When: Thurs 26th of Nov, 1p AEDT
Where: This seminar will be presented online, RSVP here.
Speaker: Dr Tim Bailey
Title: Adventures in Estimation; applying linear tools to nonlinear problems
Abstract: This talk is about things I have learned over the past several years applying linear-Gaussian estimation methods to a variety of nonlinear problems. Some are very nonlinear, or have lots of bad data, or have sensor noise of unknown magnitude. I shall discuss: why using linear-Gaussian approximation gets hate but can be great; how to design forward-pass filters to avoid over-confidence and local minima; how to use residuals and iterative gate refinement to reject bad data; how to estimate noise scale to automate tuning; and various other topics of tangential relevance. If you enjoy a rambling technical talk with no flashy video or fun stuff, then pull up a chair, grab a beverage, and spend an hour with me. If your expectations are low, you might not be disappointed.
Bio: Tim Bailey joined Sydney University as an undergraduate Mechatronics student in 1993, and remained for 20 years as a PhD candidate and research fellow with the ACFR, before spending the past six years at NICTA and CSIRO. Now he is back.