JOB: Postdoctoral Research Associate, Statistical Laboratory, Cambridge

[rbgr...@gmail.com]

Statistical Laboratory
Centre for Mathematical Sciences
University of Cambridge
Wilberforce Road
Cambridge

Postdoctoral Research Associateship in Markov Chain Monte Carlo
Methods and Applications

Applications are invited for the post of Postdoctoral Research
Associate in Markov chain Monte Carlo (MCMC) methodology and
application. The position may be taken up as early as 1 September
2008. The research associate will work with Dr. Robert B. Gramacy on
an EPSRC-funded project on trans-dimensional Markov chain simulation
for both Bayesian and classical model determination. The position may
be occupied for up to 24 months. The salary will be in the range
£25888-£33780.

The research entails developing new Monte Carlo inference methodology
for problems requiring model selection and averaging. The specific
project(s) will be negotiated with -- and tuned to the skills of the
successful applicant. Motivation may come from Dr. Gramacy's recent
work with models and data from statistical/quantitative finance,
parsimonious regression and covariance estimation in missing data
problems, partition models for spatial data, non-stationary and non-
linear regression and classification, and the sequential design of
experiments. It is anticipated that the work will require extensions
and novel applications of several of the following Monte Carlo
inference techniques: reversible jump (RJ)MCMC, simulated annealing
(SA) and tempering (ST), Markov coupled MCMC (MC3), particle filters
[i.e., sequential importance sampling (SIS)], including the scope for
parallelisation of the above algorithms.

Broadly, the aim of this project is to develop high powered algorithms
which are automatic, and adaptive, in that they enable non-experts to
fit intricate models. Ideally, the methodologies developed would
exploit modern multi-node and multi-core computing architectures
without the need for tedious pilot tuning and calibration. We may
consider the spectrum of situations where data is scarce to data sets
so massive that they cannot fit into computer memory. Such an
ambitious remit requires flexible models and implementations, while
remaining focused on portability and user friendliness.

Applicants should have expertise in applying existing MCMC
methodology, Bayesian hierarchical modelling, an interest in algorithm
design and a strong background in statistical computing (with high
proficiency in a statistical package like R and/or Matlab and a
compiled language like C/C++ or Fortran). There is a preference to
appoint a Research Associate with experience implementing one or more
of the following: RJMCMC, SA, ST, MC3, particle filters.

Applicants should send a full CV, including a list of publications and
a description of previous research experience, and completed PD18
(http://www.admin.cam.ac.uk/offices/hr/forms/pd18/) CV cover form
(parts I and III only) including the names of TWO references to Julia
Blackwell, Statistical Laboratory, DPMMS, Centre for Mathematical
Sciences, Wilberforce Road, Cambridge CB3 0WB (email:
vaca...@statslab.cam.ac.uk).

Applicants must ask their referees to write directly to Julia
Blackwell by the closing date of 1 August 2008.

The University is committed to equality of opportunity