Title:
Joint Online Parameter Estimation and Optimal Sensor Placement for Continuous Time State Space Models
Abstract:
In the talk we will discuss how to formulate this problem as a bilevel optimisation problem, and propose a solution in the form of a continuous-time, two-timescale, stochastic gradient descent algorithm. Under suitable conditions on the latent signal, the filter, and the filter derivatives, we can establish almost sure convergence of the online parameter estimates and optimal sensor placements to the stationary points of the asymptotic log-likelihood and asymptotic filter covariance, respectively. This is achieved using a more general result on for two-timescale stochastic gradient descent algorithms in continuous time under Markovian noise. We also present numerical examples for the partially observed Beneš equation and a partially observed stochastic advection-diffusion equation.
This is joint work with Louis Sharrock (from the University of Bristol).
Biography:
Worked as a researcher in Cambridge (Signal Processing and Control groups), here at imperial EEE control (2010-12), UCL Stats (2012-13) and joined imperial Stats section in Maths in 2013.
Zoom Meeting Details
- Link: https://imperial-ac-uk.zoom.us/j/96747382309?pwd=RXI0UFRpY0dTZlZVd0FXYzZQTVltUT09
- Meeting ID: 967 4738 2309
- Passcode: f2!xuS