Recently, four papers by Prof. Jianwei Huang's team at the School of Science and Engineering (SSE) of The Chinese University of Hong Kong, Shenzhen were accepted by IEEE International Conference on Computer Communications 2021 (INFOCOM 2021).

Hosted annually by IEEE, INFOCOM is one of the three top international conferences on networking in the research community. With an international reputation and wide academic influence, it has long been recommended by the China Computer Federation (CCF) as a Class A international academic conference. This year’s conference continued its rigorous double-blind review process and recorded an acceptance rate of less than 20% (252/1266).

Jianwei Huang is a Presidential Chair Professor and the Associate Dean of the School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen. He also serves as the Vice President of Shenzhen Institute of Artificial Intelligence and Robotics for Society (AIRS) and Director of the institute’s Research Center on Crowd Intelligence. Dr. Huang has published 290+ papers in leading international journals and conferences on communications and networking, with a total Google Scholar citations of 12800+ and an H-index of 58. He is the co-author of 9 Best Paper Awards, including the IEEE Marconi Prize Paper Award in Wireless Communications in 2011.

The four accepted papers of Prof. Jianwei Huang's team are:

1. Incentive Mechanism Design for Distributed Coded Machine Learning

2. Strategic Information Revelation in Crowdsourcing Systems Without Verification

3. Cost-Effective Federated Learning Design

4. Taming Time-Varying Information Asymmetry in Fresh Status Acquisition

Paper Descriptions

1. Incentive Mechanism Design for Distributed Coded Machine Learning

Authors

Ningning Ding (CUHK Ph.D.candidate; Visiting Ph.D.student at AIRS, CUHK-Shenzhen)

Zhixuan Fang (Assistant Professor, Tsinghua University)

Lingjie Duan (Assistant Professor, Singapore University of Technology and Design)

Jianwei Huang (Corresponding author)

Abstract

A distributed machine learning platform needs to recruit many heterogeneous worker nodes to finish computation simultaneously. As a result, the overall performance may be degraded due to straggling workers. By introducing redundancy into computation, coded machine learning can effectively improve the runtime performance by recovering the final computation result through the first k (out of the total n) workers who finish computation. While existing studies focus on designing efficient coding schemes, the issue of designing proper incentives to encourage worker participation is still under-explored. This paper studies the platform's optimal incentive mechanism for motivating proper workers' participation in coded machine learning, despite the incomplete information about heterogeneous workers' computation performances and costs. A key contribution of this work is to summarize workers' multi-dimensional heterogeneity as a one-dimensional metric, which guides the platform's efficient selection of workers under incomplete information with a linear computation complexity. Moreover, we prove that the optimal recovery threshold k is linearly proportional to the participator number n if we use the widely adopted MDS (Maximum Distance Separable) codes for data encoding. We also show that the platform's increased cost due to incomplete information disappears when worker number is sufficiently large, but it does not monotonically decrease in worker number.

2.Strategic Information Revelation in Crowdsourcing Systems Without Verification

Authors

Chao Huang (CUHK Ph.D.candidate; Visiting Ph.D.student at AIRS, CUHK-Shenzhen)

Haoran Yu (Associate Professor, Beijing Institute of Technology)

Jianwei Huang (Corresponding author)

Randall A. Berry (Professor, Northwestern University, USA)

Abstract

This paper focuses on crowdsourcing without verification. Specifically, crowdsourcing platforms want to motivate crowdsourced workers to provide high quality and authentic solutions, but at the same time are unable to validate the collected solutions. Most prior work assumes that the platform and workers share symmetric information. In this paper, however, we study an asymmetric information scenario in which the platform learns more information about the average accuracy of workers (e.g., the probability that a worker will provide the correct solution) and can strategically present this information to workers. Workers will use the published information to estimate their chances of receiving a crowdsourced reward.

Two types of workers are studied in this paper: (1) naive workers who fully trust the information posted on the platform, and (2) strategic workers who adjust their ideas based on the posted information. For naive workers, we find that crowdsourcing platforms should always declare a higher average accuracy of workers, thus maximizing platform revenue. However, this is not always appropriate for strategic workers, as it may reduce the credibility of the platform postings and thus the platform revenue. Interestingly, for strategic workers, it is even possible for the platform to announce a lower value to maximize revenue when the average accuracy of workers is high.

3. Cost-Effective Federated Learning Design

Authors

Bing Luo (International Joint Postdoctoral Fellow, AIRS, CUHK-Shenzhen and Yale University)

Xiang Li (CUHK-Shenzhen Ph.D.Candidate)

Shiqiang Wang (Research Fellow, IBM T. J. Watson Research Center)

Jianwei Huang (Corresponding author)

Leandros Tassiulas (Professor, Yale University)

Abstract

Federated learning (FL) is a distributed learning paradigm that enables a large number of devices to collaboratively learn a model without sharing their raw data. Despite its practical efficiency and effectiveness, the iterative on-device learning process incurs a considerable cost in terms of learning time and energy consumption, which depends crucially on the number of selected clients and the number of local iterations in each training round. In this paper, we analyze how to design adaptive FL that optimally chooses these essential control variables to minimize the total cost while ensuring convergence. Theoretically, we analytically establish the relationship between the total cost and the control variables with the convergence upper bound. To efficiently solve the cost minimization problem, we develop a low-cost sampling-based algorithm to learn the convergence related unknown parameters. We derive important solution properties that effectively identify the design principles for different metric preferences. Practically, we evaluate our theoretical results both in a simulated environment and on a hardware prototype. Experimental evidence verifies our derived properties and demonstrates that our proposed solution achieves near-optimal performance for various datasets, different machine learning models, and heterogeneous system settings.

 

4. Taming Time-Varying Information Asymmetry in Fresh Status Acquisition

Authors

Zhiyuan Wang (Postdoctoral Researcher, CUHK)

Lin Gao (Associate Professor, Harbin Institute of Technology, Shenzhen)

Jianwei Huang (Corresponding author)

Abstract

An increasing number of online platforms provide valuable real-time content (e.g., traffic congestion) to users by continuously acquiring the status of different points of interest (PoIs). In this issue, PoIs tend to have egoistic and time-varying preferences, which makes it difficult to optimize the frequency of state acquisition between individual PoIs and Platforms. In this paper, we consider a multi-stage real-time state collection problem that aims to maximize Social Welfare while ensuring the freshness of information across Platforms, where freshness is measured by the Age of Information (AoI). An auction-based multi-stage decomposition mechanism is designed to solve the time-varying asymmetric information. Experimental results based on real data show that when the platform is more concerned with maximizing revenue, each PoI can still obtain non-negative revenue in the long run.

About Jianwei Huang

Jianwei Huang is a Presidential Chair Professor and the Associate Dean of the School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen. He also serves as the Vice President of the Shenzhen Institute of Artificial Intelligence and Robotics for Society (AIRS) and Director of the institute’s Research Center on Crowd Intelligence. He is an IEEE Fellow, a Distinguished Lecturer of IEEE Communications Society, and a Clarivate Analytics Highly Cited Researcher in Computer Science. He received the Ph.D.degree from Northwestern University (USA) in 2005, worked as a Postdoc Research Associate at Princeton University (USA) during 2005-2007, and worked as Assistant/Associate/Full Professor at the Department of Information Engineering at the Chinese University of Hong Kong during 2007-2018.  

Dr. Huang has served as an Associate Editor of several JCR Q1 journals, such as IEEE Transactions on Mobile Computing, IEEE/ACM Transactions on Networking, IEEE Transactions on Network Science and Engineering, IEEE Transactions on Wireless Communications, IEEE Journal on Selected Areas in Communications - Cognitive Radio Series, and IEEE Transactions on Cognitive Communications and Networking. He has served as an Editor of Wiley Information and Communication Technology Series, Springer Encyclopedia of Wireless Networks, and Springer Handbook of Cognitive Radio. He has served as the Chair of IEEE Communications Society Cognitive Network Technical Committee and Multimedia Communications Technical Committee. He is the recipient of IEEE Communications Society Multimedia Communications Technical Committee Distinguished Service Award in 2015 and IEEE GLOBECOM Outstanding Service Award in 2010.

Prof. Jianwei Huang will be the new Editor-in-Chief of IEEE Transactions on Network Science and Engineering in January 2021.