Research

Our research is centred on resource-aware machine learning. We delve into areas including but not limited to efficient deep learning on mobile/embedded devices, efficient spatial-temporal data processing, as well as theories associated with resource-awareness and efficiency.

 

We aim to develop a methodical approach to resource-aware machine learning systems, considering four key aspects:

​

• Feasibility: We explore the boundaries of existing algorithms and hardware systems.

• Sustainability: We focus on the long-term deployment of systems, tying deeply into societal, energy, and environmental concerns.

• Scalability: Our approach is designed to be future-proof, scalable in multiple dimensions, including local computing and parallel and distributed deployments.

• Adaptability: We recognize the importance of efficient training for systems operating in changing environments.

 

Our research is dedicated to bridging the gap between the rapid evolution of deep learning algorithms and hardware computing power. For a detailed list of publications, please refer to my Google Scholar profile.