Importance Sampling with Particle Flows

Add to your list(s) Download to your calendar using vCal

• Dr Pete Bunch, Cambridge University
• Thursday 06 November 2014, 14:00-15:00
• BN3-05 (Sigproc Meeting Room), 3rd Floor of CUED.

If you have a question about this talk, please contact Fredrik Lindsten.

The Bayesian approach to inference is based upon calculating posterior probability distributions. When these are not analytically tractable, we can instead estimate lots of useful properties by sampling. However, drawing samples from a posterior distribution is also a challenging task. This talk will focus on two methods for achieving this, importance sampling and particle flow sampling.

Importance sampling is a well-established, well-used, well-studied algorithm, in which samples are drawn from an importance distribution, and then weighted so as to represent the posterior. Particle flow sampling instead uses a differential equation to move samples through the state space to positions that represent the posterior.

We discuss the relative advantages and disadvantages of these two methods, and show how they may be usefully combined to achieve better inference with a class of nonlinear Gaussian models.

This talk is part of the Signal Processing and Communications Lab Seminars series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• BN3-05 (Sigproc Meeting Room), 3rd Floor of CUED
• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge University Engineering Department Talks
• Cambridge talks
• Centre for Smart Infrastructure & Construction
• Chris Davis' list
• Computational Continuum Mechanics Group Seminars
• Featured lists
• Information Engineering Division seminar list
• Interested Talks
• School of Technology
• Signal Processing and Communications Lab Seminars
• Trust & Technology Initiative - interesting events
• bld31
• ndk22's list
• ob366-ai4er
• rp587

Note that ex-directory lists are not shown.