BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1304
DTSTAMP:20230921T105046Z
SUMMARY:DANSE - Data-driven Non-linear State Estimation in Unsupervised Lea
 rning
DESCRIPTION:Speaker: Saikat Chatterjee (KTH Royal Institute of Technology\,
  Stockholm\, Sweden)\n\nAbstract: \nWe address the tasks of Bayesian state
  estimation and forecasting for a model-free process in an unsupervised le
 arning setup. In the seminar\, we discuss our new method called DANSE -- D
 ata-driven Nonlinear State Estimation method. DANSE provides a closed-form
  posterior of the state of the model-free process\, given linear measureme
 nts of the state. In addition it provides a closed-form posterior for fore
 casting. We show how data-driven recurrent neural networks (RNNs) are used
  in the DANSE to provide closed-form prior of the state and posterior. The
  training of DANSE\, mainly learning the parameters of RNN\, is executed i
 n unsupervised learning approach. In the unsupervised learning\, we have a
 ccess to a training dataset comprising of only a set of measurement data t
 rajectories\, but we do not have any access to the state trajectories. The
 refore\, DANSE does not have access to state information in training data 
 and can not use supervised learning. Using simulated linear and non-linear
  process models (Lorenz attractor and Chen attractor)\, we evaluate the un
 supervised learning-based DANSE. We show that the proposed DANSE\, without
  knowledge of the process model and without supervised learning\, provides
  a competitive performance against model-driven methods\, such as Kalman f
 ilter (KF)\, extended KF (EKF) and unscented KF (UKF)\, and a recently pro
 posed hybrid method called KalmanNet.\n
URL:https://www.tcs.tifr.res.in/web/events/1304
DTSTART;TZID=Asia/Kolkata:20230616T160000
DTEND;TZID=Asia/Kolkata:20230616T170000
LOCATION:A201
END:VEVENT
END:VCALENDAR