Research Overview

My research philosophy: it is important to balance between solid mathematical and algorithmic theory and a diverse class of applications.

My research features theoretical framework, algorithm design and performance modeling of networked intelligent systems, where AI, computing, control and communication converge.

My research is cross-disciplinary, where tools in machine learning, artificial intelligence, deep learning, networking theory, information theory, control theory, optimization theory, probability theory, and game theory are leveraged. I am also interested in design and implementation of real-world AI systems.

In what follows, I give an overview of my research topics in recent years (with representative publications), which are (loosely) categorized into four main areas i) Distributed AI: Theory, Model, Algorithm and Applications; ii) AI and Machine Learning for Networked Systems; iii) AI for Social Benefit and iv) Principled Design to Demystify Black Box of Engineering .

Research Area I. Distributed AI: Theory, Model, Algorithm and Applications

Explainable, Controllable and Trustworthy AI with Performance Guarantee

During recent years, and especially since the stunning results Chat-GPT brought to the general public, the demands to comprehend and trust the results and output created by machine learning algorithms, to mitigate potential risks, prioritize safety and transparency for the people who interact with AI, and to explicit performance guarantee are increasing!

In this research project, we are interested in principled design of deep reinforcement learning and deep bandit learning for online interactive AI.

Y. Xu, S. Wang, H. Guo, Z. Shao and X. Liu, ‘‘Learning to Schedule Online Tasks with Bandit Feedback’’, in Proceedings of AAMAS 2024, Auckland, New Zealand, May 6-10, 2024.

S. Wang, S. Bian, X. liu, and Z. Shao, ‘‘Neural Constrained Combinatorial Bandits’’, in Proceedings of IEEE INFOCOM 2023, New York, U.S.A, May 17-20, 2023.

Edge Intelligence and On-Device Machine Learning Systems

More recently, with the proliferation of mobile computing and Internet-of-Things (IoT), billions of mobile and IoT devices are connected to the Internet, generating zillions Bytes of data at the network edge. Driving by this trend, there is an urgent need to push the AI frontiers to the network edge so as to fully unleash the potential of the edge big data. To meet this demand, edge (fog) computing, an emerging paradigm that pushes computing tasks and services from the network core to the network edge, has been widely recognized as a promising solution. The resulted new inter-discipline, edge AI or edge intelligence, is beginning to receive a tremendous amount of interest.

In this research project, we are interested in edge NLP, edge CV, and edge CG.

Recent Publications:

S. Wang, Z. Shao, and J. Lui, ‘‘Next-Word Prediction: A Perspective of Energy-Aware Distributed Inference’’, IEEE Transactions on Mobile Computing, vol. 23, no. 5, pp. 5695 - 5708, May 2024.

Y. Tang, J. Hou, X. Huang, Z. Shao, and Y. Yang, ‘‘Green Edge Intelligence Scheme for Mobile Keyboard Emoji Prediction’’, IEEE Transactions on Mobile Computing, vol. 23, no. 2, pp. 1888 - 1901, February 2024.

Research Area II. AI and Machine Learning for Networked Systems

Integration of Online Learning and Online Control(Optimization) for Networked Systems

Classcial online control(optimization) for networked systems always assume the availability of system parameters. For example, the channel statistics in edge computing and the preference in stable matching. However, in practice, such information is usually unknown and need to be learnt. The resulting interaction between online learning and online control is complex. In this research project, we adopt and further develop the constarined bandit learning methods to design the algorithms.

Recent Publications:

X. Gao, J. Wang, X. Huang, Q. Leng, Z. Shao, and Y. Yang, ‘‘Energy-Constrained Online Scheduling for Satellite-Terrestrial Integrated Networks, IEEE Transactions on Mobile Computing’’, vol. 22, no. 4, pp. 2163 - 2176, April 2023.

X. Gao, X. Huang, Y. Tang, Z. Shao , and Y. Yang, ‘‘History-Aware Online Cache Placement in Fog-Assisted IoT Systems: An Integration of Learning and Control’’, IEEE Internet of Things Journal, vol. 8, no. 19, pp. 14683-14704, October 2021.

X. Gao, X. Huang, Z. Shao , and Y. Yang, ‘‘An Integration of Online Learning and Online Control for Green Offloading in Fog-Assisted IoT Systems’’, IEEE Transactions on Green Communications and Networking, vol. 5, no. 3, pp. 1632-1646, September 2021.

X. Huang, Y. Tang, Z. Shao, and H. Xu, ‘‘Joint Switch-Controller Association and Control Devolution for SDN Systems: An Integration of Online Control and Online Learning’’, IEEE Transactions on Network and Service Management, vol. 18, no. 1, pp. 315-330, March 2021.

Predictive Scheduling for Networked Systems

The rapid development of machine learning and user behavior study have made it possible to learn and predict user behavior, e.g., mobility patterns, user preferences and software resource demand. Despite several progress, it is still open to efficiently incorporate such prediction information into network scheduling.

In this research project, we are interested in studying the fundamental benefit limit of prediction and predictive scheduling algorithm design, and apply such algorithms to practical systems including data streaming systems, edge computing systems, wireless caching networks, SDN systems, NFV systems et al.

Recent Publications:

X. Huang, Z. Shao and Y. Yang, ‘‘POTUS: Predictive Online Tuple Scheduling for Data Stream Processing Systems’’, IEEE Transactions on Cloud Computing, vol. 10, no. 4, pp. 2863 - 2875, December 2022.

X. Huang, S. Zhao, Z. Shao , H. Qian and Y. Yang, ‘‘Online User-AP Association with Predictive Scheduling in Wireless Caching Networks’’, IEEE Transactions on Mobile Computing, vol. 21, no. 6, pp. 2116 - 2129, June 2022.

X. Huang, S. Bian, X. Gao, W. Wu, Z. Shao , Y. Yang and J. Lui, ‘‘Online VNF Chaining and Scheduling with Prediction: Optimality and Trade-offs’’, IEEE/ACM Transactions on Networking, vol. 29, no. 4, pp. 1867-1880, August 2021.

X. Huang, S. Bian, Z. Shao and H. Xu, ‘‘Predictive Switch-Controller Association and Control Devolution for SDN Systems’’, IEEE/ACM Transactions on Networking, vol. 28, no. 6, pp. 2783-2796, December 2020.

X. Gao, X. Huang, S. Bian, Z. Shao, and Y. Yang, ‘‘PORA: Predictive Offloading and Resource Allocation in Dynamic Fog Computing Systems’’, IEEE Internet of Things Journal, vol. 7, no. 1, pp. 72-87, January 2020.

Research Area III. AI for Social Benefit

We are focused on advancing networked and distributed AI systems research for social impact in topics such as digital divide, the design for elderly, public health and environment conservation. We focus on fundamental research problems in multiagent networks, machine learning, reinforcement learning, game theory, bandit algorithms that are driven by these topics, ensuring a virtuous cycle of research and real-world applications. We simultaneously aim to achieve real-world social impact, often in domains with marginalized or endangered communities, and those that have not benefited from AI research in the past.

Edge Nature Language Processing for Digital Underprivileged Group

In this project, we want to help the people whose mobile devices were brought at least five years ago and could not afford upgrade to the latest mobile devices. We focus on providing NLP service to such people, an example of information accessibility via barrier-free design.

Recent Publications:

S. Wang, Z. Shao, and J. Lui, ‘‘Next-Word Prediction: A Perspective of Energy-Aware Distributed Inference’’, IEEE Transactions on Mobile Computing, vol. 23, no. 5, pp. 5695 - 5708, May 2024.

Y. Tang, J. Hou, X. Huang, Z. Shao, and Y. Yang, ‘‘Green Edge Intelligence Scheme for Mobile Keyboard Emoji Prediction’’, IEEE Transactions on Mobile Computing, vol. 23, no. 2, pp. 1888 - 1901, February 2024.

Research Area IV. Principled Design to Demystify Black Box of Engineering

There are many blackbox designs of engineering which enjoy empiricial success while no one knows why for a long time. Examples includes deep learning for AI and machine learning and fat-tree topology for data center networks. Several efforts have been made to find the hidden principles through reverse and forward engineering methods.

Understanding MLP

A Multilayer Perceptron (MLP) is the simplest yet powerful deep neural network. To understand the principle behind deep learning, it is the key step to understand MLP. Current theoretical tools include Neural Tangent Kernel (NTK) and Neural Network Gaussian Process (NNGP).

In this project, we focus on the MLP-based neural rendering, a popular application in graphics and vision. We establish the connection between theoretical understanding of MLP and the practical neural rendering algorithms.

Recent Publications:

H. Yu, A. Chen, X. Chen, L. Xu, Z. Shao, and J. Yu, ‘‘Anisotropic Fourier Features for Neural Image-Based Rendering and Relighting’’, in Proceedings of AAAI 2022 (Oral), February 22 - March 1, 2022.

When Shannon Meets Ramanujan: A Principled Design for Large-Scale Network Topology

For modern large-scale networked systems, ranging from cloud to edge computing systems, the topology design has a significant impact on the system performance in terms of scalability, cost, latency, throughput, and fault-tolerance. These performance metrics may conflict with each other and design criteria often vary across different networks. To date, there has been little theoretic foundation on topology designs from a prescriptive perspective, indicating that the current status quo of the design process is more of an art than a science.

In this project, we advocate a novel unified framework to describe, generate, and analyze topology design in a systematic fashion. By reverse-engineering existing topology designs and developing a fine-grained decomposition method for topology design, we propose a general procedure that serves as a common language to describe topology design. By proposing general criteria for the procedure, we devise a top-down approach to generate topology models, based on which we can systematically construct and analyze new topologies. To validate our approach, we leverage concrete tools based on combinatorial design theory and propose several novel topology models. With quantitative performance analysis, we reveal the trade-offs among performance metrics and generate new topologies with various advantages for different large-scale networks.

Recent Publications:

Y. Chang, X. Huang, L. Deng, Z. Shao, and J. Zhang, ‘‘Systematic Topology Design for Large-Scale Networks: A Unified Framework’’, in Proceedings of IEEE INFOCOM 2020, Toronto, Canada, July 6-9, 2020.

Markov Approximation Framework for Distributed Stochastic Algorithm Design

Many important network resource allocation problems in wireless networks,Peer-to-Peer networks, fog computing and cloud computing are combinatorial optimization problems. In general, exact solutions of these problems are computationally prohibitive. Efficient approximation algorithms either do not exist or only allow centralized implementation. The Markov approximation framework we developed is a general framework to guide distributed algorithm design for solving combinatorial network optimization problems. Within this framework, we can synthesize distributed approximation algorithms in a systematical way, which converge fast, attain close-to-optimal performances, scale with the network size, and adapt smoothly to network and user dynamics. Our past investigations on this framework include: equivalence between distributed algorithm design and distributed implementation of Markov chain;state-connectivity structure of Markov chain, state transition rate of Markov chain, perturbation of Markov chain, quantization of optimality loss caused by the perturbation, application of this framework to various network domains such as reverse-engineering BitTorrent protocol and traffic engineering of data center networks.

In this research project, we are interested in (i) speed the convergence rate of algorithms (ii) application to new networked systems.

Recent Publications:

S. Bian, X. Huang, X. Gao, Z. Shao , and Y. Yang, ‘‘Service Chain Composition with Failures in NFV Systems: A Game-Theoretic Perspective’’, IEEE Transactions on Network and Service Management, vol. 18, no. 1, pp. 224-239, March 2021.

S. Li, Z. Shao and J. Huang, ‘‘ARM: Anonymous Rating Mechanism for Discrete Power Control’’, IEEE Transactions on Mobile Computing (TMC), vol. 16, issue 2, pp.326-340, Feb. 2017.

S. Zhang, Z. Shao, M. Chen and L. Jiang, ‘‘Optimal Distributed P2P Streaming under Node Degree Bounds’’, IEEE/ACM Transactions on Networking, vol. 22, issue 3, pp.717-730, June 2014.

M. Chen, S. Liew, Z. Shao, and C. Kai, ‘‘Markov Approximation for Combinatorial Network Optimization’’, IEEE Transactions on Information Theory, vol.59, no.10, pp.6301 - 6327, Oct. 2013.