Learning from nature to create the future
Team work has helped make Victoria University a global hub for research into artificial intelligence, says Mengjie Zhang.
Artificial intelligence (AI) may seem like technology of the future, but you’re already using it on a daily basis. Whether you’re looking at your smartphone, making a bank transaction, playing a video game or driving your car, AI helped make your task possible.
Victoria University of Wellington is a global hub for research into AI, which involves machines mimicking human abilities such as learning and problem solving.
Professor of Computer Science Mengjie Zhang is a leading international AI researcher. Based in the School of Engineering and Computer Science, he is a member of Victoria’s Artificial Intelligence Group and heads the interdisciplinary Evolutionary Computation Research Group, which is the largest of its type in Australasia and in the top five in the world in terms of its representation in major journals and conference publications in recent years.
Originally from China, Professor Zhang has witnessed a rapid rise in interest in the application of AI since he arrived in New Zealand in 2000.
AI has “the very cool idea” of mimicking humans’ thinking, ideas, behaviour and learning ability, he says.
“A long-term dream in computing has been for machines to do things that are useful for humans. This could include robots that mimic human behaviour to do dangerous things or postal recognition systems that can recognise handwriting people may not be able to.
“Ten years ago, AI was not so useful. Now it can increasingly be applied to the real world, in areas ranging from immunoanalysis [lab tests that use antibodies or antigens to test for specific molecules] to Antarctic research.
“AI is still seen as a hard area to work in, but I tell my students anything new—data mining, big data, data science, cyber security or the internet of things—needs AI techniques.”
Professor Zhang’s primary area of research is evolutionary computation, which is in the computational intelligence side of AI. The term covers a range of problem-solving techniques based on the theories of biological evolution, including genetic algorithms, genetic programming, particle swarm optimisation, different evolution and learning classifier systems, and evolutionary multi-objective optimisation.
Evolutionary computation techniques are used to tackle complex problems that have too many variables for traditional computer algorithms to solve, that people have very little prior knowledge of, that are in dynamic and uncertain environments, or where the approach to solving the problem is difficult to understand.
The areas of evolutionary computation Professor Zhang is particularly interested in are:
- genetic programming, which is inspired by Darwinian natural selection, gene theory and automatic programming
- particle swarm optimisation, which is inspired by the behaviour of birds flying or fish swimming from one place to another.
“Evolutionary computation borrows from biological science, but using computer algorithms,” says Professor Zhang.
He also researches three other important fields:
- data mining/big data and machine learning, particularly big dimensionality reduction and evolutionary feature selection
- intelligent and evolutionary computer vision and image analysis
- evolutionary scheduling and combinatorial optimisation.
Professor Zhang and his team play a leading role in international research into the use of computers to duplicate the abilities of human vision. They use evolutionary algorithms to select and classify large data sets of images faster than the human eye would be able to and generate interesting features and solutions that humans can understand better than those many other algorithms generate.
Computer vision can be used to make it easier to carry out imaging tasks such as detecting tumours, analysing medical data and recognising faces in digital video images.
“Millions of pixels can be identified in an image of a human face. Not all those pixels are useful, so we try to identify the regions of most interest, such as the eyes, mouth and nose. With mathematics, we can combine these areas of interest and construct a small number of high-level but very informative features to carry out facial recognition,” says Professor Zhang.
He says dimensionality reduction and feature selection is a particularly fast-moving field of research, using machine learning to reduce the number of input features/variables being considered by automatically identifying and removing irrelevant and redundant features and constructing a small number of high-level informative features from a large number of low-level ones.
Evolutionary scheduling and combinatorial optimisation is another field where Victoria plays a leading role. This field has a large number of real-world applications, including resource planning and allocation.
“Computer hardware advances and the internet have made it possible to solve things we couldn’t have begun to solve 10 years ago,” says Professor Zhang.
Since 2005, he has been awarded four Marsden Fund grants as the sole principal investigator from Government funding managed by the Royal Society of New Zealand. His research contributes to Victoria’s ‘Spearheading our digital futures’ area of academic distinctiveness.
Professor Zhang’s many responsibilities include Associate Dean (Research and Innovation) for the Faculty of Engineering. In 2016 he was appointed Chair of the IEEE (Institute of Electrical and Electronics Engineers) Emerging Technologies Technical Committee, which identifies and nurtures new directions in technology, and he is editor or associate editor of more than 10 major international journals in his field.
He says a favourite part of his job is collaborating with colleagues and bringing people together.
“Most of the areas I research need a team to do the work. It would be very hard for a single person, and almost impossible to make breakthroughs,” he says. “Competition is good, but collaboration and cooperation are also critical to success.”
Professor Zhang says the team’s success would not have been possible without the excellent technical and administrative support it receives from the School of Engineering and Computer Science at Victoria, which has helped the University become a recognised hub for AI research—particularly in the fast growing areas of evolutionary computation, optimisation and learning.
“This is a great research environment, which is why Victoria attracts so many excellent PhD students and visiting researchers. New Zealand may be far away from anywhere, but people always want to come here.”