Schwarz, J;
Jayakumar, SM;
Pascanu, R;
Latham, PE;
Teh, YW;
(2021)
Powerpropagation: A sparsity inducing weight reparameterisation.
In:
Proceedings of the Thirty-fifth Conference on Neural Information Processing Systems.
NeurIPS
(In press).
Preview |
Text
2110.00296v2.pdf - Accepted Version Download (1MB) | Preview |
Abstract
The training of sparse neural networks is becoming an increasingly important tool for reducing the computational footprint of models at training and evaluation, as well enabling the effective scaling up of models. Whereas much work over the years has been dedicated to specialised pruning techniques, little attention has been paid to the inherent effect of gradient based training on model sparsity. In this work, we introduce Powerpropagation, a new weight-parameterisation for neural networks that leads to inherently sparse models. Exploiting the behaviour of gradient descent, our method gives rise to weight updates exhibiting a “rich get richer” dynamic, leaving low-magnitude parameters largely unaffected by learning. Models trained in this manner exhibit similar performance, but have a distribution with markedly higher density at zero, allowing more parameters to be pruned safely. Powerpropagation is general, intuitive, cheap and straight-forward to implement and can readily be combined with various other techniques. To highlight its versatility, we explore it in two very different settings: Firstly, following a recent line of work, we investigate its effect on sparse training for resource-constrained settings. Here, we combine Powerpropagation with a traditional weight-pruning technique as well as recent state-of-the-art sparse-to-sparse algorithms, showing superior performance on the ImageNet benchmark. Secondly, we advocate the use of sparsity in overcoming catastrophic forgetting, where compressed representations allow accommodating a large number of tasks at fixed model capacity. In all cases our reparameterisation considerably increases the efficacy of the off-the-shelf methods.
Type: | Proceedings paper |
---|---|
Title: | Powerpropagation: A sparsity inducing weight reparameterisation |
Event: | Thirty-fifth Conference on Neural Information Processing Systems 2021 |
Open access status: | An open access version is available from UCL Discovery |
Publisher version: | https://neurips.cc/Conferences/2021/ |
Language: | English |
Additional information: | This version is the author accepted manuscript. For information on re-use, please refer to the publisher's terms and conditions. |
UCL classification: | UCL UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Life Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Life Sciences > Gatsby Computational Neurosci Unit |
URI: | https://discovery.ucl.ac.uk/id/eprint/10139743 |
1. | China | 2 |
2. | Germany | 2 |
3. | Russian Federation | 1 |
Archive Staff Only
View Item |