November 3-4, 2006
We will consider the problem of identifying the most likely source
of a multivariate data point from among several multivariate
populations. The use of statistical depth functions for solving
this classification problem will be discussed. Statistical depth functions
provide a center-outward ordering of points in a multivariate data
cloud and hence can be considered to be multivariate analogues of ranks.
Specifically, classification through maximizing the estimated
transvariation probability of statistical depths is proposed. Considering
symmetric populations, it will be illustrated that these new
classification techniques provide lower misclassification error rates in
the case of heavy tailed distributions.
This is joint work with Nedret Billor, Asuman Turkmen and Sai Nudurupati.
Light propagation in coupled fiber arrays is described by a balanced of
diffraction and nonlinearity. At
high intensities, light is localized as a nonlinear mode propagating in a
few fibers. The imperfections
in the manufacturing of such fiber arrays account for multiplicative noise
in the governing equations.
Here we analyze how this noise affects the phenomenon of linear
(Anderson-like) and nonlinear localization.
The Aviation Systems Engineering Group at JHU/APL conducts systems
engineering and analysis to support the development and operational
employment of military aviation systems. In this endeavor technical
requirements and enabling technologies are identified that relate to
operational requirements and operational concepts. The group strives to
maintain expertise in air defense threat characterization and analyze
the survivability and effectiveness of current and future military
aviation systems. To this end we are involved in a wide array of
projects encompassing many technical disciplines.