Engineering recurrent neural networks from task-relevant manifolds and dynamics
Copyright: © 2020 Pollock, Jazayeri. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Many cognitive processes in...
| Main Authors: | , |
|---|---|
| 其他作者: | |
| 格式: | 文件 |
| 语言: | English |
| 出版: |
Public Library of Science (PLoS)
2021
|
| 在线阅读: | https://hdl.handle.net/1721.1/135247 |
| _version_ | 1826190280859058176 |
|---|---|
| author | Pollock, Eli Jazayeri, Mehrdad |
| author2 | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences |
| author_facet | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Pollock, Eli Jazayeri, Mehrdad |
| author_sort | Pollock, Eli |
| collection | MIT |
| description | Copyright: © 2020 Pollock, Jazayeri. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Many cognitive processes involve transformations of distributed representations in neural populations, creating a need for population-level models. Recurrent neural network models fulfill this need, but there are many open questions about how their connectivity gives rise to dynamics that solve a task. Here, we present a method for finding the connectivity of networks for which the dynamics are specified to solve a task in an interpretable way. We apply our method to a working memory task by synthesizing a network that implements a drift-diffusion process over a ring-shaped manifold. We also use our method to demonstrate how inputs can be used to control network dynamics for cognitive flexibility and explore the relationship between representation geometry and network capacity. Our work fits within the broader context of understanding neural computations as dynamics over relatively low-dimensional manifolds formed by correlated patterns of neurons. |
| first_indexed | 2024-09-23T08:37:51Z |
| format | Article |
| id | mit-1721.1/135247 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T08:37:51Z |
| publishDate | 2021 |
| publisher | Public Library of Science (PLoS) |
| record_format | dspace |
| spelling | mit-1721.1/1352472023-09-28T20:14:52Z Engineering recurrent neural networks from task-relevant manifolds and dynamics Pollock, Eli Jazayeri, Mehrdad Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences McGovern Institute for Brain Research at MIT Copyright: © 2020 Pollock, Jazayeri. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Many cognitive processes involve transformations of distributed representations in neural populations, creating a need for population-level models. Recurrent neural network models fulfill this need, but there are many open questions about how their connectivity gives rise to dynamics that solve a task. Here, we present a method for finding the connectivity of networks for which the dynamics are specified to solve a task in an interpretable way. We apply our method to a working memory task by synthesizing a network that implements a drift-diffusion process over a ring-shaped manifold. We also use our method to demonstrate how inputs can be used to control network dynamics for cognitive flexibility and explore the relationship between representation geometry and network capacity. Our work fits within the broader context of understanding neural computations as dynamics over relatively low-dimensional manifolds formed by correlated patterns of neurons. 2021-10-27T20:22:38Z 2021-10-27T20:22:38Z 2020 2021-03-23T18:54:50Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/135247 en 10.1371/JOURNAL.PCBI.1008128 PLoS Computational Biology Creative Commons Attribution 4.0 International license https://creativecommons.org/licenses/by/4.0/ application/pdf Public Library of Science (PLoS) PLoS |
| spellingShingle | Pollock, Eli Jazayeri, Mehrdad Engineering recurrent neural networks from task-relevant manifolds and dynamics |
| title | Engineering recurrent neural networks from task-relevant manifolds and dynamics |
| title_full | Engineering recurrent neural networks from task-relevant manifolds and dynamics |
| title_fullStr | Engineering recurrent neural networks from task-relevant manifolds and dynamics |
| title_full_unstemmed | Engineering recurrent neural networks from task-relevant manifolds and dynamics |
| title_short | Engineering recurrent neural networks from task-relevant manifolds and dynamics |
| title_sort | engineering recurrent neural networks from task relevant manifolds and dynamics |
| url | https://hdl.handle.net/1721.1/135247 |
| work_keys_str_mv | AT pollockeli engineeringrecurrentneuralnetworksfromtaskrelevantmanifoldsanddynamics AT jazayerimehrdad engineeringrecurrentneuralnetworksfromtaskrelevantmanifoldsanddynamics |