Attentive Group Equivariant Convolutional Networks

David Romero, Erik Bekkers, Jakub Tomczak, Mark Hoogendoorn
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:8188-8199, 2020.

Abstract

Although group convolutional networks are able to learn powerful representations based on symmetry patterns, they lack explicit means to learn meaningful relationships among them (e.g., relative positions and poses). In this paper, we present attentive group equivariant convolutions, a generalization of the group convolution, in which attention is applied during the course of convolution to accentuate meaningful symmetry combinations and suppress non-plausible, misleading ones. We indicate that prior work on visual attention can be described as special cases of our proposed framework and show empirically that our attentive group equivariant convolutional networks consistently outperform conventional group convolutional networks on benchmark image datasets. Simultaneously, we provide interpretability to the learned concepts through the visualization of equivariant attention maps.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-romero20a, title = {Attentive Group Equivariant Convolutional Networks}, author = {Romero, David and Bekkers, Erik and Tomczak, Jakub and Hoogendoorn, Mark}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {8188--8199}, 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/romero20a/romero20a.pdf}, url = {http://proceedings.mlr.press/v119/romero20a.html}, abstract = {Although group convolutional networks are able to learn powerful representations based on symmetry patterns, they lack explicit means to learn meaningful relationships among them (e.g., relative positions and poses). In this paper, we present attentive group equivariant convolutions, a generalization of the group convolution, in which attention is applied during the course of convolution to accentuate meaningful symmetry combinations and suppress non-plausible, misleading ones. We indicate that prior work on visual attention can be described as special cases of our proposed framework and show empirically that our attentive group equivariant convolutional networks consistently outperform conventional group convolutional networks on benchmark image datasets. Simultaneously, we provide interpretability to the learned concepts through the visualization of equivariant attention maps.} }
Endnote
%0 Conference Paper %T Attentive Group Equivariant Convolutional Networks %A David Romero %A Erik Bekkers %A Jakub Tomczak %A Mark Hoogendoorn %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-romero20a %I PMLR %P 8188--8199 %U http://proceedings.mlr.press/v119/romero20a.html %V 119 %X Although group convolutional networks are able to learn powerful representations based on symmetry patterns, they lack explicit means to learn meaningful relationships among them (e.g., relative positions and poses). In this paper, we present attentive group equivariant convolutions, a generalization of the group convolution, in which attention is applied during the course of convolution to accentuate meaningful symmetry combinations and suppress non-plausible, misleading ones. We indicate that prior work on visual attention can be described as special cases of our proposed framework and show empirically that our attentive group equivariant convolutional networks consistently outperform conventional group convolutional networks on benchmark image datasets. Simultaneously, we provide interpretability to the learned concepts through the visualization of equivariant attention maps.
APA
Romero, D., Bekkers, E., Tomczak, J. & Hoogendoorn, M.. (2020). Attentive Group Equivariant Convolutional Networks. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:8188-8199 Available from http://proceedings.mlr.press/v119/romero20a.html.

Related Material