Browse/search for people

Mr Andi Wang

Mr Andi Wang

Mr Andi Wang
MMath, BA

Senior Research Associate

Area of research

Computational Statistical Inference for Engineering and Security

Office GA.07
Fry Building,
Woodland Road, BS8 1UG
(See a map)

Summary

I am a post doc working with Christophe Andrieu and Anthony Lee, as part of the CoSInES project. I started this role in October 2019.

My research interests include

  • Monte Carlo methods, in particular continuous-time methods.
  • Adaptive methods.
  • Stochastic processes.
  • Quasi-stationarity.

I completed my PhD at Oxford, 2015-2019, on the OxWaSP program, under the supervision of Prof. David Steinsaltz and Prof. Gareth Roberts. My thesis was titled "The theory of regeneration and killing in continuous-time Monte Carlo sampling".

Prior to that I did my undergraduate and master's degree in Cambridge, 2011-2015, the Mathematical Tripos (including Part III). I was awarded the Thomas Bond Sprague Prize in Part III.

Biography

See also my personal page.

I am a post doc working with Christophe Andrieu and Anthony Lee, as part of the CoSInES project. I started this role in October 2019.

I completed my PhD at Oxford, 2015-2019, on the OxWaSP program, under the supervision of Prof. David Steinsaltz and Prof. Gareth Roberts. My thesis was titled "The theory of regeneration and killing in continuous-time Monte Carlo sampling". My research concerned several topics, all centered around continuous-time Monte Carlo methods.
 
We started off studying quasi-stationary Monte Carlo methods, building on the pioneering work of Pollock et al. (2016). Quasi-stationarity concerns the limiting behaviour of killed (reducible) Markov processes. I studied the application of quasi-stationary theory to Monte Carlo inference problems, which was published as Wang et al. (2019). Secondly, we studied an alternative method of simulating from quasi-stationary distributions, based on reinforced processes (akin to the classical Polya urn); see our work.
 
Finally, utilising a similar toolbox we considered a new continuous-time method for Monte Carlo sampling, based on introducing regenerations (in continuous time) into a given Markov process, in such a way as to enforce stationarity of a given target. The regenerative framework simplifies mathematical analysis and allows for straightforward parallelization; see our preprint.

Prior to that I did my undergraduate and master's degree in Cambridge 2011-2015, the Mathematical Tripos. I was a member of St. John's college. I wrote my Part III essay under the supervision of Prof. Richard Samworth, titled Nonparametric Inference Under Shape Constraints.

Keywords

  • Monte Carlo methods
  • Nonreversible processes
  • Quasi-stationarity

Memberships

Organisations

Research themes

Recent publications

View complete publications list in the University of Bristol publications system

Edit this profile If you are Mr Andi Wang, you can edit this page. Login required.

PDF versionDownload PDF