Thesis Topic Details

Topic ID:
460
Title:
Modelling and simulation for evolution and ecology
Supervisor:
Mark Tanaka
Research Area:
Bioinformatics
Associated Staff
Assessor:
Mike Bain
Topic Details
Status:
Active
Type:
Research
Programs:
BIOM BINF
Group Suitable:
No
Industrial:
No
Pre-requisites:
--
Description:
The rise of evolutionary computation and agent-based modelling has opened new pathways to study the evolutionary and ecological bases of complex animal behaviours such as foraging, communication, and learning. There are several potential projects available for honours work, including:

* Finding solutions to an evolutionary game theory problem of great complexity using genetic algorithms. Game theory is used to model the logical underpinnings to animal communication, such as when signalling systems should be reliable ('honest'), and when signal production should be costly (handicapping). However, game theory models of any useful complexity are too difficult to solve analytically, and so computational methods are required. Previous work has applied genetic algorithms (GAs) to the problem of conventional signalling in animal communication; conventional signalling arises when signals carry no cost except the response of the receiver; an animal that signals that it is strong may find itself in a fight it cannot win, which is a cost imposed by the receiver. The current project aims to extend this work by solving a much more difficult conventional signalling problem based on a system of frog calling communication using GAs.

* As part of an overseas collaboration, we are investigating the learning system of an agricultural pest, the Cabbage White butterfly Pieries rapae. These butterflies show genetic correlations between learning, attention, and movement, such that dispersive genotypes search a narrow range of potential hosts (plants to reproduce on), while less dispersive genotypes sample much more broadly among potential hosts. Less dispersive genotypes are also better learners, with larger areas of neural tissue dedicated to learning and sensory regions. Individual-based models (parameterised by field data collected on the model species) will be used to understand the evolutionary bases of these correlations as a function of environmental and strategic pressures.

* Modelling co-evolutionary pressures between parasitoid wasps and their aphid prey. Wasps must solve the problem of optimal travel time between the plants on which aphids live (with travel time affecting the sort of prey that they accept as part of their diet), while aphid colonies can respond to wasp choices by manipulating their spatial distribution. Previous work has looked at one side of this question in isolation (wasp choice when aphid behaviour is held fixed); the aim of this project is to model this system with an individual-based model coupled to co-evolving GAs to shed light on the general principles of predator patch choice and prey response.

* One of the ways that animal cognitive learning processes have been modelled is through the use of learning rules, mathematical descriptions of the value that animals should place upon behavioural options as a function of the reward to those options. Past work has searched for evolutionarily stable learning rules (rules that, if adopted by a population, bring individuals to the optimal behaviour in a way that cannot be invaded by mutant rules), but this work has focused on the few already published rules. Such a restricted set of candidates represents a tiny portion of the possible space of rules, and limits our ability to conclude that these rules are, in fact, evolutionarily stable. This project aims to leverage the use of automated program discovery techniques from the genetic programming toolkit to search for new learning rules that may be evolutionarily stable by more thoroughly traversing the search space; the outcome of this project may shed light on fundamental principles of learning in animal cognition.

Students interested in these projects should have strong programming skills; statistical knowledge and experience in analysis would be a plus, but is not necessary. Techniques used may include evolutionary computation and individual-based modelling, model visualisation (3D graphics), and large dataset manipulation and analysis.
Comments:
Dr Steven Hamblin, who is a research associate in Tanaka's group, will be the day-to-day supervisor for this project.
Past Student Reports
 
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.