Abstract: Machine learning algorithms often solve optimization problems. When model(s) are constructed to ﬁt the data, they are trained by solving a core optimization problem. This helps to learn parameters with respect to the selected loss functions and possibly some regularization functions. In the process of model selection and validation, the core optimization problem may be solved many times. This entwining of machine learning and optimization allows researchers to embrace advances in mathematical programming to study the speed, accuracy and robustness of machine learning algorithms. In this tutorial, we will investigate how popular machine learning algorithms including classiﬁcation, regression and clustering can be posed as optimization problems and solved using using linear, quadratic, conic, and semideﬁnite programs. Illustrative examples in the R programming language using the MOSEK package will be presented.
Biography: Haimonti Dutta is an Assistant Professor in the Department of Management Science and Systems (MSS), School of Management, University at Buffalo, New York. Prior to her current appointment she served as an Associate Research Scientist at the Center for Computational Learning Systems (CCLS) at Columbia University, NY where she headed the Scalable Analytics Research Group. She is affiliated to the Institute for Data Sciences (IDSE) at Columbia University and served as an adjunct assistant professor at the Indraprastha Institute of Information Technology (IIIT-Delhi) between 2014-2016. Her research focuses on mining big data. She is interested in distributed and parallel algorithms for machine learning and distributed optimization. The federal government (including the National Science Foundation, National Endowment of Humanities), private foundations (Epilepsy Research Foundation) and industry partners (including Amazon Web Services, EMC, Mathworks Inc, The Consolidated Edison Company of New York) have generously funded her research. haimonti@buﬀalo.edu