Timothy Sumer

Eamonn Keogh: "A Trillion here, a Trillion there: Scaling Time Series Data Mining to a Trillion Time Series" A video from the UC Berkeley Conference: From Data to Knowledge: Machine-Learning with Real-time and Streaming Applications (May 7-11, 2012). Abstract Eamonn Keogh (Computer Science and Engineering Dept., University of California, Riverside) In this talk I will argue the following claims. 1) Similarity search is the fundamental operation for mining time series data, and virtually any task, classification, clustering, rule finding, anomaly detection etc., can be efficiently and effectively solved once the similarity search problem is solved. 2) While there are dozens of alternative distance measures for similarity search, a 50-year old idea, Dynamic Time Warping (DTW) is exceptionally hard to beat. 3) DTWs often touted lethargy is no more. With four simple new ideas, we can exactly search billions of time series in a minute under DTW, using off-the-shelf comp