Date: Wed, December 5, 14:30-
Place: Room As311, As Block, IIS, The University of Tokyo

Invited Speaker: Prof. Danilo P. Mandic (Imperial College London)

Title: Complex Valued Nonlinear Forecasting Using Augmented Statistics

Abstract:
In engineering applications, complex valued signals are either complex
by design (communications) or they are made complex by convenience of
representation (directional signals, vector fields, sensor arrays). One
phenomenon which benefits from its complex valued representation is
complex wind field. This talk will adress nonlinear predictive modelling
of wind field and will propose forecasting models which take into
account the dependence between the speed and direction components. To
produce enhanced statistical models of the intermittent (and hence
non-Gaussian) wind signal, some new developments in complex statitics
will be employed. A class of "augmented" algorithms will be introduced
and the links between the circularity and "properness" of a random
complex variable will be highlighted. Further, surrogate data testing
for appropriateness of complex valued representations will be
introduced. Simulations on a range of wind regimes will be presented to
illustrate the benefits of complex valued representation of wind field.