Human-like systematic generalization through a meta-learning neural network
The power of human language and thought arises from systematic compositionality—the algebraic ability to understand and produce novel combinations from known components. Fodor and Pylyshyn1 famously argued that artificial neural networks lack this capacity and are therefore not viable models of the...
| Autores: | , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2023 |
| País: | España |
| Institución: | Universitat Pompeu Fabra |
| Repositorio: | Repositorio Digital de la UPF |
| OAI Identifier: | oai:repositori.upf.edu:10230/58561 |
| Acceso en línea: | http://hdl.handle.net/10230/58561 http://dx.doi.org/10.1038/s41586-023-06668-3 |
| Access Level: | acceso abierto |
| Palabra clave: | Xarxes neuronals (Informàtica) Metaaprenentatge Neurociències |
| id |
ES_a4ad33d28f8bb70d1cfa14e9e25e9bbd |
|---|---|
| oai_identifier_str |
oai:repositori.upf.edu:10230/58561 |
| network_acronym_str |
ES |
| network_name_str |
España |
| repository_id_str |
|
| spelling |
Human-like systematic generalization through a meta-learning neural networkBaroni, MarcoLake, Brenden M.Xarxes neuronals (Informàtica)MetaaprenentatgeNeurociènciesThe power of human language and thought arises from systematic compositionality—the algebraic ability to understand and produce novel combinations from known components. Fodor and Pylyshyn1 famously argued that artificial neural networks lack this capacity and are therefore not viable models of the mind. Neural networks have advanced considerably in the years since, yet the systematicity challenge persists. Here we successfully address Fodor and Pylyshyn’s challenge by providing evidence that neural networks can achieve human-like systematicity when optimized for their compositional skills. To do so, we introduce the meta-learning for compositionality (MLC) approach for guiding training through a dynamic stream of compositional tasks. To compare humans and machines, we conducted human behavioural experiments using an instruction learning paradigm. After considering seven different models, we found that, in contrast to perfectly systematic but rigid probabilistic symbolic models, and perfectly flexible but unsystematic neural networks, only MLC achieves both the systematicity and flexibility needed for human-like generalization. MLC also advances the compositional skills of machine learning systems in several systematic generalization benchmarks. Our results show how a standard neural network architecture, optimized for its compositional skills, can mimic human systematic generalization in a head-to-head comparison.Nature Research202320232023info:eu-repo/semantics/articleinfo:eu-repo/semantics/acceptedVersionapplication/pdfapplication/pdfhttp://hdl.handle.net/10230/58561http://dx.doi.org/10.1038/s41586-023-06668-3reponame:Repositorio Digital de la UPFinstname:Universitat Pompeu FabraInglésNature. 2023 Nov;623(7985):115–21.© The Author(s) 2023. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.http://creativecommons.org/licenses/by/4.0/info:eu-repo/semantics/openAccessoai:repositori.upf.edu:10230/585612026-06-12T07:21:37Z |
| dc.title.none.fl_str_mv |
Human-like systematic generalization through a meta-learning neural network |
| title |
Human-like systematic generalization through a meta-learning neural network |
| spellingShingle |
Human-like systematic generalization through a meta-learning neural network Baroni, Marco Xarxes neuronals (Informàtica) Metaaprenentatge Neurociències |
| title_short |
Human-like systematic generalization through a meta-learning neural network |
| title_full |
Human-like systematic generalization through a meta-learning neural network |
| title_fullStr |
Human-like systematic generalization through a meta-learning neural network |
| title_full_unstemmed |
Human-like systematic generalization through a meta-learning neural network |
| title_sort |
Human-like systematic generalization through a meta-learning neural network |
| dc.creator.none.fl_str_mv |
Baroni, Marco Lake, Brenden M. |
| author |
Baroni, Marco |
| author_facet |
Baroni, Marco Lake, Brenden M. |
| author_role |
author |
| author2 |
Lake, Brenden M. |
| author2_role |
author |
| dc.subject.none.fl_str_mv |
Xarxes neuronals (Informàtica) Metaaprenentatge Neurociències |
| topic |
Xarxes neuronals (Informàtica) Metaaprenentatge Neurociències |
| description |
The power of human language and thought arises from systematic compositionality—the algebraic ability to understand and produce novel combinations from known components. Fodor and Pylyshyn1 famously argued that artificial neural networks lack this capacity and are therefore not viable models of the mind. Neural networks have advanced considerably in the years since, yet the systematicity challenge persists. Here we successfully address Fodor and Pylyshyn’s challenge by providing evidence that neural networks can achieve human-like systematicity when optimized for their compositional skills. To do so, we introduce the meta-learning for compositionality (MLC) approach for guiding training through a dynamic stream of compositional tasks. To compare humans and machines, we conducted human behavioural experiments using an instruction learning paradigm. After considering seven different models, we found that, in contrast to perfectly systematic but rigid probabilistic symbolic models, and perfectly flexible but unsystematic neural networks, only MLC achieves both the systematicity and flexibility needed for human-like generalization. MLC also advances the compositional skills of machine learning systems in several systematic generalization benchmarks. Our results show how a standard neural network architecture, optimized for its compositional skills, can mimic human systematic generalization in a head-to-head comparison. |
| publishDate |
2023 |
| dc.date.none.fl_str_mv |
2023 2023 2023 |
| dc.type.none.fl_str_mv |
info:eu-repo/semantics/article info:eu-repo/semantics/acceptedVersion |
| format |
article |
| status_str |
acceptedVersion |
| dc.identifier.none.fl_str_mv |
http://hdl.handle.net/10230/58561 http://dx.doi.org/10.1038/s41586-023-06668-3 |
| url |
http://hdl.handle.net/10230/58561 http://dx.doi.org/10.1038/s41586-023-06668-3 |
| dc.language.none.fl_str_mv |
Inglés |
| language_invalid_str_mv |
Inglés |
| dc.relation.none.fl_str_mv |
Nature. 2023 Nov;623(7985):115–21. |
| dc.rights.none.fl_str_mv |
http://creativecommons.org/licenses/by/4.0/ info:eu-repo/semantics/openAccess |
| rights_invalid_str_mv |
http://creativecommons.org/licenses/by/4.0/ |
| eu_rights_str_mv |
openAccess |
| dc.format.none.fl_str_mv |
application/pdf application/pdf |
| dc.publisher.none.fl_str_mv |
Nature Research |
| publisher.none.fl_str_mv |
Nature Research |
| dc.source.none.fl_str_mv |
reponame:Repositorio Digital de la UPF instname:Universitat Pompeu Fabra |
| instname_str |
Universitat Pompeu Fabra |
| reponame_str |
Repositorio Digital de la UPF |
| collection |
Repositorio Digital de la UPF |
| repository.name.fl_str_mv |
|
| repository.mail.fl_str_mv |
|
| _version_ |
1869415517130850304 |
| score |
15,812455 |