Emergence of Sparse Representations from Noise

Trenton Bricken, Rylan Schaeffer, Bruno Olshausen, Gabriel Kreiman
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:3148-3191, 2023.

Abstract

A hallmark of biological neural networks, which distinguishes them from their artificial counterparts, is the high degree of sparsity in their activations. This discrepancy raises three questions our work helps to answer: (i) Why are biological networks so sparse? (ii) What are the benefits of this sparsity? (iii) How can these benefits be utilized by deep learning models? Our answers to all of these questions center around training networks to handle random noise. Surprisingly, we discover that noisy training introduces three implicit loss terms that result in sparsely firing neurons specializing to high variance features of the dataset. When trained to reconstruct noisy-CIFAR10, neurons learn biological receptive fields. More broadly, noisy training presents a new approach to potentially increase model interpretability with additional benefits to robustness and computational efficiency.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-bricken23a, title = {Emergence of Sparse Representations from Noise}, author = {Bricken, Trenton and Schaeffer, Rylan and Olshausen, Bruno and Kreiman, Gabriel}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {3148--3191}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/bricken23a/bricken23a.pdf}, url = {https://proceedings.mlr.press/v202/bricken23a.html}, abstract = {A hallmark of biological neural networks, which distinguishes them from their artificial counterparts, is the high degree of sparsity in their activations. This discrepancy raises three questions our work helps to answer: (i) Why are biological networks so sparse? (ii) What are the benefits of this sparsity? (iii) How can these benefits be utilized by deep learning models? Our answers to all of these questions center around training networks to handle random noise. Surprisingly, we discover that noisy training introduces three implicit loss terms that result in sparsely firing neurons specializing to high variance features of the dataset. When trained to reconstruct noisy-CIFAR10, neurons learn biological receptive fields. More broadly, noisy training presents a new approach to potentially increase model interpretability with additional benefits to robustness and computational efficiency.} }
Endnote
%0 Conference Paper %T Emergence of Sparse Representations from Noise %A Trenton Bricken %A Rylan Schaeffer %A Bruno Olshausen %A Gabriel Kreiman %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-bricken23a %I PMLR %P 3148--3191 %U https://proceedings.mlr.press/v202/bricken23a.html %V 202 %X A hallmark of biological neural networks, which distinguishes them from their artificial counterparts, is the high degree of sparsity in their activations. This discrepancy raises three questions our work helps to answer: (i) Why are biological networks so sparse? (ii) What are the benefits of this sparsity? (iii) How can these benefits be utilized by deep learning models? Our answers to all of these questions center around training networks to handle random noise. Surprisingly, we discover that noisy training introduces three implicit loss terms that result in sparsely firing neurons specializing to high variance features of the dataset. When trained to reconstruct noisy-CIFAR10, neurons learn biological receptive fields. More broadly, noisy training presents a new approach to potentially increase model interpretability with additional benefits to robustness and computational efficiency.
APA
Bricken, T., Schaeffer, R., Olshausen, B. & Kreiman, G.. (2023). Emergence of Sparse Representations from Noise. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:3148-3191 Available from https://proceedings.mlr.press/v202/bricken23a.html.

Related Material