Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling
Abstract Primate visual cortex exhibits key organizational principles: cortical magnification, eccentricity-dependent receptive field size and spatial frequency tuning as well as radial bias. We provide compelling evidence that these principles arise from the interplay of the non-uniform distributio...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2024-04-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-024-59376-x |
| _version_ | 1797199569622138880 |
|---|---|
| author | Danny da Costa Lukas Kornemann Rainer Goebel Mario Senden |
| author_facet | Danny da Costa Lukas Kornemann Rainer Goebel Mario Senden |
| author_sort | Danny da Costa |
| collection | DOAJ |
| description | Abstract Primate visual cortex exhibits key organizational principles: cortical magnification, eccentricity-dependent receptive field size and spatial frequency tuning as well as radial bias. We provide compelling evidence that these principles arise from the interplay of the non-uniform distribution of retinal ganglion cells, and a quasi-uniform convergence rate from the retina to the cortex. We show that convolutional neural networks outfitted with a retinal sampling layer, which resamples images according to retinal ganglion cell density, develop these organizational principles. Surprisingly, our results indicate that radial bias is spatial-frequency dependent and only manifests for high spatial frequencies. For low spatial frequencies, the bias shifts towards orthogonal orientations. These findings introduce a novel hypothesis about the origin of radial bias. Quasi-uniform convergence limits the range of spatial frequencies (in retinal space) that can be resolved, while retinal sampling determines the spatial frequency content throughout the retina. |
| first_indexed | 2024-04-24T07:17:50Z |
| format | Article |
| id | doaj.art-d6572f12247c42f3a589d5a54ab683b4 |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-04-24T07:17:50Z |
| publishDate | 2024-04-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-d6572f12247c42f3a589d5a54ab683b42024-04-21T11:14:09ZengNature PortfolioScientific Reports2045-23222024-04-0114111410.1038/s41598-024-59376-xConvolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal samplingDanny da Costa0Lukas Kornemann1Rainer Goebel2Mario Senden3Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht UniversityDepartment of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht UniversityDepartment of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht UniversityDepartment of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht UniversityAbstract Primate visual cortex exhibits key organizational principles: cortical magnification, eccentricity-dependent receptive field size and spatial frequency tuning as well as radial bias. We provide compelling evidence that these principles arise from the interplay of the non-uniform distribution of retinal ganglion cells, and a quasi-uniform convergence rate from the retina to the cortex. We show that convolutional neural networks outfitted with a retinal sampling layer, which resamples images according to retinal ganglion cell density, develop these organizational principles. Surprisingly, our results indicate that radial bias is spatial-frequency dependent and only manifests for high spatial frequencies. For low spatial frequencies, the bias shifts towards orthogonal orientations. These findings introduce a novel hypothesis about the origin of radial bias. Quasi-uniform convergence limits the range of spatial frequencies (in retinal space) that can be resolved, while retinal sampling determines the spatial frequency content throughout the retina.https://doi.org/10.1038/s41598-024-59376-xRetinal samplingGanglion cellsConvolutional neural networksReceptive field mappingSpatial frequency tuningRadial bias |
| spellingShingle | Danny da Costa Lukas Kornemann Rainer Goebel Mario Senden Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling Scientific Reports Retinal sampling Ganglion cells Convolutional neural networks Receptive field mapping Spatial frequency tuning Radial bias |
| title | Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling |
| title_full | Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling |
| title_fullStr | Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling |
| title_full_unstemmed | Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling |
| title_short | Convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling |
| title_sort | convolutional neural networks develop major organizational principles of early visual cortex when enhanced with retinal sampling |
| topic | Retinal sampling Ganglion cells Convolutional neural networks Receptive field mapping Spatial frequency tuning Radial bias |
| url | https://doi.org/10.1038/s41598-024-59376-x |
| work_keys_str_mv | AT dannydacosta convolutionalneuralnetworksdevelopmajororganizationalprinciplesofearlyvisualcortexwhenenhancedwithretinalsampling AT lukaskornemann convolutionalneuralnetworksdevelopmajororganizationalprinciplesofearlyvisualcortexwhenenhancedwithretinalsampling AT rainergoebel convolutionalneuralnetworksdevelopmajororganizationalprinciplesofearlyvisualcortexwhenenhancedwithretinalsampling AT mariosenden convolutionalneuralnetworksdevelopmajororganizationalprinciplesofearlyvisualcortexwhenenhancedwithretinalsampling |