10:00 AM-12:00 PM
Room: Ballroom III
Nonstationarity is one of the most relevant open problems in time series analysis. For many systems, nonstationarity is an intrinsic feature representing the time variation of the system's dynamics. The latter point of view leads to powerful new concepts to detect, analyze and cope with nonstationarity. Some of them, such as cross-prediction errors and recurrence plots, rely on deterministic structure underlying the data. Instead of segmenting data sets into almost stationary phases, the implicit reconstruction of the values of drifting parameters can restore stationarity requirements for applications such as noise reduction and signal classification.
Nonlinear time series analysis has become a relevant tool for inverse problem. A new approach towards the problem of nonstationarity is of high relevance to broader applications of these methods, e.g. in the treatment of medical data, speech processing, and failure prediction in technical devices. In this minisyposium, the speakers will present relevant recent progress in this area which could stimulate a new philosophy, and consider certain types of nonstationarity as part of the (potentially nondeterministic) dynamics and treat the series as an entity..
Organizers: Holger Kantz
Max Planck Institute for the Physics of Complex Systems, Dresden, Germany
Thomas Schreiber
University of Wuppertal, Germany
LMH, 1/7/99, MMD, 2/9/99