Weight-Sharing Regularization

Mehran Shakerinava, Motahareh MS Sohrabi, Siamak Ravanbakhsh, Simon Lacoste-Julien
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, PMLR 238:4204-4212, 2024.

Abstract

Weight-sharing is ubiquitous in deep learning. Motivated by this, we propose a “weight-sharing regularization” penalty on the weights $w \in \mathbb{R}^d$ of a neural network, defined as $\mathcal{R}(w) = \frac{1}{d - 1}\sum_{i > j}^d |w_i - w_j|$. We study the proximal mapping of $\mathcal{R}$ and provide an intuitive interpretation of it in terms of a physical system of interacting particles. We also parallelize existing algorithms for $\mathrm{prox}_{\mathcal{R}}$ (to run on GPU) and find that one of them is fast in practice but slow ($O(d)$) for worst-case inputs. Using the physical interpretation, we design a novel parallel algorithm which runs in $O(\log^3 d)$ when sufficient processors are available, thus guaranteeing fast training. Our experiments reveal that weight-sharing regularization enables fully connected networks to learn convolution-like filters even when pixels have been shuffled while convolutional neural networks fail in this setting. Our code is available on \href{https://github.com/motahareh-sohrabi/weight-sharing-regularization}{github}.

Cite this Paper

BibTeX
@InProceedings{pmlr-v238-shakerinava24a, title = { Weight-Sharing Regularization }, author = {Shakerinava, Mehran and MS Sohrabi, Motahareh and Ravanbakhsh, Siamak and Lacoste-Julien, Simon}, booktitle = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics}, pages = {4204--4212}, year = {2024}, editor = {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen}, volume = {238}, series = {Proceedings of Machine Learning Research}, month = {02--04 May}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v238/shakerinava24a/shakerinava24a.pdf}, url = {https://proceedings.mlr.press/v238/shakerinava24a.html}, abstract = { Weight-sharing is ubiquitous in deep learning. Motivated by this, we propose a “weight-sharing regularization” penalty on the weights $w \in \mathbb{R}^d$ of a neural network, defined as $\mathcal{R}(w) = \frac{1}{d - 1}\sum_{i > j}^d |w_i - w_j|$. We study the proximal mapping of $\mathcal{R}$ and provide an intuitive interpretation of it in terms of a physical system of interacting particles. We also parallelize existing algorithms for $\mathrm{prox}_{\mathcal{R}}$ (to run on GPU) and find that one of them is fast in practice but slow ($O(d)$) for worst-case inputs. Using the physical interpretation, we design a novel parallel algorithm which runs in $O(\log^3 d)$ when sufficient processors are available, thus guaranteeing fast training. Our experiments reveal that weight-sharing regularization enables fully connected networks to learn convolution-like filters even when pixels have been shuffled while convolutional neural networks fail in this setting. Our code is available on \href{https://github.com/motahareh-sohrabi/weight-sharing-regularization}{github}. } }
Endnote
%0 Conference Paper %T Weight-Sharing Regularization %A Mehran Shakerinava %A Motahareh MS Sohrabi %A Siamak Ravanbakhsh %A Simon Lacoste-Julien %B Proceedings of The 27th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2024 %E Sanjoy Dasgupta %E Stephan Mandt %E Yingzhen Li %F pmlr-v238-shakerinava24a %I PMLR %P 4204--4212 %U https://proceedings.mlr.press/v238/shakerinava24a.html %V 238 %X Weight-sharing is ubiquitous in deep learning. Motivated by this, we propose a “weight-sharing regularization” penalty on the weights $w \in \mathbb{R}^d$ of a neural network, defined as $\mathcal{R}(w) = \frac{1}{d - 1}\sum_{i > j}^d |w_i - w_j|$. We study the proximal mapping of $\mathcal{R}$ and provide an intuitive interpretation of it in terms of a physical system of interacting particles. We also parallelize existing algorithms for $\mathrm{prox}_{\mathcal{R}}$ (to run on GPU) and find that one of them is fast in practice but slow ($O(d)$) for worst-case inputs. Using the physical interpretation, we design a novel parallel algorithm which runs in $O(\log^3 d)$ when sufficient processors are available, thus guaranteeing fast training. Our experiments reveal that weight-sharing regularization enables fully connected networks to learn convolution-like filters even when pixels have been shuffled while convolutional neural networks fail in this setting. Our code is available on \href{https://github.com/motahareh-sohrabi/weight-sharing-regularization}{github}.
APA
Shakerinava, M., MS Sohrabi, M., Ravanbakhsh, S. & Lacoste-Julien, S.. (2024). Weight-Sharing Regularization . Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:4204-4212 Available from https://proceedings.mlr.press/v238/shakerinava24a.html.

Related Material