Researcher Highlights
The nature-inspired design of machines

Dr Nasimul Noman is using nature as a blueprint for improving machine autonomy. It’s a field of study known as evolutionary machine learning and it has the potential to solve many real-world problems.

According to Dr Nasimul Noman, optimisation is part of daily life. Motorists want cars that are fuel efficient, environmentally friendly and stylish; investors seek to maximise the return on their financial portfolio while minimising risk; medical professionals need a range of drugs with optimum dosages for treatment that elicit little to no side effects.

Each problem, Nasimul states, has traditionally been solved using different optimisation algorithms. In this age of machine learning, machine learning algorithms need to be optimised from different perspectives, and to make machines smarter, smarter optimisation techniques must be used.

His research is in evolutionary machine learning, a subfield in which evolutionary algorithms interplay with machine learning. Specifically, Nasimul develops computational algorithms for designing optimal and more proficient machine learning algorithms, and for solving some of our real-world optimisation problems.

“In the field of machine learning, optimisation plays a key role in designing efficient and effective algorithms for extracting valuable information through analysis of data,” explains Nasimul.

Nasimul believes it’s possible to exploit many of nature’s processes in instilling robust and intelligent behaviours in computational algorithms. Using the natural world as his reference point, he hopes computers will one day design and synthesise intelligent systems automatically without human intervention—in other words, learn and improve on their own.

“My long-term goal is to develop computational frameworks for evolving robust and resilient intelligent systems through computational evolution.”

A leaf out of nature’s book

Like others in his field, Nasimul believes that nature is the supreme optimiser. He was introduced to the field of evolutionary algorithms—computational programs that mimic natural evolution—during his masters at the University of Dhaka in Bangladesh.

Subsequent research for his PhD afforded a valuable placement at one of Tokyo University’s laboratories studying computational intelligence that imitates clever, problem-solving techniques found in nature.

“Since my PhD, I have been developing various nature-inspired algorithms to optimise machine learning models and to solve different optimisation problems in life.

“I truly believe we can greatly benefit by understanding and applying the principles used in different spheres of nature for solving various complex problems and machine intelligence is not an exception.”

World-proofing algorithms

When it comes to solving the problems of our world, not only must solutions be optimal, they must withstand the challenges and disruptions that inevitably occur. For example, some minor distortions can never deceive human eyes in identifying an object in an image but can easily trick a machine.

Designing systems that are robust and resilient in automated decision-making is also important from a security perspective. Finding such designs can be very costly, but as part of his research, Nasimul is working towards developing more effective and efficient meta-algorithms that deliver optimal, robust algorithmic solutions. Already his work has had real-life applications.

His developed algorithms have been used in different fields, such as in machine learning for creating optimal deep neural networks, for feature selection and designing classifiers; in systems biology for designing combination therapies for disease, for automatic design and reconstruction of genetic and reaction networks; and in engineering for optimal scheduling of machines and for economic load distribution of power systems.

Shaping the future

Solving problems that benefit others is Nasimul’s greatest motivation for continuing his research.

Several useful applications for machine-learning algorithms already exist—from modelling spam filters to drug designs. New iterations, he stresses, must be more robust and secure—and he’s working to make them so. In the meantime, when Nasimul reflects on his research, he feels inspired.

“Working in a research area that can help shape our future makes me feel excited and proud. I feel committed and motivated because I believe I can draw inspiration from nature in solving problems that are important to us.”