Abstract: The proliferation of edge devices and the gigantic amount of data they generate are distributed everywhere. Such distributed data fuel the intelligence at the edge where data reside. In this talk, I will present my works on how to enable intelligence on edge devices by leveraging the power of deep learning.

 

First, I will present my work on designing a task-agnostic privacy-respecting data crowdsourcing framework for learning a feature extractor that can hide user-specified private information from intermediate representations while retaining the high utility of extracted features. Those features can be safely aggregated to train deep neural works for any learning tasks.

 

Second, I will shift from the centralized setting to the distributed setting for the collaborative learning on edge devices. In particular, I will present my work on designing a personalized federated learning framework that can jointly improve communication and computation efficiency.

 

I will also outline future research directions for building billion-scale networked and trustworthy intelligent ecosystem, such as developing ambient intelligent applications, designing scalable and adaptive machine learning algorithms, intelligently employing heterogeneous hardware resources, etc.

Short Bio: Ang Li is a Ph.D. candidate in the Department of Electrical and Computer Engineering at Duke University working Prof. Yiran Chen. His research interests lie in the intersection of machine learning and edge computing, with a focus on building large-scale networked and trustworthy intelligent systems to solve practical problems in a collaborative, scalable, secure, and ubiquitous manner. He received the Best Student Paper Award in KDD’20. His research is also applied to commercial products by companies.