tinyML Talks: Software/Hardware Co-design for Tiny AI Systems

“Software/Hardware Co-design for Tiny AI Systems”

Yiran Chen
Chair
ACM SIGDA

The advancement of Artificial Intelligence (AI) and its swift deployment on resource-constrained tiny systems relies on both design quality and design efficiency of models. In this talk, we first introduce efficient AI models via hardware-friendly model compression and topology-aware Neural Architecture Search to optimize quality-efficiency trade-off on AI models. Then, we involve cross-optimization design and efficient distributed learning to brew swift and scalable AI systems with specialized hardware. Finally, we demonstrate the enhancement on quality-efficiency trade-off on alternative applications and scenarios, such as Electronic Design Automation (EDA) and Adversarial Machine Learning. Through these explorations, we present our vision on the future of the full stack of tiny AI solutions.

source