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:
[log in to unmask]).
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
|