___arxiv_org_abs_2309_06629
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle The Relational Bottleneck as an Inductive Bias for Efficient Abstraction
Webb, Taylor W.
Frankland, Steven M.
Altabaa, Awni
Segert, Simon
Krishnamurthy, Kamesh
Campbell, Declan
Russin, Jacob
Giallanza, Tyler
Dulberg, Zack
O'Reilly, Randall
Lafferty, John
Cohen, Jonathan D.
Artificial Intelligence
Neural and Evolutionary Computing
A central challenge for cognitive science is to explain how abstract concepts are acquired from limited experience. This has often been framed in terms of a dichotomy between connectionist and symbolic cognitive models. Here, we highlight a recently emerging line of work that suggests a novel reconciliation of these approaches, by exploiting an inductive bias that we term the relational bottleneck. In that approach, neural networks are constrained via their architecture to focus on relations between perceptual inputs, rather than the attributes of individual inputs. We review a family of models that employ this approach to induce abstractions in a data-efficient manner, emphasizing their potential as candidate models for the acquisition of abstract concepts in the human mind and brain.
title The Relational Bottleneck as an Inductive Bias for Efficient Abstraction
topic Artificial Intelligence
Neural and Evolutionary Computing
url https://arxiv.org/abs/2309.06629