Near-Tight Margin-Based Generalization Bounds for Support Vector Machines

Allan Grønlund, Lior Kamma, Kasper Green Larsen
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:3779-3788, 2020.

Abstract

Support Vector Machines (SVMs) are among the most fundamental tools for binary classification. In its simplest formulation, an SVM produces a hyperplane separating two classes of data using the largest possible margin to the data. The focus on maximizing the margin has been well motivated through numerous generalization bounds. In this paper, we revisit and improve the classic generalization bounds in terms of margins. Furthermore, we complement our new generalization bound by a nearly matching lower bound, thus almost settling the generalization performance of SVMs in terms of margins.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-gronlund20a, title = {Near-Tight Margin-Based Generalization Bounds for Support Vector Machines}, author = {Gr{\o}nlund, Allan and Kamma, Lior and Larsen, Kasper Green}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {3779--3788}, 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/gronlund20a/gronlund20a.pdf}, url = {http://proceedings.mlr.press/v119/gronlund20a.html}, abstract = {Support Vector Machines (SVMs) are among the most fundamental tools for binary classification. In its simplest formulation, an SVM produces a hyperplane separating two classes of data using the largest possible margin to the data. The focus on maximizing the margin has been well motivated through numerous generalization bounds. In this paper, we revisit and improve the classic generalization bounds in terms of margins. Furthermore, we complement our new generalization bound by a nearly matching lower bound, thus almost settling the generalization performance of SVMs in terms of margins.} }
Endnote
%0 Conference Paper %T Near-Tight Margin-Based Generalization Bounds for Support Vector Machines %A Allan Grønlund %A Lior Kamma %A Kasper Green Larsen %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-gronlund20a %I PMLR %P 3779--3788 %U http://proceedings.mlr.press/v119/gronlund20a.html %V 119 %X Support Vector Machines (SVMs) are among the most fundamental tools for binary classification. In its simplest formulation, an SVM produces a hyperplane separating two classes of data using the largest possible margin to the data. The focus on maximizing the margin has been well motivated through numerous generalization bounds. In this paper, we revisit and improve the classic generalization bounds in terms of margins. Furthermore, we complement our new generalization bound by a nearly matching lower bound, thus almost settling the generalization performance of SVMs in terms of margins.
APA
Grønlund, A., Kamma, L. & Larsen, K.G.. (2020). Near-Tight Margin-Based Generalization Bounds for Support Vector Machines. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:3779-3788 Available from http://proceedings.mlr.press/v119/gronlund20a.html.

Related Material