"Events and Patterns in Noisy Sequences"
Tingjian Ge, UMass
Wednesday, October 30, 2013 at 12:00 Noon
Room 506 (CIT 5th Floor)
Sequence data is prevalent: from texts to ECG signals and time series, from smartphone and social network data to highway sensory data. In a practical setting, such data is produced at a high rate by unreliable devices, is often communicated through noisy channels (e.g., wireless networks), or is derived using incomplete knowledge based on raw data (e.g., trajectory mapping and forecasting). Consequently, noise and incompleteness in data are the norm rather than the exception. Some queries are sensitive to the noise or uncertainty in data - event pattern matching is such a type of query.
In this talk, I will describe two lines of work that we have recently done on efficient online complex event pattern matching in noisy streams: (1) subsequence matching with added negation semantics and window constraints, where errors of each element in the sequence are independently modeled, and (2) the more general regular expression matching with negation and window semantics, where arbitrary error correlation models can be assumed. While (1) is more efficient, (2) is more powerful. Both are fast enough to be online in real time for all the practical applications that we are aware of.
Tingjian Ge is an assistant professor in the Computer Science Department of the University of Massachusetts at Lowell. He received a Ph.D. from Brown University in 2009. Prior to that, he got his Bachelor's and Master's degrees from Tsinghua University and UC Davis, respectively, and worked in database companies Informix and IBM. From 2009 to 2011 he worked as an assistant professor at the University of Kentucky. His research areas are in data management, with topics including noisy and uncertain data, and data security and privacy. He is a recipient of the NSF CAREER Award in 2012. He serves as a Program Committee member in major database and data mining conferences such as ICDE, VLDB, and ICDM.
Host: Ugur Cetintemel