Page 60 - University of Pretoria Research Review 2017

58
Algorithms – finding the set of optimal trade-offs
Mardé Helbig, Department of Computer Science
The work of the Computational Intelligence Research Group (CIRG) in the Department of Computer Science focuses on computational intelligence and nature- inspired algorithms, which include swarm intelligence, evolutionary computation and neural networks. The research of Dr Mardé Helbig focuses on using these algorithms to solve a type of optimisation problem referred to as dynamic multi-objective optimisation problems.
Where objectives and/or constraints change over
time, and where improving on one objective results in worsening at least one of the other objectives, there can be no single solution. Algorithms, and especially nature-inspired algorithms, help find the set of optimal trade-off solutions. This type of algorithm typically uses a population of entities to search for solutions and delivers multiple possible solutions after a single run. The algorithm also does not need gradient information to find the solutions. The nature-inspired elements
of the algorithm evolve the solutions over time. Once the solutions are available, the decision-maker is
presented with the set of trade-off solutions from which to select one solution, based on her or his expert knowledge.
Recently, CIRG researchers have developed approaches where
Dr Mardé Helbig is one of the recipients of the UP Exceptional Young Researcher Award, and in 2017 was selected as a member of the executive committee of the South African Young Academy of Science (SAYAS).
Unsplash (Ray Hennessy)
a decision-maker guides the algorithm’s search by incorporating her or his knowledge during the search process. They have applied algorithms to the inverse kinematics problem of an animated character, for example, by specifying a position where a hand or
foot should be placed and then calculating the angles of all the joints to bring the hand or foot as close as possible to the specified position. This process can also be applied to the field of robotics, while algorithms have been developed to determine when to buy or sell foreign currency, for example.
The next step is to find suitable ways to visualise the set of trade-off solutions, especially when
the problem has more than three objectives. Helbig has published on visualising trade-offs with
Complex problems in real-world contexts are dynamic and invariably have conflicting objectives which means there can be no single solution and trade-offs are necessary. The same is the case with optimisation problems where either an objective function or constraint can vary over time.