About the Project
Genetic algorithms are used in a huge number of diverse applications spanning pop music, clinical decision support, big data and automated design. Originally proposed by Turing in 1950, the genetic algorithm was first used by Holland in the 1960’s to investigate evolution but since this time the approach has since been successfully adapted to solve a range of problems. As the application of genetic algorithms has broadened, improvements and additions to the available methodologies are required to solve increasingly diverse and complex problems. Most of the top genetic algorithms are derived from two distinct methodologies with adaptations for specific problem types, but neglect current evolutionary theory. Concurrently contemporary evolutionary biology is undergoing impassioned debate into how epigenetic factors including learned behaviours, social memes and subconscious physiological cues modify inheritance across generations.
Recently a new methodology has been developed at the University of Southampton using a mechanism inspired by multi-level selection theory. The new methodology retains the use of individual selection, which all of the current genetic algorithms use, but adds selection at the collective level that increases the performance of the current methodologies on a wide range of problems. The aim of this research will be to extend the Multi-Level Selection Genetic Algorithm, MLSGA, to give further increases in performance. The focus will be on inspiring improvements to this from state-of-the-art evolutionary theory and the biological sciences resulting in a powerful general solver that will be tested on a range of problems. To test the generality and performance a range of tests will be used including standard optimisation benchmarking sets, boat hull design, composite structures, Atari 2600 games, genetic programming and scheduling problems.
The following areas will form an initial focus:
• the evolution of trait co-variation, including modularity;
• collective reproduction;
• benchmarking the performance on real-world applications.
It is proposed mechanisms that speed the rate of evolution in the natural world, like multi-level selection, will increase the speed of the genetic algorithm. We will continue to extend the new methodology using inspiration from epigenetics and evolutionary theory. This will provide a boost to the already excellent performance and provide evidence of the potential for the method.
Key Facts:
Entry requirements: first or upper second-class degree or equivalent
Assessment: Nine month and 18 month reports, viva voce and thesis examination
Start date: typically September
Applying: www.southampton.ac.uk/postgraduate/pgstudy/howdoiapplypg.html
If you wish to discuss any details of the project informally, please contact Adam Sobey, Fluid Structure Interactions research group, Email: [Email Address Removed], Tel: +44 (0) 2380 59 7773.
Funding Notes
This project is in competition with others for funding; the projects which receive the best applicants will be awarded a full studentship. This 3 year studentship covers home-rate tuition fees and provides an annual tax-free stipend at the standard EPSRC rate, which is £14,777 for 2018/19.
The funding is only available to UK citizens or EU citizens who have been resident in the UK for at least 3 years prior to the start of the studentship and not mainly for the purpose of receiving full-time education. For further guidance on funding, please contact [Email Address Removed]
Continue with Facebook
