BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stationary Subspace Analysis - Frank Meinecke\, TU Berlin
DTSTART:20100216T140000Z
DTEND:20100216T150000Z
UID:TALK22845@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:*Abstract:* Multivariate non-stationary time series are encoun
 tered in many application domains\, ranging from climate research to neuro
 science. Often\, the non-stationarity is a problem for classical data anal
 ysis methods. In machine learning\, it might limit the generalization abil
 ity from training- to test set. So for accurate predictions\, the goal is 
 to control or eliminate the non-stationarity. On the other hand\, temporal
  changes in the dynamics can be of high interest to understand the analyze
 d system. Often they allow to find change points (i.e. data points where t
 he system switches between different states) that yield a temporal segment
 ation. Here\, the stationary part of the data is uninformative and only ma
 kes the segmentation problem harder. To tackle these problems\, we propose
  a novel technique\, Stationary Subspace Analysis (SSA)\, that decomposes 
 a multivariate timeseries into its stationary and non-stationary parts. Th
 e method is based on two assumptions: (a) the observed signals are linear 
 superpositions of stationary-- and non-stationary sources\; and (b) the no
 n-stationarity is measurable in the first two moments. We characterize the
  properties of the resulting algorithm and present results on simulated an
 d real data. 
LOCATION:Small public lecture room\, Microsoft Research Ltd\, 7 J J Thomso
 n Avenue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR