Working Set Selection to Accelerate SVR Training

Pablo Rivas
Proceedings of 2nd Workshop on Diversity in Artificial Intelligence (AIDBEI), PMLR 142:35-38, 2021.

Abstract

With the increasing demand for robust and resilient machine learning models, support vector machines (SVMs) are regaining attention. One of the significant problems in SVMs is finding the support vectors as soon as possible during the optimization process. This paper describes a methodology to accelerate the training by making certain assumptions on the data and find the support vectors near the convex hull of every class group. Results suggest that the methodology can provide an advantage over traditional training for larger datasets with specific statistical properties. We focus on the particular case of support vector machines for regression.

Cite this Paper

BibTeX
@InProceedings{pmlr-v142-rivas21a, title = {Working Set Selection to Accelerate SVR Training}, author = {Rivas, Pablo}, booktitle = {Proceedings of 2nd Workshop on Diversity in Artificial Intelligence (AIDBEI)}, pages = {35--38}, year = {2021}, editor = {Lamba, Deepti and Hsu, William H.}, volume = {142}, series = {Proceedings of Machine Learning Research}, month = {09 Feb}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v142/rivas21a/rivas21a.pdf}, url = {https://proceedings.mlr.press/v142/rivas21a.html}, abstract = {With the increasing demand for robust and resilient machine learning models, support vector machines (SVMs) are regaining attention. One of the significant problems in SVMs is finding the support vectors as soon as possible during the optimization process. This paper describes a methodology to accelerate the training by making certain assumptions on the data and find the support vectors near the convex hull of every class group. Results suggest that the methodology can provide an advantage over traditional training for larger datasets with specific statistical properties. We focus on the particular case of support vector machines for regression.} }
Endnote
%0 Conference Paper %T Working Set Selection to Accelerate SVR Training %A Pablo Rivas %B Proceedings of 2nd Workshop on Diversity in Artificial Intelligence (AIDBEI) %C Proceedings of Machine Learning Research %D 2021 %E Deepti Lamba %E William H. Hsu %F pmlr-v142-rivas21a %I PMLR %P 35--38 %U https://proceedings.mlr.press/v142/rivas21a.html %V 142 %X With the increasing demand for robust and resilient machine learning models, support vector machines (SVMs) are regaining attention. One of the significant problems in SVMs is finding the support vectors as soon as possible during the optimization process. This paper describes a methodology to accelerate the training by making certain assumptions on the data and find the support vectors near the convex hull of every class group. Results suggest that the methodology can provide an advantage over traditional training for larger datasets with specific statistical properties. We focus on the particular case of support vector machines for regression.
APA
Rivas, P.. (2021). Working Set Selection to Accelerate SVR Training. Proceedings of 2nd Workshop on Diversity in Artificial Intelligence (AIDBEI), in Proceedings of Machine Learning Research 142:35-38 Available from https://proceedings.mlr.press/v142/rivas21a.html.

Related Material