COLLOQUIUM
             Computer Science Department, Boston University

Speaker: Spiros Papadimitriou
         Carnegie-Mellon University

Date:  Wednesday, March 30
Time:  9:00
Place: Room MCS 135, 111 Cummington Street 
       (for directions, see www.cs.bu.edu/colloquium)

Title: Parameter-Free Spatial and Stream Mining

Abstract:

Data mining is the extraction of useful information from large
collections of data.  Besides more traditional settings, in recently
emerging applications in stream processing, we cannot even hope to store
all the data.  This talk is about finding patterns in spatial and in
stream data and has two parts.
 
The first part focuses on outlier detection in multi-dimensional sets of
points.  Our method, LOCI, can quickly detect outliers and groups of
outliers.  In addition to previous methods, it provides an automatic,
data-dictated cut-off to determine whether a point is an outlier.  Also,
besides just an outlier score, it can provide a summary of information
about the data in the vicinity of each point.
 
In the second part of the talk, we focus on numerical stream data.
Given multiple streams, we present a method that can incrementally find
correlations and hidden variables, which summarize the key trends in the
entire stream collection.  It can do this quickly, with no buffering of
stream values, and does not require tuning any special parameters.  The
discovered trends can also be used to spot outliers and do efficient
forecasting.

Host: George Kollios