A major challenge in deep learning is how to develop models which are compact, lightweight and power efficient so that they can be effectively deployed on devices that billions of users use like XR glasses, smart-phones, and tablets.
Prominent methods for achieving all these goals are developing efficient architectures via Neural Architecture Search, Network Pruning and Quantization (including Binary Networks).
Despite recent successes in all these areas, efficiency always comes at the cost of reduced accuracy. This PhD project will undertake fundamental research in the area-efficient Deep Learning for developing computationally efficient yet powerful models for perception and/or generation building upon prior work by Dr. Georgios Tzimiropoulos and Professor Ioannis Patras (the supervisors).