Computer Science Distinguished Lecture
| Date/Time: | Thursday, 27 Feb 2014 at 3:40 pm |
|---|---|
| Location: | 2245 Coover Hall |
| Cost: | Free |
| Phone: | 515-294-6516 |
| Channel: | College of Liberal Arts and Sciences |
| Categories: | Lectures |
| Actions: | Download iCal/vCal | Email Reminder |
Bio
Eamonn Keogh is a full professor at the University of California Riverside. His research interests include data mining and artificial intelligence, with applications to medicine and entomology. He is one of only two people (Along with Christos Faloutsos) to have won at least one best paper award in SIGMOD, SIGKDD, ICDM and SDM, the four main venues for data mining research.
Abstract
Time series data is pervasive, produced by virtually every human endeavor, including medicine, industry and entertainment. Because of its ubiquity, the data mining community has devoted significant energy over the last two decades to techniques for extracting useful and actionable knowledge from such data.
In this talk I will review progress in time series data mining, and offers insights into future challenges and how they might be solved. As an example of progress, I will explain how the paradoxical idea that: the clustering of time series requires us to ignore most of the data, has enabled us to build significantly better classifiers for activity recognition and medical monitoring.
As an example of future directions for research, I will outline my arguments that Minimum Description Length may allow us to rank and compare models of different sizes, lengths sampling rates etc. This is currently an unsolved problem, and arguably the bottleneck in meaningful time series data mining.
I will illustrate my ideas with examples from entomology, activity/gesture recognition and medicine.