Institute for Mathematical Sciences                                        Programs & Activities

 

   
 

Online registration form
   


Enquiries

      General

      Scientific aspects


   

Workshop on Recent Advances in Bayesian Computation
(20 - 22 September 2010)

Organizing Committee · Visitors and Participants · Overview · Activities · Venue

 

 Organizing Committee

 

Chair

 

Members

  • Leontine Alkema (National University of Singapore)
  • Alex Cook (National University of Singapore)
  • Robert Kohn (University of New South Wales)

 

 Visitors and Participants

 

 

 Overview

 

Over the last 15 years there has been an explosion in the use of Bayesian methods in applied statistics. Due to advances in technology for data collection in many fields of science, engineering and the social sciences, applied statisticians increasingly have to deal with problems of combining data and information from different sources. This leads naturally to the use of richly structured hierarchical models, and advances in Bayesian computational methods have meant that a Bayesian approach is often the most easily implemented one for inference in such models. The purpose of this workshop is to bring together leading researchers in the area of Bayesian computational methods to discuss challenges and opportunities in the area, with a focus on dealing with large data sets.

 

The breakthrough technology that has allowed routine use of Bayesian methods in complex problems is Markov chain Monte Carlo (MCMC). Although MCMC has a long history of use in statistical physics its potential for general purpose Bayesian inference in statistics was only fully realized in the 1990's, starting with Gelfand and Smith (1990) and following some earlier applications in spatial statistics (Geman and Geman, 1984). New theoretical developments in MCMC are now appearing more slowly than previously but there is continuing widespread use of MCMC in applications. Although MCMC has allowed statisticians to handle complex models often with hundreds or thousands of parameters, one weakness is that MCMC is very computationally intensive, and for large data sets the use of MCMC methods may be completely infeasible. In short, while MCMC allows computation for complex models with small to moderate sized data sets, computation for complex models and very large data sets remains a challenge.

 

Many alternative strategies for computation with large datasets and complex models are being actively explored. These include clever methods for sampling large datasets, adaptive MCMC methods that attempt to improve on traditional MCMC approaches, advanced importance sampling techniques and deterministic methods such as variational approximation (developed actively within the machine learning community) and Laplace approximation. It is likely that advances in Bayesian computation for large data sets will involve at least to some extent imaginative combinations of the approaches discussed above. A key goal of the workshop is to look at the field of Bayesian computation very broadly and to bring together researchers from different areas to enable a cross-fertilization of ideas.

 

Gelfand, A. and Smith, A. (1990). Sampling based approaches to calculating marginal densities. J. American Statist. Assoc., 85, 398-409.

 

Geman, S. and Geman, D. (1984). Stochastic relaxation, Gibbs distributions and the Bayesian restoration of images. IEEE Trans. Pattern Anal. Mach. Intell., 6, 721-741.

 

 Activities

 

The workshop will be held over 3 days from 20-22 September 2010. There will be talks from approximately 10 overseas invited speakers and several local speakers. Although all the talks are by invitation, there will also be a poster session on the afternoon of 21 September which is open to all attendees.


Monday, 20 Sep 2010

09:00am - 09:20am

Registration

09:20am - 09:30am

Opening Remarks

Louis Chen, Institute for Mathematical Sciences
David Nott, National University of Singapore

 

Chair: Robert Kohn, University of New South Wales

09:30am - 10:15am

The r-inla.org project: an overview

HÃ¥vard Rue, Norwegian University of Science and Technology, Norway

10:15am - 10:45am

--- Coffee Break ---

10:45am - 11:30am

EM, variational bayes and expectation propagation

Mike Titterington, University of Glasgow, UK

11:30am - 01:30pm

--- Lunch ---

 

Chair: Chenlei Leng, National University of Singapore

01:30pm - 02:15pm

Skew-normal variational approximations for Bayesian inference

John Ormerod, University of Sydney, Australia

02:15pm - 03:00pm

Variational bayes for elaborate distributions

Matt Wand, University of Wollongong, Australia

03:00pm - 03:30pm

--- Coffee Break ---

 

Chair: Minh Ngoc Tran, National University of Singapore

03:30pm - 04:15pm

Large-scale Bayesian logistic regression

David Madigan, Columbia University, USA

Tuesday, 21 Sep 2010

09:15am - 09:30am

Registration

 

Chair: Leontine Alkema, National University of Singapore

09:30am - 10:15am

Bayesian hypothesis testing in latent variable models

Jun Yu, Singapore Management University

10:15am - 10:45am

--- Coffee Break ---

10:45am - 11:30am

Optimizing MCMC algorithms in high dimensions: a new perspective.

Natesh Pillai, Harvard University, USA

11:30am - 01:30pm

--- Lunch ---

 

Chair: Yanan Fan, University of New South Wales, Australia

01:30pm - 02:15pm

Bayesian computation on graphics cards

Chris Holmes, University of Oxford, UK

02:15pm - 03:00pm

Finite dimensional simulation methods for infinite dimensional posteriors

Jim Griffin, University of Kent, UK

03:00pm - 03:30pm

--- Coffee Break ---

 

Chair: Scott Sisson, University of New South Wales, Australia

03:30pm - 04:15pm

Variational Bayes for spatial data analysis

Clare McGrory, Queensland University of Technology, Australia

04:15pm - 05:00pm

Regression density estimation with variational methods and stochastic approximation

David Nott, National University of Singapore

Wednesday, 22 Sep 2010

09:15am - 09:30am

Registration

 

Chair: Alex Cook, National University of Singapore

09:30am - 10:15am

Recent advances in ABC (Approximate Bayesian Computation) methodology

Jean-Michel Marin, Université Montpellier 2, France

10:15am - 10:45am

--- Coffee Break ---

10:45am - 11:30am

Adaptive Monte Carlo sampling and model uncertainty

Merlise Clyde, Duke University, USA

11:30am - 01:30pm

--- Lunch ---

 

Chair: Siew Li Linda Tan, National University of Singapore

01:30pm - 02:15pm

The expected auxiliary variable method for Bayesian computation

Arnaud Doucet, University of British Columbia, Canada

02:15pm - 03:00pm

Flexible modeling of conditional distributions using smooth mixtures of asymmetric student t densities

Robert Kohn, University of New South Wales, Australia

03:00pm - 03:30pm

--- Coffee Break ---

 

Chair: Roman Carrasco, National University of Singapore

03:30pm - 04:15pm

Help! Fitting process-based models to infectious disease data, Bayesianly

Alex Cook, National University of Singapore

04:15pm - 05:00pm

Adaptive optimal scaling of Metropolis-Hastings algorithms

Scott Sisson, University of New South Wales, Australia



Students and researchers who are interested in attending these activities and who do not require financial aid are requested to complete the online registration form.

The following do not need to register:

  • Those invited to participate.
  • Those applying for financial support.


 Venue

 

 

Organizing Committee · Visitors and Participants · Overview · Activities · Venue

top
Best viewed with IE 7 and above