My research themes

Main direction: Decentralise social computing and mobile computing

All the mainstream online social networking (OSN) service providers have a cloud-based infrastructure that is designed with a logically centralized architecture controlled by a single authority, i.e., the OSN service provider. With massive user data controlled by the provider, users are susceptible to loss of control over ownership of their data and privacy, and users may receive false social information due to deceptive advertisements and unfair censorship.

My research mainly concentrates on the design of the architecture for decentralized OSN (DOSN) which is created through participation by a set of autonomous OSN users collaborating with each other without a centralized repository. The social data are stored in local devices and controlled by end-users rather than OSN providers. The key challenge I am working on is how to create efficient device-to-device communication protocols and overlay networks to harness the resources in hundreds of millions of end devices without centralised servers.

Subordinate direction: Efficient AI for Edge computing

The IoT revolution leads the development of the next generation of computing and communications. The IoT edge computing aims to utilize computing processing power on the edge devices, provide interoperable networks and communication protocols and ultimately achieve greater computational flexibility and availability. However, due to the resource constraints at the IoT end devices, it is extremely challenging to deploy the state-of-the-art computation-intensive machine learning and deep learning algorithms up and running on the edge.

Efficient AI and machine learning in edge computing environments is another research direction. I specifically look at the incorporation of deep learning into a fully decentralised environment, supporting real-time artificial intelligence (AI) on resource-constrained edge devices. Network representation learning and knowledge representation learning for heterogamous IoT big data are the key research threads. The research aims at enabling computational expensive AI algorithms trained in end-to-end devices to support timely and accurate decision-making on the edge.