Adaptive MCMC and Bayesian time-frequency analysis

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

• Richard Everitt, Dept of Statistics, University of Bristol
• Wednesday 03 February 2010, 14:00-15:00
• LR4, Engineering, Department of.

If you have a question about this talk, please contact Rachel Fogg.

The spectral density is an important quantity in time series analysis. Not only is it used in frequency domain analyses of signals, it is also relevant to Markov chain Monte Carlo (MCMC) since the integrated autocorrelation time (the spectral density evaluated at frequency zero) is related to the efficiency of MCMC estimators.

In this talk we introduce recursions for estimating the spectral density (and hence also the autocorrelation time) online. We then show how these recursions may be used for two different purposes:

1. We present an adaptive MCMC algorithm, in which a computer can use the online estimates of the autocorrelation time in order to automatically tune the MCMC algorithm; 2. We demonstrate how to perform an online Bayesian estimation of a time-frequency representation of a signal (using a particle filter) through use of online estimates of the Page spectrum of the signal.

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)
• 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
• LR4, Engineering, Department of
• 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.