Algorithms are shaping the way we live. At its most simple, an algorithm is a step-by-step process for solving a problem. Some are everyday problems, others are more complex.
Let’s say you want to buy a car and you have two objectives in mind: comfort and affordability. The more comfortable a vehicle, the higher the cost, and therein lies the conflict between the two goals. Here’s a trickier example: how can you cost electricity in a way that guides responsible behaviour on the part of users, yet still make a profit? In a manufacturing plant, how can you drive up profits while also meeting clients’ tight deadlines? This is where algorithms can help.
When Dr Mardé Helbig, senior lecturer in the department of computer science at the University of Pretoria, found that little work had been done on solving problems with conflicting objectives that change over time, known as dynamic multi-objective optimisation problems (DMOOPs), she began to focus on solving DMOOPs using vector-evaluated particle swarm optimisation. These are algorithms which are simulated or inspired by biological behaviours of animal or birds and have been used to find the optimal solution to a given problem.
“Many real world optimisation problems are dynamic,” says Helbig. “They have more than one objective, with at least two of those objectives in conflict with one another, and at least one objective or constraint changing over time,” she says. “Research in this area can be applied to optimising the treatment of water based on what it’s going to be used for: the scheduling of jobs at a production plant or the routing of vehicles, for example.”
The main goals of her research were to develop an algorithm that can solve DMOOPs efficiently, work out how to measure whether an algorithm can solve DMOOPs competently, and to set benchmarks for their development and evaluation.
The benchmark suite characteristics, performance measures and approach to compare the performance of various algorithms when solving DMOOPs has provided the DMOO community with a platform for the evaluation and standardisation of newly proposed algorithms.
“It’s about a new way to compare algorithms that determines which one is better, and finding ways to adapt them when changes occur, which they are bound to,” she says of her ground-breaking research that led to her election as a member of the South African Young Academy of Science (Sayas) in 2017 and as a member of the executive committee of Sayas in 2017