Directions of Curvature as an Explanation for Loss of Plasticity

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lewandowski, Alex, Tanaka, Haruto, Schuurmans, Dale, Machado, Marlos C.
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916423845019648
author Lewandowski, Alex
Tanaka, Haruto
Schuurmans, Dale
Machado, Marlos C.
author_facet Lewandowski, Alex
Tanaka, Haruto
Schuurmans, Dale
Machado, Marlos C.
contents Loss of plasticity is a phenomenon in which neural networks lose their ability to learn from new experience. Despite being empirically observed in several problem settings, little is understood about the mechanisms that lead to loss of plasticity. In this paper, we offer a consistent explanation for loss of plasticity: Neural networks lose directions of curvature during training and that loss of plasticity can be attributed to this reduction in curvature. To support such a claim, we provide a systematic investigation of loss of plasticity across continual learning tasks using MNIST, CIFAR-10 and ImageNet. Our findings illustrate that loss of curvature directions coincides with loss of plasticity, while also showing that previous explanations are insufficient to explain loss of plasticity in all settings. Lastly, we show that regularizers which mitigate loss of plasticity also preserve curvature, motivating a simple distributional regularizer that proves to be effective across the problem settings we considered.
format Preprint
id arxiv_https___arxiv_org_abs_2312_00246
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Directions of Curvature as an Explanation for Loss of Plasticity
Lewandowski, Alex
Tanaka, Haruto
Schuurmans, Dale
Machado, Marlos C.
Machine Learning
Loss of plasticity is a phenomenon in which neural networks lose their ability to learn from new experience. Despite being empirically observed in several problem settings, little is understood about the mechanisms that lead to loss of plasticity. In this paper, we offer a consistent explanation for loss of plasticity: Neural networks lose directions of curvature during training and that loss of plasticity can be attributed to this reduction in curvature. To support such a claim, we provide a systematic investigation of loss of plasticity across continual learning tasks using MNIST, CIFAR-10 and ImageNet. Our findings illustrate that loss of curvature directions coincides with loss of plasticity, while also showing that previous explanations are insufficient to explain loss of plasticity in all settings. Lastly, we show that regularizers which mitigate loss of plasticity also preserve curvature, motivating a simple distributional regularizer that proves to be effective across the problem settings we considered.
title Directions of Curvature as an Explanation for Loss of Plasticity
topic Machine Learning
url https://arxiv.org/abs/2312.00246