Cargando…

Multitask learning over shared subspaces

This paper uses constructs from machine learning to define pairs of learning tasks that either shared or did not share a common subspace. Human subjects then learnt these tasks using a feedback-based approach and we hypothesised that learning would be boosted for shared subspaces. Our findings broad...

Descripción completa

Detalles Bibliográficos
Autores principales: Menghi, Nicholas, Kacar, Kemal, Penny, Will
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Public Library of Science 2021
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8284664/
https://www.ncbi.nlm.nih.gov/pubmed/34228719
http://dx.doi.org/10.1371/journal.pcbi.1009092
_version_ 1783723431949238272
author Menghi, Nicholas
Kacar, Kemal
Penny, Will
author_facet Menghi, Nicholas
Kacar, Kemal
Penny, Will
author_sort Menghi, Nicholas
collection PubMed
description This paper uses constructs from machine learning to define pairs of learning tasks that either shared or did not share a common subspace. Human subjects then learnt these tasks using a feedback-based approach and we hypothesised that learning would be boosted for shared subspaces. Our findings broadly supported this hypothesis with either better performance on the second task if it shared the same subspace as the first, or positive correlations over task performance for shared subspaces. These empirical findings were compared to the behaviour of a Neural Network model trained using sequential Bayesian learning and human performance was found to be consistent with a minimal capacity variant of this model. Networks with an increased representational capacity, and networks without Bayesian learning, did not show these transfer effects. We propose that the concept of shared subspaces provides a useful framework for the experimental study of human multitask and transfer learning.
format Online
Article
Text
id pubmed-8284664
institution National Center for Biotechnology Information
language English
publishDate 2021
publisher Public Library of Science
record_format MEDLINE/PubMed
spelling pubmed-82846642021-07-28 Multitask learning over shared subspaces Menghi, Nicholas Kacar, Kemal Penny, Will PLoS Comput Biol Research Article This paper uses constructs from machine learning to define pairs of learning tasks that either shared or did not share a common subspace. Human subjects then learnt these tasks using a feedback-based approach and we hypothesised that learning would be boosted for shared subspaces. Our findings broadly supported this hypothesis with either better performance on the second task if it shared the same subspace as the first, or positive correlations over task performance for shared subspaces. These empirical findings were compared to the behaviour of a Neural Network model trained using sequential Bayesian learning and human performance was found to be consistent with a minimal capacity variant of this model. Networks with an increased representational capacity, and networks without Bayesian learning, did not show these transfer effects. We propose that the concept of shared subspaces provides a useful framework for the experimental study of human multitask and transfer learning. Public Library of Science 2021-07-06 /pmc/articles/PMC8284664/ /pubmed/34228719 http://dx.doi.org/10.1371/journal.pcbi.1009092 Text en © 2021 Menghi et al https://creativecommons.org/licenses/by/4.0/This is an open access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/) , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
spellingShingle Research Article
Menghi, Nicholas
Kacar, Kemal
Penny, Will
Multitask learning over shared subspaces
title Multitask learning over shared subspaces
title_full Multitask learning over shared subspaces
title_fullStr Multitask learning over shared subspaces
title_full_unstemmed Multitask learning over shared subspaces
title_short Multitask learning over shared subspaces
title_sort multitask learning over shared subspaces
topic Research Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8284664/
https://www.ncbi.nlm.nih.gov/pubmed/34228719
http://dx.doi.org/10.1371/journal.pcbi.1009092
work_keys_str_mv AT menghinicholas multitasklearningoversharedsubspaces
AT kacarkemal multitasklearningoversharedsubspaces
AT pennywill multitasklearningoversharedsubspaces