Investigation into the Applications of Quantum Machine Intelligence in Smart Living

Project ID: SST_2_3

With its current growth rate, Quantum Computing is likely to become a major paradigm over the next decade, particularly in the field of computational intelligence. Although the real quantum computers are still not publicly available, theoretical research on quantum algorithms either for future implementations or for the existing simulators is growing. However, there is a research gap in exploring the practical applications of “quantum machine intelligence” in the field of intelligent environments including smart cities and smart homes.

Considering the needs for computational powers for such applications, this project will investigate how machine learning algorithms can be enhanced, adapted or inspired by quantum computing and be applied in future smart environments.

Smart living largely relies on employing machine learning algorithms, e.g., to classify human activities or the living environment attributes, face/gesture recognition, logistics and route planning, and to provide necessary actions in the environment. A known challenge in machine learning is the heuristic nature of the underlying parameters/hyperparameters’ search methods, due to their very large search spaces, which usually leads to non-optimal solutions with a lower performance than necessary. In deep learning for instance, the classical methods may not converge to optimum attribute sets or network topologies, whereas there could be a better solution if different attributes would have been discovered. This is particularly the case in the lack of enough training data in the smart living scenarios.

Enters quantum computing with built-in parallelism, there is a potential that machine learning search spaces, even very large ones, can be much more effectively searched for the optimum solutions, than by classical heuristic search. For example, the possible research questions can be (but not limited to:)

- Can quantum circuits be employed for hyperparameter optimisation of a regression model used in fall detection or in optimising logistics route planning in smart cities?

- Can the power of quantum annealing in optimally solve polynomial constrained/unconstrained binary optimisation problems be used to optimise the performance of deep learning systems used in assisted living?

We have already conducted some initial research in Quantum Computing, as well as extensive research on assisted living, so that this project can use the existing research infrastructure on both areas of quantum computing and assistive living to comparatively investigate quantum machine intelligence benefits in an unexplored application area. The department has access to experimental quantum computers, large-scale FPGA cluster, High-Power Computing facilities, and collaboration with Microsoft Azure Quantum which can collectively support the candidate to development of this project. The research will be in collaboration with the supervisory team, NTU research group and external collaborators.

The successful candidates should have a good machine learning background and be preferably familiar with quantum computing. The candidate should have a good research and communication skills for publishing the research outputs and for presenting them in conferences or meetings.

Supervisory Team:

Dr Amir Pourabdollah (Senior Lecturer and AI course leader, Computer Science)

Dr Colin Wilmott (Senior Lecturer, Mathematics)

Dr Ismahane Cheheb (Early Career Lecturer, Computer Science)

Prof Ahmad Lotfi (HoD and CIA research group leader, Computer Science)