University Of Pretoria Computer Science Department

Nature Inspired Computing Optimization Group

This research groups focusses on taking analogies from nature to solve problems in the following areas:

Computer science

These include problems in computer security, e.g. network intrusion detection, grammatical inference, e.g. derivation of automata and grammars, automated programming, e.g. automated object-oriented programming, game playing, e.g. deriving game playing strategies for board games.

Operations research

These include both discrete and continuous combinatorial optimization problems. Discrete problems include timetabling and scheduling, vehicle routing, travelling salesman, packing problems and continuous problems such as function optimization.

Data science

Includes various real-world problems requiring the finding patterns in the data and/or classifying the data. The following techniques are currently being investigated to solve these problems:

Applications of genetic programming

Genetic programming is a variation of genetic algorithms which search a program space rather than a solution space. Research involves applying genetic programming to solve the different problems. In doing so this also leads to investigating various theoretical aspects to overcome any shortcomings of genetic programming in solving these problems.

Evolutionary algorithm hyper-heuristics

Hyper-heuristics aim to provide more generalized solutions to problems by exploring a heuristic space rather than a solution space. While a majority of the research has focused on using single point search techniques for selection hyper-heuristics, this research investigates a multipoint search, namely, genetic algorithms for this purpose. The use of genetic programming and variations thereof are studied for generation hyper-heuristics.

Automated design of machine learning and search techniques

There is a large scale initiative towards automating the design of machine learning and search techniques in aim to allow domain experts to have access to off the shelf tools which they did not need knowledge of to use in their particular domain to solve problems. This will also free the research of the time consuming tasking of design when implementing these techniques. This area examines various techniques, such as meta-GAs and hyper-heuristics for automated design.

Biologically-inspired methodologies

This area focuses on taking analogies from nature to solve the different problems. This has led to the derivation of the developmental approach which takes an analogy from cell biology to solve combinatorial optimization problems by working in the partial solution space. This technique is being developed further based of the development of multicellular organisms.
All content copyright © Department of Computer Science, School of IT, University of Pretoria, South Africa