Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream?
Neural computations along the ventral visual stream, -- which culminates in the inferior temporal (IT) cortex -- enable humans and monkeys to recognize objects quickly. Primate IT is organized topographically: nearby neurons have similar response properties. Yet the best models of the ventral visual...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
Cognitive Computational Neuroscience
2021
|
| Online Access: | https://hdl.handle.net/1721.1/130332 |
| _version_ | 1826211175907459072 |
|---|---|
| author | Jozwik, Kamila Maria Lee, Hyo-Dong Kanwisher, Nancy DiCarlo, James |
| author2 | McGovern Institute for Brain Research at MIT |
| author_facet | McGovern Institute for Brain Research at MIT Jozwik, Kamila Maria Lee, Hyo-Dong Kanwisher, Nancy DiCarlo, James |
| author_sort | Jozwik, Kamila Maria |
| collection | MIT |
| description | Neural computations along the ventral visual stream, -- which culminates in the inferior temporal (IT) cortex -- enable humans and monkeys to recognize objects quickly. Primate IT is organized topographically: nearby neurons have similar response properties. Yet the best models of the ventral visual stream - deep artificial neural networks (ANNs) – have “IT” layers that lack topography. We built Topographic Deep ANNs (TDANNs) by incorporating a proxy wiring cost alongside the standard ImageNet categorization cost in the two “IT-like” layers of AlexNet (Lee et al., 2018), by specifying that “neurons” that have similar response properties should be physically close to each other. This cost both induced topographic structure and altered tuning characteristics of model IT neurons. We presented 2560 naturalistic images to monkeys and to ANNs. We found that, relative to the base (nontopographic) model, the “neurons” in the “IT” layer of some of the TDANN models matched actual IT neurons slightly better, and the dimensionality of the TDANN “IT” neural population was much closer to that of the measured monkey IT neural population. We also found that, while TDANNs did not show a statistically significant better match to human object discrimination behavior, detailed analysis suggests a trend in that direction. Taken together, TDANNs may better capture properties of IT cortex and wiring costs might be the cause of topographic organization in primate IT. |
| first_indexed | 2024-09-23T15:01:47Z |
| format | Article |
| id | mit-1721.1/130332 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T15:01:47Z |
| publishDate | 2021 |
| publisher | Cognitive Computational Neuroscience |
| record_format | dspace |
| spelling | mit-1721.1/1303322022-10-02T00:07:05Z Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? Jozwik, Kamila Maria Lee, Hyo-Dong Kanwisher, Nancy DiCarlo, James McGovern Institute for Brain Research at MIT Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Neural computations along the ventral visual stream, -- which culminates in the inferior temporal (IT) cortex -- enable humans and monkeys to recognize objects quickly. Primate IT is organized topographically: nearby neurons have similar response properties. Yet the best models of the ventral visual stream - deep artificial neural networks (ANNs) – have “IT” layers that lack topography. We built Topographic Deep ANNs (TDANNs) by incorporating a proxy wiring cost alongside the standard ImageNet categorization cost in the two “IT-like” layers of AlexNet (Lee et al., 2018), by specifying that “neurons” that have similar response properties should be physically close to each other. This cost both induced topographic structure and altered tuning characteristics of model IT neurons. We presented 2560 naturalistic images to monkeys and to ANNs. We found that, relative to the base (nontopographic) model, the “neurons” in the “IT” layer of some of the TDANN models matched actual IT neurons slightly better, and the dimensionality of the TDANN “IT” neural population was much closer to that of the measured monkey IT neural population. We also found that, while TDANNs did not show a statistically significant better match to human object discrimination behavior, detailed analysis suggests a trend in that direction. Taken together, TDANNs may better capture properties of IT cortex and wiring costs might be the cause of topographic organization in primate IT. Wellcome Trust (Award 206521/Z/17/Z) National Institutes of Health (Grant DP1HD091947) Simons Foundation (Grants SCGB 325500, 542965) NSF (Award CCF-1231216) 2021-04-01T15:13:07Z 2021-04-01T15:13:07Z 2019-12 2019-09 2021-03-29T16:15:20Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/130332 Jozwik, Kamila Maria et al. "Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream?" 2019 Conference on Cognitive Computational Neuroscience, September 2019, Berlin, Germany, Cognitive Computational Neuroscience, December 2019. en http://dx.doi.org/10.32470/ccn.2019.1019-0 2019 Conference on Cognitive Computational Neuroscience Creative Commons Attribution 3.0 unported license https://creativecommons.org/licenses/by/3.0/ application/pdf Cognitive Computational Neuroscience Cognitive Computational Neuroscience |
| spellingShingle | Jozwik, Kamila Maria Lee, Hyo-Dong Kanwisher, Nancy DiCarlo, James Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? |
| title | Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? |
| title_full | Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? |
| title_fullStr | Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? |
| title_full_unstemmed | Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? |
| title_short | Are Topographic Deep Convolutional Neural Networks Better Models of the Ventral Visual Stream? |
| title_sort | are topographic deep convolutional neural networks better models of the ventral visual stream |
| url | https://hdl.handle.net/1721.1/130332 |
| work_keys_str_mv | AT jozwikkamilamaria aretopographicdeepconvolutionalneuralnetworksbettermodelsoftheventralvisualstream AT leehyodong aretopographicdeepconvolutionalneuralnetworksbettermodelsoftheventralvisualstream AT kanwishernancy aretopographicdeepconvolutionalneuralnetworksbettermodelsoftheventralvisualstream AT dicarlojames aretopographicdeepconvolutionalneuralnetworksbettermodelsoftheventralvisualstream |