Doctoral student Haoran Sun is a recipient of the best student paper contest at the 2018 Asilomar Conference on Signals, Systems, and Computers, for the paper “Distributed Non-Convex First-Order Optimization and Information Processing: Lower Complexity Bounds and Rate Optimal Algorithms.” A prestigious prize among signal processing conferences, Haoran’s paper competed against 86 student papers, and came in at third place, among the 8 finalists.
Haoran’s work addresses decentralization of information processing. The next decade will see an estimated 50 billion connected smart devices providing data, services, and ubiquitous real-time information, touching all aspects of our lives, from healthcare to entertainment. Such a scenario necessitates a paradigmatic shift in the way that information processing, computation, and resource management are handled. One promising solution is to move away from the centralized client-server protocol, towards decentralized processing at the network edge. Such decentralization can effectively manage the increasing number of distributed devices and the surge in data, and meet the stringent latency requirements.
Haoran’s research focuses on such a distributed setting and addresses the question of identifying and achieving the best possible performance for distributed optimization and machine learning. The conference paper presents methods and algorithms capable of using large scale distributed resources such as data and computational power, to perform fast, decentralized, and scalable computation. In the paper, Haoran derives the fundamental performance limits for a class of challenging distributed optimization problems, where multiple nodes collectively optimize certain non-convex functions using local data. He presents an optimal algorithm, which enables the nodes to find high-quality solutions using the least amount of communication and computational resources.
Haoran Sun earned his Bachelor of Science in Automatic Control from Beijing Institute of Technology, China, in 2015, and earned his Master of Science in Industrial Engineering from Iowa State University in 2017. He is currently pursuing his doctoral degree under the guidance of Prof. Mingyi Hong. His research interests include optimization, machine learning, and its applications in signal processing and wireless communications.