报告题目：Empower Intelligent Edge Computing: Opportunities and Challenges
The upheaval of AI applications has spawned a great many innovations across the whole stack, including the design of neural network (NN) accelerators, runtime frameworks, novel NN algorithms, and diverse scenario-specific solutions. The active use and interaction of these innovations on edge computing, which has strict requirements on latency and energy consumption, have raised unique challenges and opportunities compared to the cloud-side counterpart.This talk will center on these challenges and opportunities, and introduce the works of our team towards empowering intelligent edge computing. The talk will cover our projects on the edge NN-runtime implementation which fully consider the features of heterogeneous edge hardware to gain substantial speedup, the hardware-aware NN algorithm design to achieve real efficiency on diverse edge NN deployment platforms, and the end-to-end solutions for specific edge AI applications.
Ting Cao is a Senior Researcher in Microsoft Research Asia (MSRA). Her research interests include deep learning system and algorithm design, hardware/software co-design, high-level language implementation, management of heterogeneous hardware, and big data systems. She received her PhD from Research School of Computer Science, the Australian National University. Before joining MSRA, she had been working in the State Key Lab of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences, and then the Compiler and Computing Language Lab in 2012 Labs, Huawei Technologies, where she was engaged in the OpenArk (Fangzhou) compiler project. She had publications in a series of top venues, including ISCA, ASPLOS, PLDI, MobiCom, MobiSys, etc. She has also served in the program committee in several top conferences, such as PLDI, OOPSLA, VEE and ISMM.