Year: 2010
Type: other
Abstract: We investigate the application of kernel density \nestimators to pattern-recognition problems. These estimators \nhave a number of attractive properties for data analysis in \npattern recognition, but the particular characteristics of patternrecognition \nproblems also place some non-trivial requirements on \nkernel density estimation – especially on the algorithm used to \ncompute bandwidths. We introduce a new algorithm for variable \nbandwidth estimation, investigate some of its properties, and \nshow that it performs competitively on a wide range of tasks, \nparticularly in spaces of high dimensionality.
Author: Etienne Barnard
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2790709351