Introducing topography in convolutional neural networks. Poli, M., Dupoux, E., & Riad, R. October, 2022. arXiv:2211.13152 [cs, eess]![Introducing topography in convolutional neural networks [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Parts of the brain that carry sensory tasks are organized topographically: nearby neurons are responsive to the same properties of input signals. Thus, in this work, inspired by the neuroscience literature, we proposed a new topographic inductive bias in Convolutional Neural Networks (CNNs). To achieve this, we introduced a new topographic loss and an efficient implementation to topographically organize each convolutional layer of any CNN. We benchmarked our new method on 4 datasets and 3 models in vision and audio tasks and showed equivalent performance to all benchmarks. Besides, we also showcased the generalizability of our topographic loss with how it can be used with different topographic organizations in CNNs. Finally, we demonstrated that adding the topographic inductive bias made CNNs more resistant to pruning. Our approach provides a new avenue to obtain models that are more memory efficient while maintaining better accuracy.
@misc{poli_introducing_2022,
title = {Introducing topography in convolutional neural networks},
copyright = {All rights reserved},
url = {http://arxiv.org/abs/2211.13152},
doi = {10.48550/arXiv.2211.13152},
abstract = {Parts of the brain that carry sensory tasks are organized topographically: nearby neurons are responsive to the same properties of input signals. Thus, in this work, inspired by the neuroscience literature, we proposed a new topographic inductive bias in Convolutional Neural Networks (CNNs). To achieve this, we introduced a new topographic loss and an efficient implementation to topographically organize each convolutional layer of any CNN. We benchmarked our new method on 4 datasets and 3 models in vision and audio tasks and showed equivalent performance to all benchmarks. Besides, we also showcased the generalizability of our topographic loss with how it can be used with different topographic organizations in CNNs. Finally, we demonstrated that adding the topographic inductive bias made CNNs more resistant to pruning. Our approach provides a new avenue to obtain models that are more memory efficient while maintaining better accuracy.},
urldate = {2023-03-04},
publisher = {arXiv},
author = {Poli, Maxime and Dupoux, Emmanuel and Riad, Rachid},
month = oct,
year = {2022},
note = {arXiv:2211.13152 [cs, eess]},
keywords = {Computer Science - Machine Learning, Computer Science - Neural and Evolutionary Computing, Computer Science - Sound, Electrical Engineering and Systems Science - Audio and Speech Processing},
}
Downloads: 0
{"_id":"5iTDTMNot5782oMJS","bibbaseid":"poli-dupoux-riad-introducingtopographyinconvolutionalneuralnetworks-2022","author_short":["Poli, M.","Dupoux, E.","Riad, R."],"bibdata":{"bibtype":"misc","type":"misc","title":"Introducing topography in convolutional neural networks","copyright":"All rights reserved","url":"http://arxiv.org/abs/2211.13152","doi":"10.48550/arXiv.2211.13152","abstract":"Parts of the brain that carry sensory tasks are organized topographically: nearby neurons are responsive to the same properties of input signals. Thus, in this work, inspired by the neuroscience literature, we proposed a new topographic inductive bias in Convolutional Neural Networks (CNNs). To achieve this, we introduced a new topographic loss and an efficient implementation to topographically organize each convolutional layer of any CNN. We benchmarked our new method on 4 datasets and 3 models in vision and audio tasks and showed equivalent performance to all benchmarks. Besides, we also showcased the generalizability of our topographic loss with how it can be used with different topographic organizations in CNNs. Finally, we demonstrated that adding the topographic inductive bias made CNNs more resistant to pruning. Our approach provides a new avenue to obtain models that are more memory efficient while maintaining better accuracy.","urldate":"2023-03-04","publisher":"arXiv","author":[{"propositions":[],"lastnames":["Poli"],"firstnames":["Maxime"],"suffixes":[]},{"propositions":[],"lastnames":["Dupoux"],"firstnames":["Emmanuel"],"suffixes":[]},{"propositions":[],"lastnames":["Riad"],"firstnames":["Rachid"],"suffixes":[]}],"month":"October","year":"2022","note":"arXiv:2211.13152 [cs, eess]","keywords":"Computer Science - Machine Learning, Computer Science - Neural and Evolutionary Computing, Computer Science - Sound, Electrical Engineering and Systems Science - Audio and Speech Processing","bibtex":"@misc{poli_introducing_2022,\n\ttitle = {Introducing topography in convolutional neural networks},\n\tcopyright = {All rights reserved},\n\turl = {http://arxiv.org/abs/2211.13152},\n\tdoi = {10.48550/arXiv.2211.13152},\n\tabstract = {Parts of the brain that carry sensory tasks are organized topographically: nearby neurons are responsive to the same properties of input signals. Thus, in this work, inspired by the neuroscience literature, we proposed a new topographic inductive bias in Convolutional Neural Networks (CNNs). To achieve this, we introduced a new topographic loss and an efficient implementation to topographically organize each convolutional layer of any CNN. We benchmarked our new method on 4 datasets and 3 models in vision and audio tasks and showed equivalent performance to all benchmarks. Besides, we also showcased the generalizability of our topographic loss with how it can be used with different topographic organizations in CNNs. Finally, we demonstrated that adding the topographic inductive bias made CNNs more resistant to pruning. Our approach provides a new avenue to obtain models that are more memory efficient while maintaining better accuracy.},\n\turldate = {2023-03-04},\n\tpublisher = {arXiv},\n\tauthor = {Poli, Maxime and Dupoux, Emmanuel and Riad, Rachid},\n\tmonth = oct,\n\tyear = {2022},\n\tnote = {arXiv:2211.13152 [cs, eess]},\n\tkeywords = {Computer Science - Machine Learning, Computer Science - Neural and Evolutionary Computing, Computer Science - Sound, Electrical Engineering and Systems Science - Audio and Speech Processing},\n}\n\n","author_short":["Poli, M.","Dupoux, E.","Riad, R."],"key":"poli_introducing_2022","id":"poli_introducing_2022","bibbaseid":"poli-dupoux-riad-introducingtopographyinconvolutionalneuralnetworks-2022","role":"author","urls":{"Paper":"http://arxiv.org/abs/2211.13152"},"keyword":["Computer Science - Machine Learning","Computer Science - Neural and Evolutionary Computing","Computer Science - Sound","Electrical Engineering and Systems Science - Audio and Speech Processing"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"misc","biburl":"https://bibbase.org/zotero/mxmplx","dataSources":["aXmRAq63YsH7a3ufx"],"keywords":["computer science - machine learning","computer science - neural and evolutionary computing","computer science - sound","electrical engineering and systems science - audio and speech processing"],"search_terms":["introducing","topography","convolutional","neural","networks","poli","dupoux","riad"],"title":"Introducing topography in convolutional neural networks","year":2022}