PhiMac Seminar - John Armstrong - Projecting SDEs on Submanifolds and the Optimal Gaussian Filter


Speaker:  John Armstrong, King's Colleg

Title:  Projecting SDEs on Submanifolds and the OptimalGaussian Filter 

Abstract:  We will ask what it means to optimallyapproximate an SDE in a space with an SDE on a submanifold. This is motivatedby attempts to approximate the solution to the non-linear filtering problemusing non-linear families of distributions such as the family of Gaussiandistributions. One approach that has been used in the past is to project thevector coefficients of the SDE written in Stratonovich form, but while this mayappear intuitive we will see that it is sub-optimal and will describealternative projection methods. As an application we will see how to write downoptimal Gaussian approximations to the non-linear filtering problem.

Location:  Hamilton Hall, Room 410

Go Back
McMaster University - Faculty of Science | Math & Stats