Two Routes to Scalable Credit Assignment without Weight Symmetry

Daniel Kunin, Aran Nayebi, Javier Sagastuy-Brena, Surya Ganguli, Jonathan Bloom, Daniel Yamins
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:5511-5521, 2020.

Abstract

The neural plausibility of backpropagation has long been disputed, primarily for its use of non-local weight transport — the biologically dubious requirement that one neuron instantaneously measure the synaptic weights of another. Until recently, attempts to create local learning rules that avoid weight transport have typically failed in the large-scale learning scenarios where backpropagation shines, e.g. ImageNet categorization with deep convolutional networks. Here, we investigate a recently proposed local learning rule that yields competitive performance with backpropagation and find that it is highly sensitive to metaparameter choices, requiring laborious tuning that does not transfer across network architecture. Our analysis indicates the underlying mathematical reason for this instability, allowing us to identify a more robust local learning rule that better transfers without metaparameter tuning. Nonetheless, we find a performance and stability gap between this local rule and backpropagation that widens with increasing model depth. We then investigate several non-local learning rules that relax the need for instantaneous weight transport into a more biologically-plausible "weight estimation" process, showing that these rules match state-of-the-art performance on deep networks and operate effectively in the presence of noisy updates. Taken together, our results suggest two routes towards the discovery of neural implementations for credit assignment without weight symmetry: further improvement of local rules so that they perform consistently across architectures and the identification of biological implementations for non-local learning mechanisms.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-kunin20a, title = {Two Routes to Scalable Credit Assignment without Weight Symmetry}, author = {Kunin, Daniel and Nayebi, Aran and Sagastuy-Brena, Javier and Ganguli, Surya and Bloom, Jonathan and Yamins, Daniel}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {5511--5521}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/kunin20a/kunin20a.pdf}, url = {https://proceedings.mlr.press/v119/kunin20a.html}, abstract = {The neural plausibility of backpropagation has long been disputed, primarily for its use of non-local weight transport — the biologically dubious requirement that one neuron instantaneously measure the synaptic weights of another. Until recently, attempts to create local learning rules that avoid weight transport have typically failed in the large-scale learning scenarios where backpropagation shines, e.g. ImageNet categorization with deep convolutional networks. Here, we investigate a recently proposed local learning rule that yields competitive performance with backpropagation and find that it is highly sensitive to metaparameter choices, requiring laborious tuning that does not transfer across network architecture. Our analysis indicates the underlying mathematical reason for this instability, allowing us to identify a more robust local learning rule that better transfers without metaparameter tuning. Nonetheless, we find a performance and stability gap between this local rule and backpropagation that widens with increasing model depth. We then investigate several non-local learning rules that relax the need for instantaneous weight transport into a more biologically-plausible "weight estimation" process, showing that these rules match state-of-the-art performance on deep networks and operate effectively in the presence of noisy updates. Taken together, our results suggest two routes towards the discovery of neural implementations for credit assignment without weight symmetry: further improvement of local rules so that they perform consistently across architectures and the identification of biological implementations for non-local learning mechanisms.} }
Endnote
%0 Conference Paper %T Two Routes to Scalable Credit Assignment without Weight Symmetry %A Daniel Kunin %A Aran Nayebi %A Javier Sagastuy-Brena %A Surya Ganguli %A Jonathan Bloom %A Daniel Yamins %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-kunin20a %I PMLR %P 5511--5521 %U https://proceedings.mlr.press/v119/kunin20a.html %V 119 %X The neural plausibility of backpropagation has long been disputed, primarily for its use of non-local weight transport — the biologically dubious requirement that one neuron instantaneously measure the synaptic weights of another. Until recently, attempts to create local learning rules that avoid weight transport have typically failed in the large-scale learning scenarios where backpropagation shines, e.g. ImageNet categorization with deep convolutional networks. Here, we investigate a recently proposed local learning rule that yields competitive performance with backpropagation and find that it is highly sensitive to metaparameter choices, requiring laborious tuning that does not transfer across network architecture. Our analysis indicates the underlying mathematical reason for this instability, allowing us to identify a more robust local learning rule that better transfers without metaparameter tuning. Nonetheless, we find a performance and stability gap between this local rule and backpropagation that widens with increasing model depth. We then investigate several non-local learning rules that relax the need for instantaneous weight transport into a more biologically-plausible "weight estimation" process, showing that these rules match state-of-the-art performance on deep networks and operate effectively in the presence of noisy updates. Taken together, our results suggest two routes towards the discovery of neural implementations for credit assignment without weight symmetry: further improvement of local rules so that they perform consistently across architectures and the identification of biological implementations for non-local learning mechanisms.
APA
Kunin, D., Nayebi, A., Sagastuy-Brena, J., Ganguli, S., Bloom, J. & Yamins, D.. (2020). Two Routes to Scalable Credit Assignment without Weight Symmetry. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:5511-5521 Available from https://proceedings.mlr.press/v119/kunin20a.html.

Related Material