PhD, Hong Kong University of Science and Technology
Online Stochastic Optimisation for Large-Scale Machine Learning Problems in Big Data
Support vector machine (SVM) is considered as the standard technique for a wide range of data classification problems in many different fields, such as cancer diagnostic in bioinformatics, image classification, face recognition in computer vision and text categorisation in document processing. However, in a big data environment, computing the SVM classifier amounts to solve a large-scale optimisation which is mathematically complex and computationally expensive and the existing optimisation methods would not be fast enough. The main objective of this talk is to propose an accelerated stochastic optimisation method to solve SVMs for big data applications, which benefits from fast convergence, low complexity and easy implementation. The convergence analysis of the new method is theoretically investigated and also using some real-world data sets, the proposed algorithm is compared to the existing state of the art algorithms. Numerical results show that the proposed method significantly outperforms the existing schemes with orders of magnitude higher convergence rate.
Naeimeh Omidvar received her B.Sc. and M.Sc. both in Communication Systems Engineering from Sharif University of Technology, Tehran, Iran, in 2009 and 2011, respectively. She is currently pursuing her Ph.D. studies in the Hong Kong University of Science and Technology and Sharif University of Technology, under a double-degree Ph.D. program. Her research interests include multi-timescale stochastic optimisations, large-scale optimisation for big data, green networking, data networking algorithms and protocols, cross-layer radio resource management for future networks, and game theory. She is also involved in various industrial projects at Huawei-HKUST Joint Innovation Lab at HKUST, Hong Kong.
Watch video from here.