Bilkent University
Department of Computer Engineering


Multimodal Data Fusion and Online Statistical Inference


Yasin Ylmaz
University of Michigan

The advancing sensing and computing technologies enable collecting and processing, in real time, abundance of data in a variety of forms, i.e., big data. Considering the emerging concepts of Cyber-Physical Systems (CPS), Internet of Things (IoT), and intelligent systems such as autonomous vehicles, smart grid, smart home, and wearable technology, there is a rapidly growing need for big data analysis techniques for fast and accurate decision making. We already see applications of big data analytics in many areas such as healthcare, energy, finance, marketing, and transportation. Three commonly accepted fundamental challenges in big data analytics are (i) tackling large datasets, (ii) fusing multimodal data (i.e., different data types), and (iii) online statistical inference from streaming data. In this talk, I will present our efforts in both exploratory and confirmatory data analysis to address the above challenges. Specially, I will present a novel generative latent variable model called Multimodal Factor Analysis; a computationally efficient expectation-maximization (EM) algorithm to learn the model parameters; and results from two applications, namely event detection in Twitter, and online anomaly detection in nuclear fuel cycles. Finally, I will briefly mention our previous works on online statistical inference under privacy constraints and constraints on critical resources such as energy and communication bandwidth, with applications to wireless sensor networks, communication systems, and smart grid.

Bio: Yasin Yilmaz received the B.Sc., M.Sc., and Ph.D. degrees in Electrical Engineering from Middle East Technical University, Ankara, Turkey in 2008, Koc University, Istanbul, Turkey in 2010, and Columbia University, New York, NY in 2014, respectively. He is currently a postdoctoral research fellow at the University of Michigan, Ann Arbor. His research interests include big data analytics, machine learning, statistical signal processing, and their applications to cyber-physical security, intelligent systems, communication systems, energy systems, and social networks. He received the Collaborative Research Award from Columbia University in 2015 for his broad spectrum of collaborations, from statistics to nuclear engineering, to applied physics, to public and international affairs. His research has resulted in a number of papers in top-tier journals and conferences, invited talks, a book chapter, and a patent application.


DATE: 05 May 2016, Thursday @ 13:00