Meta-Learning Bidirectional Update Rules

Mark Sandler, Max Vladymyrov, Andrey Zhmoginov, Nolan Miller, Tom Madams, Andrew Jackson, Blaise Agüera Y Arcas
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:9288-9300, 2021.

Abstract

In this paper, we introduce a new type of generalized neural network where neurons and synapses maintain multiple states. We show that classical gradient-based backpropagation in neural networks can be seen as a special case of a two-state network where one state is used for activations and another for gradients, with update rules derived from the chain rule. In our generalized framework, networks have neither explicit notion of nor ever receive gradients. The synapses and neurons are updated using a bidirectional Hebb-style update rule parameterized by a shared low-dimensional "genome". We show that such genomes can be meta-learned from scratch, using either conventional optimization techniques, or evolutionary strategies, such as CMA-ES. Resulting update rules generalize to unseen tasks and train faster than gradient descent based optimizers for several standard computer vision and synthetic tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-sandler21a, title = {Meta-Learning Bidirectional Update Rules}, author = {Sandler, Mark and Vladymyrov, Max and Zhmoginov, Andrey and Miller, Nolan and Madams, Tom and Jackson, Andrew and Arcas, Blaise Ag{\"u}era Y}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {9288--9300}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/sandler21a/sandler21a.pdf}, url = {https://proceedings.mlr.press/v139/sandler21a.html}, abstract = {In this paper, we introduce a new type of generalized neural network where neurons and synapses maintain multiple states. We show that classical gradient-based backpropagation in neural networks can be seen as a special case of a two-state network where one state is used for activations and another for gradients, with update rules derived from the chain rule. In our generalized framework, networks have neither explicit notion of nor ever receive gradients. The synapses and neurons are updated using a bidirectional Hebb-style update rule parameterized by a shared low-dimensional "genome". We show that such genomes can be meta-learned from scratch, using either conventional optimization techniques, or evolutionary strategies, such as CMA-ES. Resulting update rules generalize to unseen tasks and train faster than gradient descent based optimizers for several standard computer vision and synthetic tasks.} }
Endnote
%0 Conference Paper %T Meta-Learning Bidirectional Update Rules %A Mark Sandler %A Max Vladymyrov %A Andrey Zhmoginov %A Nolan Miller %A Tom Madams %A Andrew Jackson %A Blaise Agüera Y Arcas %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-sandler21a %I PMLR %P 9288--9300 %U https://proceedings.mlr.press/v139/sandler21a.html %V 139 %X In this paper, we introduce a new type of generalized neural network where neurons and synapses maintain multiple states. We show that classical gradient-based backpropagation in neural networks can be seen as a special case of a two-state network where one state is used for activations and another for gradients, with update rules derived from the chain rule. In our generalized framework, networks have neither explicit notion of nor ever receive gradients. The synapses and neurons are updated using a bidirectional Hebb-style update rule parameterized by a shared low-dimensional "genome". We show that such genomes can be meta-learned from scratch, using either conventional optimization techniques, or evolutionary strategies, such as CMA-ES. Resulting update rules generalize to unseen tasks and train faster than gradient descent based optimizers for several standard computer vision and synthetic tasks.
APA
Sandler, M., Vladymyrov, M., Zhmoginov, A., Miller, N., Madams, T., Jackson, A. & Arcas, B.A.Y.. (2021). Meta-Learning Bidirectional Update Rules. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:9288-9300 Available from https://proceedings.mlr.press/v139/sandler21a.html.

Related Material