Modern Bayesian Inference in Implicit Probabilistic Models

Back to search results
Recent advances in generative adversarial networks have sparked tremendous excitement about the more general area of implicit probabilistic models. These models are only defined via simulations from an unknown (implied) distribution and provide a much more flexible data-modelling approach than traditional prescribed probabilistic models. However, their generality comes at the expense of extremely difficult inference challenges.

This project will develop new methods for inference in implicit probabilistic models and is expected to have a significant impact in areas such as evolutionary biology, ecology, high-energy physics, financial models and portfolio optimisation, where the combination of simulators with data-driven approaches is ubiquitous.

Website

The ideal candidate should have completed a degree in a quantitative discipline such as Computer Science, Statistics, Physics, Electronic Engineering or related disciplines. He/she should have a strong Mathematics and Statistics background with excellent programming skills and proficiency in programming languages such as Python, Matlab, R or C++. He/She must be a highly-motivated student, proactive, curious and enthusiastic about scientific research.

Supervisory team
Edwin
Bonilla

Engineering
Computer Science and Engineering
Robert
Kohn

Business School
Economics
Scott
Sisson

Science
Mathematics & Statistics
Register to Apply
Non-UNSW staff/students must Register to Apply
e.bonilla@unsw.edu.au