Title: From Chaos to Harmony: Making Sense of Massive Imperfect Data
Speaker: Ting Wang, Research Staff Member, IBM
Abstract: When analyzing massive data collected from many diverse sources, we often face chaotic information, i.e., observations that are imprecise, incomplete, or inconsistent. We typically rely on data redundancy to compensate for its imperfection. However, this simple “quantity-for-quality” approach performs poorly when (i) timeliness is required or (ii) data sources are interdependent.
In this talk, I will focus on two of my research efforts to tackle these limitations from data mining perspective. First, I will discuss how to exploit spatiotemporal correlation naturally existing in data to address the timeliness issue. This work promotes a powerful primitive to support a range of “finding needle in haystack” analysis.
The second part of the talk details how to uncover interdependence of data sources and how to separate bias incurred by this entanglement.
This work provides insights towards more reliable and effective ways of marshaling information from a multitude of entangled sources. I will finish with discussing my plan of contributing to the teaching in the department.
Speaker Bio: Ting Wang is currently a Research Staff Member at IBM Thomas J. Watson Research Center. His research invents new concepts and methods empowering large-scale data mining as well as bridges disciplinary boundaries for application of these advances to privacy, security, and trust issues. Ting is the recipient of a major innovation award for “Big Data Analytics for Cybersecurity” from IBM. Prior to joining IBM Research, Ting completed his Ph.D. in Computer Science at Georgia Institute of Technology.