CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1165
Title: Fast computation of the angular halfspace depth Authors:  Stanislav Nagy - Charles University (Czech Republic) [presenting]
Rainer Dyckerhoff - University of Cologne (Germany)
Petra Laketa - Charles University (Czech Republic)
Abstract: The angular halfspace depth is a nonparametric tool for the analysis of directional data. That depth was proposed already in 1987, but its widespread use has been hampered in practice by significant computational issues. We address these problems by considering a simple projection scheme that allows reducing the computation of the angular depth to the task of evaluating a variant of the usual halfspace depth in a linear space. Efficient algorithms for exact computation and approximation of the angular halfspace depth are developed.