A survey of robust optimization based machine learning with special reference to support vector machines

Abstract

This paper gives an overview of developmentsin the field of robust optimization in machine learning (ML) in general and Support Vector Machine (SVM)/Support Vector Regression (SVR) models in particular. This survey comprises of researches in which robustness is sought against uncertainty. This uncertainty is in the values of parameters of the given model or it can be in the data. In this work, we have discussed how robust optimization has entered in the field of machine learning. Here, we investigate the contribution of various researchers in dealing with different types of uncertainties arising in the problem of maximizing or minimizing the objective function. We deal with the variants of SVM/SVR in more detail, although we have also covered supervised, unsupervised ML and various other aspects of ML. Also, we present an extensive study of research carried out in the applications of robust optimization in other areas like energy and power systems, networking, transportation etc. as well. © 2019, Springer-Verlag GmbH Germany, part of Springer Nature.