Edge

This thrust focuses on EdgeAI, a new paradigm meant to help achieving the large-scale deployment of AI/ML techniques on edge devices targeting Internet-of-Things (IoT) applications.

The research in this thrust aims at developing new learning models, algorithms, and prototypes that address important challenges of edge computing such as hardware- and energy-constrained IoT devices, distributed training and inference across multiple IoT devices, data security and privacy. This thrust core activities will help enable the seamless and strong integration between the cloud and the physical world.

Key Challenges

To enable rapid adoption of deep learning at the edge, there are a few major challenges that need to be addressed: hardware-constrained IoT devices (e.g., model compression to work with devices that have limited computational, memory, and power resources), data security and privacy (e.g., federated learning to enable on-device training), and network-aware deep learning (e.g., communication-aware model compression to enable distributed inference).

Synergies

This thrust will work synergistically with the Cloud and Applications thrusts to address the challenges involved in enabling the widespread deployment of deep learning on networked edge devices (e.g., resource scarcity, energy efficiency, security/privacy etc.), thus making the cloud-edge computing continuum a reality.

Edge Lead

Edge Faculty