Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks
Much of the information the brain processes and stores is temporal in nature—a spoken word or a handwritten signature, for example, is defined by how it unfolds in time. However, it remains unclear how neural circuits encode complex time-varying patterns. We show that by tuning the weights of a recu...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
eLife Sciences Publications Ltd
2018-03-01
|
| Series: | eLife |
| Subjects: | |
| Online Access: | https://elifesciences.org/articles/31134 |
| _version_ | 1829096705066270720 |
|---|---|
| author | Vishwa Goudar Dean V Buonomano |
| author_facet | Vishwa Goudar Dean V Buonomano |
| author_sort | Vishwa Goudar |
| collection | DOAJ |
| description | Much of the information the brain processes and stores is temporal in nature—a spoken word or a handwritten signature, for example, is defined by how it unfolds in time. However, it remains unclear how neural circuits encode complex time-varying patterns. We show that by tuning the weights of a recurrent neural network (RNN), it can recognize and then transcribe spoken digits. The model elucidates how neural dynamics in cortical networks may resolve three fundamental challenges: first, encode multiple time-varying sensory and motor patterns as stable neural trajectories; second, generalize across relevant spatial features; third, identify the same stimuli played at different speeds—we show that this temporal invariance emerges because the recurrent dynamics generate neural trajectories with appropriately modulated angular velocities. Together our results generate testable predictions as to how recurrent networks may use different mechanisms to generalize across the relevant spatial and temporal features of complex time-varying stimuli. |
| first_indexed | 2024-04-11T09:01:58Z |
| format | Article |
| id | doaj.art-8f9c4f3b6a194650a94bebb09619c995 |
| institution | Directory Open Access Journal |
| issn | 2050-084X |
| language | English |
| last_indexed | 2024-04-11T09:01:58Z |
| publishDate | 2018-03-01 |
| publisher | eLife Sciences Publications Ltd |
| record_format | Article |
| series | eLife |
| spelling | doaj.art-8f9c4f3b6a194650a94bebb09619c9952022-12-22T04:32:45ZengeLife Sciences Publications LtdeLife2050-084X2018-03-01710.7554/eLife.31134Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networksVishwa Goudar0https://orcid.org/0000-0002-6612-3076Dean V Buonomano1https://orcid.org/0000-0002-8528-9231Departments of Neurobiology, University of California, Los Angeles, Los Angeles, United StatesDepartments of Neurobiology, University of California, Los Angeles, Los Angeles, United States; Integrative Center for Learning and Memory, University of California, Los Angeles, Los Angeles, United States; Departments of Psychology, University of California, Los Angeles, Los Angeles, United StatesMuch of the information the brain processes and stores is temporal in nature—a spoken word or a handwritten signature, for example, is defined by how it unfolds in time. However, it remains unclear how neural circuits encode complex time-varying patterns. We show that by tuning the weights of a recurrent neural network (RNN), it can recognize and then transcribe spoken digits. The model elucidates how neural dynamics in cortical networks may resolve three fundamental challenges: first, encode multiple time-varying sensory and motor patterns as stable neural trajectories; second, generalize across relevant spatial features; third, identify the same stimuli played at different speeds—we show that this temporal invariance emerges because the recurrent dynamics generate neural trajectories with appropriately modulated angular velocities. Together our results generate testable predictions as to how recurrent networks may use different mechanisms to generalize across the relevant spatial and temporal features of complex time-varying stimuli.https://elifesciences.org/articles/31134temporal scalingrecurrent neural networksneural dynamicssensorimotor |
| spellingShingle | Vishwa Goudar Dean V Buonomano Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks eLife temporal scaling recurrent neural networks neural dynamics sensorimotor |
| title | Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks |
| title_full | Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks |
| title_fullStr | Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks |
| title_full_unstemmed | Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks |
| title_short | Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks |
| title_sort | encoding sensory and motor patterns as time invariant trajectories in recurrent neural networks |
| topic | temporal scaling recurrent neural networks neural dynamics sensorimotor |
| url | https://elifesciences.org/articles/31134 |
| work_keys_str_mv | AT vishwagoudar encodingsensoryandmotorpatternsastimeinvarianttrajectoriesinrecurrentneuralnetworks AT deanvbuonomano encodingsensoryandmotorpatternsastimeinvarianttrajectoriesinrecurrentneuralnetworks |