Stochastic Layer-Wise Precision in Deep Neural Networks

Griffin Lacey, Graham W. Taylor, Shawki Areibi
Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, PMLR R16:662-671, 2018.

Abstract

Low precision weights, activations, and gradi- ents have been proposed as a way to improve the computational efficiency and memory foot- print of deep neural networks. Recently, low precision networks have even shown to be more robust to adversarial attacks. How- ever, typical implementations of low precision DNNs use uniform precision across all lay- ers of the network. In this work, we explore whether a heterogeneous allocation of preci- sion across a network leads to improved per- formance, and introduce a learning scheme where a DNN stochastically explores multi- ple precision configurations through learning. This permits a network to learn an optimal pre- cision configuration. We show on convolu- tional neural networks trained on MNIST and ILSVRC12 that even though these nets learn a uniform or near-uniform allocation strat- egy respectively, stochastic precision leads to a favourable regularization effect improving generalization.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR16-lacey18a, title = {Stochastic Layer-Wise Precision in Deep Neural Networks}, author = {Lacey, Griffin and Taylor, Graham W. and Areibi, Shawki}, booktitle = {Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence}, pages = {662--671}, year = {2018}, editor = {Globerson, Amir and Silva, Ricardo}, volume = {R16}, series = {Proceedings of Machine Learning Research}, month = {06--10 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r16/main/assets/lacey18a/lacey18a.pdf}, url = {https://proceedings.mlr.press/r16/lacey18a.html}, abstract = {Low precision weights, activations, and gradi- ents have been proposed as a way to improve the computational efficiency and memory foot- print of deep neural networks. Recently, low precision networks have even shown to be more robust to adversarial attacks. How- ever, typical implementations of low precision DNNs use uniform precision across all lay- ers of the network. In this work, we explore whether a heterogeneous allocation of preci- sion across a network leads to improved per- formance, and introduce a learning scheme where a DNN stochastically explores multi- ple precision configurations through learning. This permits a network to learn an optimal pre- cision configuration. We show on convolu- tional neural networks trained on MNIST and ILSVRC12 that even though these nets learn a uniform or near-uniform allocation strat- egy respectively, stochastic precision leads to a favourable regularization effect improving generalization.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Stochastic Layer-Wise Precision in Deep Neural Networks %A Griffin Lacey %A Graham W. Taylor %A Shawki Areibi %B Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2018 %E Amir Globerson %E Ricardo Silva %F pmlr-vR16-lacey18a %I PMLR %P 662--671 %U https://proceedings.mlr.press/r16/lacey18a.html %V R16 %X Low precision weights, activations, and gradi- ents have been proposed as a way to improve the computational efficiency and memory foot- print of deep neural networks. Recently, low precision networks have even shown to be more robust to adversarial attacks. How- ever, typical implementations of low precision DNNs use uniform precision across all lay- ers of the network. In this work, we explore whether a heterogeneous allocation of preci- sion across a network leads to improved per- formance, and introduce a learning scheme where a DNN stochastically explores multi- ple precision configurations through learning. This permits a network to learn an optimal pre- cision configuration. We show on convolu- tional neural networks trained on MNIST and ILSVRC12 that even though these nets learn a uniform or near-uniform allocation strat- egy respectively, stochastic precision leads to a favourable regularization effect improving generalization. %Z Reissued by PMLR on 04 October 2026.
APA
Lacey, G., Taylor, G.W. & Areibi, S.. (2018). Stochastic Layer-Wise Precision in Deep Neural Networks. Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R16:662-671 Available from https://proceedings.mlr.press/r16/lacey18a.html. Reissued by PMLR on 04 October 2026.

Related Material