Location: Streatham Campus, University of Exeter, EX4 4QJ
We are inviting applications for a College of Engineering, Mathematical and Physical Sciences funded PhD studentship in Machine Learning and Optimisation.
Evolutionary algorithms (EAs) use processes inspired by biological evolution to solve optimisation problems. However, they require many evaluations of the function being optimised. In recent years EAs have been combined with machine learning algorithms to automatically learn approximations to the expensive function. These surrogates that emulate the expensive problem are relatively cheap to evaluate and can be used to guide the evolutionary search.
As part of an EPSRC-funded project on using evolutionary algorithms to optimise designs using computational fluid dynamics, this PhD project will focus on investigating and developing new machine learning methods for learning the surrogate model. It will focus on using novel loss functions as alternatives to the traditional mean squared error. The project will also explore the interaction between the evolutionary search, Bayesian optimisation and surrogate models. The results will be applicable across the wide range of evolutionary optimisers using surrogate modelling.
You will join a team of researchers working on surrogate modelling in a department with a strong reputation for evolutionary algorithms and machine learning. The project will blend computational and theoretical work and you will have to opportunity to try out your algorithms on a range of real-world problems.
Contact for Informal enquiries: : Professor Richard Everson ([email protected]
, tel: 01392 724065), Dr Jonathan Fieldsend ([email protected]
; tel: 01392 722090) or Dr Gavin Tabor ([email protected]
, tel: 01392 723662)
Applicants should have or expect to achieve at least a 2:1 honours degree, or equivalent, in Computer Science, Mathematics, or an aligned subject. It would be beneficial for applicants to have had experience of one or more of: machine learning, nature inspired computing, high performance computing.
The closing date for applications is midnight on 26 February 2016.