|
Activity Number:
|
553
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305437 |
|
Title:
|
Multivariate Data Adaptive Compression and Density Estimation
|
|
Author(s):
|
Edward J. Wegman*+ and Roger Shores
|
|
Companies:
|
George Mason University and U.S. Census Bureau
|
|
Address:
|
MS 6A2, Center for Computational Data Sciences, Fairfax, VA, 22030,
|
|
Keywords:
|
nonparametric density estimation ; data compression ; delauney triangularization ; upper and lower bounds ; voronoi tiling
|
|
Abstract:
|
High-dimensional data compression and density estimation often depend on binning. Standard binning usually involves the use of hyper-rectangular bins. This type of binning has two problems: 1) the number of bins grows exponentially with dimension and 2) with fixed size bins and fixed data set size, some bins may be empty while other bins may oversmooth the data. We propose a data adaptive binning procedure which addresses both issues. In particular we discuss bounds on the number of bins as a function of the number of tessellating points and the dimension.
|
