Sparse and Structured Hopfield Networks

Saul José Rodrigues Dos Santos, Vlad Niculae, Daniel C Mcnamee, Andre Martins
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:43368-43388, 2024.

Abstract

Modern Hopfield networks have enjoyed recent interest due to their connection to attention in transformers. Our paper provides a unified framework for sparse Hopfield networks by establishing a link with Fenchel-Young losses. The result is a new family of Hopfield-Fenchel-Young energies whose update rules are end-to-end differentiable sparse transformations. We reveal a connection between loss margins, sparsity, and exact memory retrieval. We further extend this framework to structured Hopfield networks via the SparseMAP transformation, which can retrieve pattern associations instead of a single pattern. Experiments on multiple instance learning and text rationalization demonstrate the usefulness of our approach.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-santos24a, title = {Sparse and Structured Hopfield Networks}, author = {Santos, Saul Jos\'{e} Rodrigues Dos and Niculae, Vlad and Mcnamee, Daniel C and Martins, Andre}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {43368--43388}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/santos24a/santos24a.pdf}, url = {https://proceedings.mlr.press/v235/santos24a.html}, abstract = {Modern Hopfield networks have enjoyed recent interest due to their connection to attention in transformers. Our paper provides a unified framework for sparse Hopfield networks by establishing a link with Fenchel-Young losses. The result is a new family of Hopfield-Fenchel-Young energies whose update rules are end-to-end differentiable sparse transformations. We reveal a connection between loss margins, sparsity, and exact memory retrieval. We further extend this framework to structured Hopfield networks via the SparseMAP transformation, which can retrieve pattern associations instead of a single pattern. Experiments on multiple instance learning and text rationalization demonstrate the usefulness of our approach.} }
Endnote
%0 Conference Paper %T Sparse and Structured Hopfield Networks %A Saul José Rodrigues Dos Santos %A Vlad Niculae %A Daniel C Mcnamee %A Andre Martins %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-santos24a %I PMLR %P 43368--43388 %U https://proceedings.mlr.press/v235/santos24a.html %V 235 %X Modern Hopfield networks have enjoyed recent interest due to their connection to attention in transformers. Our paper provides a unified framework for sparse Hopfield networks by establishing a link with Fenchel-Young losses. The result is a new family of Hopfield-Fenchel-Young energies whose update rules are end-to-end differentiable sparse transformations. We reveal a connection between loss margins, sparsity, and exact memory retrieval. We further extend this framework to structured Hopfield networks via the SparseMAP transformation, which can retrieve pattern associations instead of a single pattern. Experiments on multiple instance learning and text rationalization demonstrate the usefulness of our approach.
APA
Santos, S.J.R.D., Niculae, V., Mcnamee, D.C. & Martins, A.. (2024). Sparse and Structured Hopfield Networks. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:43368-43388 Available from https://proceedings.mlr.press/v235/santos24a.html.

Related Material