An end-to-end approach for the verification problem: learning the right distance

Joao Monteiro, Isabela Albuquerque, Jahangir Alam, R Devon Hjelm, Tiago Falk
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:7022-7033, 2020.

Abstract

In this contribution, we augment the metric learning setting by introducing a parametric pseudo-distance, trained jointly with the encoder. Several interpretations are thus drawn for the learned distance-like model’s output. We first show it approximates a likelihood ratio which can be used for hypothesis tests, and that it further induces a large divergence across the joint distributions of pairs of examples from the same and from different classes. Evaluation is performed under the verification setting consisting of determining whether sets of examples belong to the same class, even if such classes are novel and were never presented to the model during training. Empirical evaluation shows such method defines an end-to-end approach for the verification problem, able to attain better performance than simple scorers such as those based on cosine similarity and further outperforming widely used downstream classifiers. We further observe training is much simplified under the proposed approach compared to metric learning with actual distances, requiring no complex scheme to harvest pairs of examples.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-monteiro20a, title = {An end-to-end approach for the verification problem: learning the right distance}, author = {Monteiro, Joao and Albuquerque, Isabela and Alam, Jahangir and Hjelm, R Devon and Falk, Tiago}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {7022--7033}, 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/monteiro20a/monteiro20a.pdf}, url = {https://proceedings.mlr.press/v119/monteiro20a.html}, abstract = {In this contribution, we augment the metric learning setting by introducing a parametric pseudo-distance, trained jointly with the encoder. Several interpretations are thus drawn for the learned distance-like model’s output. We first show it approximates a likelihood ratio which can be used for hypothesis tests, and that it further induces a large divergence across the joint distributions of pairs of examples from the same and from different classes. Evaluation is performed under the verification setting consisting of determining whether sets of examples belong to the same class, even if such classes are novel and were never presented to the model during training. Empirical evaluation shows such method defines an end-to-end approach for the verification problem, able to attain better performance than simple scorers such as those based on cosine similarity and further outperforming widely used downstream classifiers. We further observe training is much simplified under the proposed approach compared to metric learning with actual distances, requiring no complex scheme to harvest pairs of examples.} }
Endnote
%0 Conference Paper %T An end-to-end approach for the verification problem: learning the right distance %A Joao Monteiro %A Isabela Albuquerque %A Jahangir Alam %A R Devon Hjelm %A Tiago Falk %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-monteiro20a %I PMLR %P 7022--7033 %U https://proceedings.mlr.press/v119/monteiro20a.html %V 119 %X In this contribution, we augment the metric learning setting by introducing a parametric pseudo-distance, trained jointly with the encoder. Several interpretations are thus drawn for the learned distance-like model’s output. We first show it approximates a likelihood ratio which can be used for hypothesis tests, and that it further induces a large divergence across the joint distributions of pairs of examples from the same and from different classes. Evaluation is performed under the verification setting consisting of determining whether sets of examples belong to the same class, even if such classes are novel and were never presented to the model during training. Empirical evaluation shows such method defines an end-to-end approach for the verification problem, able to attain better performance than simple scorers such as those based on cosine similarity and further outperforming widely used downstream classifiers. We further observe training is much simplified under the proposed approach compared to metric learning with actual distances, requiring no complex scheme to harvest pairs of examples.
APA
Monteiro, J., Albuquerque, I., Alam, J., Hjelm, R.D. & Falk, T.. (2020). An end-to-end approach for the verification problem: learning the right distance. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:7022-7033 Available from https://proceedings.mlr.press/v119/monteiro20a.html.

Related Material