My research is on machine learning systems: the theory, algorithms and software that make AI reliable, efficient and verifiable at scale, with a current focus on AI for Science, Healthcare AI and AI-native Wireless Edge. Openings: one postdoctoral scholar to work on agentic AI, with a strong publication record, and one PhD student strong in software engineering and mathematics (linear algebra and optimization). Please send your CV and transcripts to cheewei.tan@ntu.edu.sg.

(1) Trustworthy LLM Agents and AI Systems

The question in deploying an LLM agent is not whether the model can answer but whether the system around it can tell when it should not. We develop provenance-aware evidence assessment for agentic retrieval-augmented generation [1], with a clinical instantiation in (3), graph-based RAG for automated fact-checking (TrumorGPT) [2], evaluation methodology for LLMs [3], adversarial robustness of AI-generated-text detectors [4], and risk awareness in LLM agents by probing their safety boundaries [5]. In each case the agent is treated as a system with measurable failure modes and given the instrumentation to certify its outputs. Current work studies low-rank adaptation (LoRA) from an optimization standpoint, asking what a fine-tuning update should preserve and what it can discard. The agents and copilots we build as software are described in (2).

[1] S. Sankeerthana and C. W. Tan, PROVE-RAG: Provenance-Aware Evidence-State Assessment for Trustworthy Agentic Retrieval-Augmented Generation, submitted, 2026.
[2] C. N. Hang, P.-D. Yu and C. W. Tan, TrumorGPT: Graph-Based Retrieval-Augmented Large Language Model for Fact-Checking, IEEE Transactions on Artificial Intelligence, Vol. 6, No. 11, pp. 3148-3162, 2025.
[3] Md. T. R. Laskar, S. Alqahtani, M. S. Bari, M. Rahman, Md. A. M. Khan, H. Khan, I. Jahan, A. Bhuiyan, C. W. Tan, Md. R. Parvez, E. Hoque, S. Joty, J. X. Huang, A Systematic Survey and Critical Review on Evaluating Large Language Models: Challenges, Limitations, and Recommendations, EMNLP, 2024.
[4] R. Tu, X. Kang, C. W. Tan, C.-H. Chi and K.-Y. Lam, Sentences Based Adversarial Attack on AI-Generated Text Detectors, IEEE Transactions on Big Data, Vol. 12, No. 1, pp. 80-91, 2026.
[5] Y. Jiang, H. Peng, Y. Zheng, Z. Liu, K.-Y. Lam and C. W. Tan, Enhancing Risk Awareness in LLM Agents via Probing Safety Boundaries, IEEE ICASSP, 2026.

(2) AI Agents and Copilots

Much of our work ships as software, and we are interested in how software is designed once language models are part of it, what some call Software 3.0: programs that are specified in natural language, generated and tested by models, and kept correct by the tools around them. We build AI copilots for programming that retrieve context before they generate and test what they generate. Our open-source copilot on GitHub, which carries code generation through to software testing with context-based retrieval-augmented generation [1], received the 2025 IEEE Software Best Paper Award; our open-source codebase has been assimilated into GitHub Copilot since March 2025 and is used by many developers. Alongside it are a local-cloud copilot that routes between on-device and cloud models (CAMP) [2] and Copilot for Xcode, an early in-IDE assistant driven by cloud-hosted LLMs [3]. Beneath the copilots is the question of how to align a code-generation model with the people who use it, which we treat as Bayesian inference over crowd-sourced human feedback [4], and a review of natural language generation and understanding of big code [5]. Current work extends the copilot into an autoresearch agent that proposes, runs, verifies and profiles its own experiments, with GPU-kernel generation as the first testbed. These copilots are also the instruments of our education research in (9).

[1] Y. Wang, S. Guo and C. W. Tan, From Code Generation to Software Testing: AI Copilot With Context-Based Retrieval-Augmented Generation, IEEE Software, Vol. 42, No. 4, pp. 34-42, 2025 (IEEE Software Best Paper Award, 2025).
[2] Y. Wang, S. Guo and C.-W. Tan, Contextual Augmented Multi-Model Programming (CAMP): A Local-Cloud Copilot Solution, IEEE Conference on Artificial Intelligence, 2025.
[3] C. W. Tan, S. Guo, M. F. Wong and Y. Chen, Copilot for Xcode: Exploring AI-Assisted Programming by Prompting Cloud-based Large Language Models, Symposium on Large Language Models, 2023.
[4] M. F. Wong and C. W. Tan, Aligning Crowd-Sourced Human Feedback for Reinforcement Learning on Code Generation by Large Language Models, IEEE Transactions on Big Data, 2025 (Honorable Mention Paper Award, Foundation Models and Generative AI in IEEE Conference on Artificial Intelligence, 2024).
[5] M. F. Wong, S. Guo, C.-N. Hang, S.-W. Ho and C. W. Tan, Natural Language Generation and Understanding of Big Code for AI-Assisted Programming: A Review, Entropy, 25(6): 888, 2023.

(3) Healthcare AI

Healthcare is where an AI system's efficiency must be optimized and AI safety is paramount. Our work spans two scales. At the scale of Electronic Health Record (EHR) data query, we build medical LLM systems that gate their own answers on evidence sufficiency [1] and ground them in provenance [2]. At the scale of populations, we work with clinical researchers on the spread of infectious diseases and antimicrobial resistance, developing AI-native decision intelligence for early warning, contact tracing and AI-health governance [3-5], building on a decade of work on contagion source detection and epidemic centrality that culminated in a Foundations and Trends monograph [6-8] where the same contagion mathematics governs infectious diseases and misinformation, and are useful to study future outbreaks.

[1] S. Sankeerthana and C. W. Tan, Detecting Evidence Insufficiency for Gated Hallucination Mitigation in Medical Large Language Models, submitted, 2026.
[2] S. Sankeerthana and C. W. Tan, PROVE-RAG: Provenance-Aware Evidence-State Assessment for Trustworthy Agentic Retrieval-Augmented Generation, submitted, 2026.
[3] J. Wang, P. Yang, G. Li, X. Wang, Y. Sun, Z. Feng, Y. Gao, Y. Li, W. Guo, Z. Li, T. Zhang, J. Ma, C. W. Tan, Z. Cao, D. Zeng, Reframing Epidemic Early Warning: Toward AI-Native Decision Intelligence for Public Health Governance, npj Digital Public Health, accepted, 2026.
[4] C.-W. Tan, P.-D. Yu, S. Chen and H. V. Poor, DeepTrace: Learning to Optimize Contact Tracing in Epidemic Networks With Graph Neural Networks, IEEE Transactions on Signal and Information Processing over Networks, Vol. 11, pp. 97-113, 2025.
[5] C. N. Hang, P.-D. Yu, S. Chen, C. W. Tan and G. Chen, MEGA: Machine Learning-Enhanced Graph Analytics for Infodemic Risk Management, IEEE Journal of Biomedical and Health Informatics, Vol. 27, No. 12, pp. 6100-6111, 2023.
[6] C. W. Tan and P.-D. Yu, Contagion Source Detection in Epidemic and Infodemic Outbreaks: Mathematical Analysis and Network Algorithms, Foundations and Trends in Networking, Vol. 13, No. 2-3, pp. 107-251, 2023.
[7] P.-D. Yu, C. W. Tan and H. Fu, Epidemic Source Detection in Contact Tracing Networks: Epidemic Centrality in Graphs and Message-Passing Algorithms, IEEE Journal of Selected Topics in Signal Processing, Vol. 16, No. 2, pp. 234-249, 2022.
[8] Z. Fei, Y. Ryeznik, O. Sverdlov, C. W. Tan and W. K. Wong, An Overview of Healthcare Data Analytics with Applications to the COVID-19 Pandemic, IEEE Transactions on Big Data, Vol. 8, No. 6, pp. 1463-1480, 2022.

(4) Optimization and Differentiable Programming

Differentiable programming is the idea at the heart of deep learning: the paradigm behind PyTorch, JAX and Julia treats a whole program as a computational graph and trains it by backpropagation, so an iterative solver or a fixed-point map becomes a trainable module of the same deep learning software. We use it to learn how to design optimization algorithms, embedding first-order methods and their duals inside these frameworks and training end to end (see the figure, from [1]); first-order methods are also what make optimization GPU-accelerated, since their iterations are the matrix-vector products and proximal steps that GPUs run best. Our INFOCOM 2026 paper DIFFRACT neuralizes utility maximization over standard interference functions, differentiating through the fixed point by the implicit function theorem rather than unrolling it [2]. The same machinery serves as a proof engine: automated reasoning by convex optimization, solved with GPU-accelerated ADMM at the scale of millions of constraints, proves or disproves information inequalities and is released as the open-source AITIP solver [3].

[1] L. Tao, X. Tong and C. W. Tan, Learning to Optimize by Differentiable Programming, arXiv 2601.16510, 2026.
[2] C. W. Tan et al., DIFFRACT: Neuralized Utility Maximization for Wireless Networks by Differentiable Programming, IEEE INFOCOM, 2026.
[3] S. W. Ho, L. Ling, C. W. Tan and R. W. Yeung, Proving and Disproving Information Inequalities: Theory and Scalable Algorithms, IEEE Transactions on Information Theory, Vol. 66, No. 9, pp. 5522-5536, 2020 (AITIP open-source code).

(5) AI-Native Wireless Networks

In an AI-native network the optimizer is itself a learned system that must run in real time and come with a guarantee. Our approach is learning to optimize: unroll a provably convergent algorithm into a neural architecture, or train a reinforcement-learning policy to drive its block-coordinate steps, so that the learned system inherits the fixed-point structure of the algorithm. Recent results include deep reinforcement learning-based block coordinate descent for downlink weighted sum-rate maximization [1], deep unfolding of fixed-point algorithms for weighted sum-rate maximization [2], and adversarial water-filling as a minimax resource-allocation foundation model, which allocates power against worst-case channel uncertainty and applies to low Earth orbit (LEO) satellite networks, where channels change on the timescale of a satellite pass [3-4].

[1] S. Chen, C. W. Tan and H. V. Poor, Deep Reinforcement Learning-Based Block Coordinate Descent for Downlink Weighted Sum-Rate Maximization on AI-Native Wireless Networks, IEEE Transactions on Wireless Communications, Vol. 25, pp. 3658-3674, 2026.
[2] J. C. Riedel, C.-W. Tan and G. Caire, Deep Unfolding of Fixed-Point Based Algorithm for Weighted Sum Rate Maximization, IEEE ISIT, 2025.
[3] X. Tong, C. W. Tan and H. V. Poor, Adversarial Water-Filling: Theory, Algorithms and Foundation Model, arXiv 2605.26163, 2026.
[4] X. Tong, C.-W. Tan and H. V. Poor, Adversarial Water-Filling: Minimax Resource Allocation Optimization with Proximal Decomposition in Open RAN, IEEE Globecom, 2025.

(6) Privacy-Preserving Federated Learning

Training where the data lives means training under adversaries, churn, bandwidth limits and privacy obligations, and a federated system is only as trustworthy as the guarantees it can produce. We build federated and edge-learning systems that come with certificates: efficient and certified recovery of a global model from poisoning attacks, so a compromised run can be repaired rather than restarted [1]; primal-dual optimization that certifies the right to be forgotten for samples and labels in vertical federated learning [2], and efficient federated unlearning with adaptive differential-privacy preservation [3]; and a credible and fair federated learning framework based on blockchain, in which contribution and reward are auditable [4]. On the communication side, we surveyed coded federated learning for communication-efficient edge computing, where coding theory trades redundancy for straggler tolerance [5].

[1] Y. Jiang, J. Shen, Z. Liu, C. W. Tan and K.-Y. Lam, Towards Efficient and Certified Recovery from Poisoning Attacks in Federated Learning, IEEE Transactions on Information Forensics and Security, Vol. 20, pp. 2632-2647, 2025.
[2] Y. Jiang, X. Tong, Z. Liu, X. Zhang, K.-Y. Lam and C.-W. Tan, Certifying the Right to Be Forgotten: Primal-Dual Optimization for Sample and Label Unlearning in Vertical Federated Learning, IEEE Transactions on Information Forensics and Security, Vol. 20, pp. 13143-13158, 2025.
[3] Y. Jiang, X. Tong, Z. Liu, H. Ye, C. W. Tan and K.-Y. Lam, Efficient Federated Unlearning with Adaptive Differential Privacy Preservation, IEEE International Conference on Big Data, 2024.
[4] L. Chen, D. Zhao, L. Tao, K. Wang, S. Qian, X. Zeng and C. W. Tan, A Credible and Fair Federated Learning Framework Based on Blockchain, IEEE Transactions on Artificial Intelligence, Vol. 6, No. 2, pp. 301-316, 2025.
[5] Y. Zhang, T. Gao, C. Li and C. W. Tan, Coded Federated Learning for Communication-Efficient Edge Computing: A Survey, IEEE Open Journal of the Communications Society, Vol. 5, pp. 4098-4124, 2024.

(7) Cloud Computing and AI Data Centers

AI is trained and served in data centers, and how computation is paid for and how fast it moves through the network are systems questions in their own right. On the economics, we studied how a user should bid for spot-priced cloud resources in online secondary markets [1] and when a virtual service provider that resells capacity is viable [2]; the mobile and spectrum side of this work is in (11). Three further lines address the computation itself. For graph analytics at scale we developed parallel triangle-counting algorithms with pruning and hierarchical clustering on MapReduce for the DARPA-MIT-Amazon Graph Challenge [3], with the MapReduce/Hadoop code released on GitHub. For parallel computing with unreliable workers we showed that fault tolerance and network coding are the same problem, and derived optimal scheduling of coded tasks so that stragglers and failures do not stall the job [4]. Our recent work turns to the data-center and wide-area network: accelerating distributed graph learning by moving multicast and aggregation into the network fabric [5], the kind of in-network computing that AI training clusters now depend on, and learning the routing policies of data-center and backbone networks from observed traffic [6].

[1] L. Zheng, C. Joe-Wong, C. W. Tan, M. Chiang and X. Wang, How to Bid the Cloud?, ACM SIGCOMM Computer Communication Review, Vol. 45, No. 4, pp. 71-84, 2015.
[2] L. Zheng, C. Joe-Wong, C. G. Brinton, C. W. Tan, S. Ha and M. Chiang, On the Viability of a Cloud Virtual Service Provider, ACM SIGMETRICS, 2016.
[3] C. Y. Kuo, C. N. Hang, P.-D. Yu and C. W. Tan, Parallel Counting of Triangles in Large Graphs: Pruning and Hierarchical Clustering Algorithms, IEEE High Performance Extreme Computing Conference (HPEC), pp. 1-6, 2018 (MIT-Amazon Graph Challenge Honorable Mention Paper Award).
[4] C. Li, Y. Zhang and C. W. Tan, Fault-Tolerant Computation Meets Network Coding: Optimal Scheduling in Parallel Computing, IEEE Transactions on Communications, Vol. 71, No. 7, pp. 3847-3860, 2023 (open-source code).
[5] Z. Li, J. Huang, Y. Li, J. Liu, J. Zhang, H. Li, X. Zhu, S. Zhou, J. Shao, X. Lu, Q. Su, J. Wang, C. W. Tan, Y. Cui and K. Chen, Accelerating Distributed Graph Learning by Using Collaborative In-Network Multicast and Aggregation, USENIX Annual Technical Conference, 2025.
[6] K. Wang and C. W. Tan, Reverse Engineering Segment Routing Policies and Link Costs With Inverse Reinforcement Learning and EM, IEEE Transactions on Machine Learning in Communications and Networking, Vol. 3, pp. 1014-1029, 2025.

(8) Network Science of Misinformation and Infodemics

Who started it? We treat rumor source detection as statistical inference on a graph, with network centrality as the estimator, and have developed the maximum-likelihood theory and scalable algorithms for Markovian spreading models [1-3], and citation-network measures of scholarly novelty and viral influence [4]. Combined with the analysis of political leaning from retweet networks [5], this gives an end-to-end pipeline from detecting a spreading falsehood to locating its origin; it shares its mathematics with the epidemic work in (3).

[1] P.-D. Yu and C. W. Tan, Unraveling the Viral Spread of Misinformation: Maximum-Likelihood Estimation and Starlike Tree Approximation in Markovian Spreading Models, IEEE Transactions on Signal Processing, Vol. 73, pp. 446-461, 2025.
[2] C. W. Tan and P.-D. Yu, Contagion Source Detection in Epidemic and Infodemic Outbreaks: Mathematical Analysis and Network Algorithms, Foundations and Trends in Networking, Vol. 13, No. 2-3, pp. 107-251, 2023.
[3] Z. Wang, W. Dong, W. Zhang and C. W. Tan, Rumor Source Detection with Multiple Observations: Fundamental Limits and Algorithms, ACM SIGMETRICS, 2014; journal version in IEEE JSTSP, Vol. 9, No. 4, pp. 663-677, 2015.
[4] C. N. Hang, P.-D. Yu, C.-W. Tan and D. M. Chiu, When Ideas Go Viral: Measuring Scholarly Novelty and Viral Influence via Citation Network Analysis, IEEE Globecom, 2025.
[5] F. M. F. Wong, C. W. Tan, S. Sen and M. Chiang, Quantifying Political Leaning from Tweets, Retweets and Retweeters, IEEE Transactions on Knowledge and Data Engineering, Vol. 28, No. 8, pp. 2158-2172, 2016.

(9) AI in Education

AI assistance can build skill or replace it, and the difference is a systems question: what the tool withholds, when it intervenes, and how it measures whether the learner is still learning. We formalized this as an optimal knowledge-tracing problem for AI-assisted personalized education [1], and built and deployed the tools to study it: the programming copilots of (2), browser-sized agents that run small language models with WebLLM entirely in the student's browser for tool-augmented programming education [2], strategic LLM gaming agents for interactive learning [3-4], an LLM-driven generator of personalized assessment [5], and classroom-flipping designs that turn peer instruction and just-in-time teaching into an optimization problem [6-7]. Our study of ChatGPT and self-regulated learning [8] was among the earliest controlled classroom evidence on the question.

[1] Y. Wang, P.-D. Yu and C. W. Tan, Future-Proofing Programmers: Optimal Knowledge Tracing for Artificial Intelligence-Assisted Personalized Education, IEEE Signal Processing Magazine, Vol. 43, No. 1, pp. 69-82, 2026 (Special Issue on AI for Education).
[2] S. Bogdan, Y. Wang and C. W. Tan, Browser-Sized Agents: Small Language Models and WebLLM for Tool-Augmented AI-Assisted Programming Education, 2026.
[3] C. W. Tan, Y. Wang and S. Guo, Nemobot Games: Crafting Strategic AI Gaming Agents for Interactive Learning with Large Language Models, IEEE Transactions on Games, to appear, 2026.
[4] Y. Wang, S. Guo, L. Ling and C. W. Tan, Nemobot: Crafting Strategic Gaming LLM Agents for K-12 AI Education, ACM Learning at Scale, 2024.
[5] C. N. Hang, C. W. Tan and P.-D. Yu, MCQGen: A Large Language Model-Driven MCQ Generator for Personalized Learning, IEEE Access, Vol. 12, pp. 102261-102273, 2024.
[6] C. W. Tan, Large Language Model-Driven Classroom Flipping: Empowering Student-Centric Peer Questioning with Flipped Interaction, IEEE CAI Workshop on AI for Education, 2024.
[7] J. Li, L. Ling and C. W. Tan, Blending Peer Instruction with Just-In-Time Teaching: Jointly Optimal Task Scheduling with Feedback for Classroom Flipping, ACM Learning at Scale, 2021.
[8] D. T. K. Ng, C. W. Tan and J. K. L. Leung, Empowering Student Self-Regulated Learning and Science Education through ChatGPT: A Pioneering Pilot Study, British Journal of Educational Technology, 55(4), pp. 1328-1353, 2024 (top 10 most-cited papers published by the journal in 2024).

(10) Network Optimization

Before machine learning entered the loop, the systems we studied were networks whose resource-allocation problems are nonconvex and whose decisions are made by many nodes with local information, and much of what we now do with learned optimizers rests on what we proved about them. Two themes run through this work: nonconvexity, and decentralized optimization in which each node updates using only what it can measure. In wireless networks under interference, we showed that power control can be posed and solved as geometric programming [1], and then developed the nonlinear Perron-Frobenius theory of utility maximization [2-9], which turned a representative class of nonconvex problems on fairness and efficiency, from max-min SINR, max-min rate and outage fairness under Rayleigh fading to sum-rate maximization, into fixed-point computations with fast, provably convergent and naturally decentralized algorithms, resolving several long-open cases along the way (e.g., Kandukuri and Boyd, TWC 2002); the monograph [2] gives the overview and code is available on GitHub. The same fixed-point tools gave optimal spectrum management and network duality in cognitive radio networks [8, 10] and, for the Internet, a characterization of the cost of not splitting flows in routing [11].

[1] M. Chiang, C. W. Tan, D. P. Palomar, D. O'Neill and D. Julian, Power Control by Geometric Programming, IEEE Transactions on Wireless Communications, Vol. 6, No. 7, pp. 2640-2651, 2007 (highly cited, Web of Science, 2013).
[2] C. W. Tan, Wireless Network Optimization by Perron-Frobenius Theory, Foundations and Trends in Networking, Now Publishers, 2015.
[3] C. W. Tan, S. Friedland and S. H. Low, Nonnegative Matrix Inequalities and Their Application to Nonconvex Power Control Optimization, SIAM Journal on Matrix Analysis and Applications, Vol. 32, No. 3, pp. 1030-1055, 2011.
[4] C. W. Tan, M. Chiang and R. Srikant, Fast Algorithms and Performance Bounds for Sum Rate Maximization in Wireless Networks, IEEE/ACM Transactions on Networking, Vol. 21, No. 3, pp. 706-719, 2013 (conference version in IEEE INFOCOM, 2009).
[5] C. W. Tan, M. Chiang and R. Srikant, Maximizing Sum Rate and Minimizing MSE on Multiuser Downlink: Optimality, Fast Algorithms and Equivalence via Max-min SIR, IEEE Transactions on Signal Processing, Vol. 59, No. 12, pp. 6127-6143, 2011.
[6] C. W. Tan, Optimal Power Control in Rayleigh-Fading Heterogeneous Wireless Networks, IEEE/ACM Transactions on Networking, Vol. 24, No. 2, pp. 940-953, 2015 (conference version in IEEE INFOCOM, 2011).
[7] L. Zheng, D. W. H. Cai and C. W. Tan, Max-min Fairness Rate Control in Wireless Networks: Optimality and Algorithms by Perron-Frobenius Theory, IEEE Transactions on Mobile Computing, Vol. 17, No. 1, pp. 127-140, 2017 (conference version in IEEE INFOCOM, 2012).
[8] C. W. Tan, S. Friedland and S. H. Low, Spectrum Management in Multiuser Cognitive Wireless Networks: Optimality and Algorithm, IEEE Journal on Selected Areas in Communications, Vol. 29, No. 2, pp. 421-430, 2011.
[9] L. Zheng, Y. W. Hong, C. W. Tan, C. L. Hsieh and C. H. Lee, Wireless Max-min Utility Fairness with General Monotonic Constraints by Perron-Frobenius Theory, IEEE Transactions on Information Theory, Vol. 62, No. 12, pp. 7283-7298, 2016 (conference version in IEEE INFOCOM, 2014).
[10] L. Zheng and C. W. Tan, Cognitive Radio Network Duality and Algorithms for Utility Maximization, IEEE Journal on Selected Areas in Communications, Vol. 31, No. 3, pp. 500-513, 2013.
[11] M. Wang, C. W. Tan, W. Xu and A. Tang, Cost of Not Splitting in Routing: Characterization and Estimation, IEEE/ACM Transactions on Networking, Vol. 19, No. 6, pp. 1849-1859, 2011.

(11) Economic Theory and Computation in Networks

Networks and clouds are shared resources, so who gets them is decided by prices as much as by algorithms, and the pricing rule is itself something to design and compute. We studied how a user should bid for spot-priced cloud resources in online secondary markets such as Amazon Web Services [1], and whether a virtual service provider that resells capacity can be economically viable [2-3], with the MapReduce/Hadoop implementations released on GitHub. On the mobile side, we designed data-pricing protocols that let users trade and customize data plans, and quantified when such trading benefits both users and operators [4]. In wireless spectrum, price-driven access algorithms expose the tradeoff between energy and infeasibility when many secondary users contend for the same band [5]. In each case the economic mechanism and the optimization algorithm are designed together.

[1] L. Zheng, C. Joe-Wong, C. W. Tan, M. Chiang and X. Wang, How to Bid the Cloud?, ACM SIGCOMM Computer Communication Review, Vol. 45, No. 4, pp. 71-84, 2015.
[2] S. Liu, C. Joe-Wong, J. Chen, C. Brinton, C. W. Tan and L. Zheng, Economic Viability of a Virtual ISP, IEEE/ACM Transactions on Networking, Vol. 28, No. 2, pp. 902-916, 2020 (conference version in IEEE INFOCOM, 2017).
[3] L. Zheng, C. Joe-Wong, C. G. Brinton, C. W. Tan, S. Ha and M. Chiang, On the Viability of a Cloud Virtual Service Provider, ACM SIGMETRICS, 2016.
[4] L. Zheng, C. Joe-Wong, C. W. Tan, S. Ha and M. Chiang, Customized Data Plans for Mobile Users: Feasibility and Benefits of Data Trading, IEEE Journal on Selected Areas in Communications, Vol. 35, No. 4, pp. 949-963, 2017 (conference version, Secondary Markets for Mobile Data: Feasibility and Benefits of Traded Data Plans, in IEEE INFOCOM, 2015, pp. 1580-1588).
[5] X. Zhai, L. Zheng and C. W. Tan, Energy-Infeasibility Tradeoff in Cognitive Radio Networks: Price-Driven Spectrum Access Algorithms, IEEE Journal on Selected Areas in Communications, Vol. 32, No. 3, pp. 528-538, 2014.

(12) Cyber-Physical Systems Optimization

Physical infrastructures pose optimization problems on very large graphs, where a decision at one node propagates to the rest. In the smart grid, we established uniqueness and algorithms for optimal power flow in resistive networks [1], and valley-filling algorithms that schedule electric-vehicle charging to flatten demand [2]. Cascading failure is the opposite problem, protecting a network from its own connectivity, which we formulated as a poset-constrained graph problem with algorithms that scale to large infrastructures [3].

[1] C. W. Tan, D. W. H. Cai and X. Lou, Resistive Network Optimal Power Flow: Uniqueness and Algorithms, IEEE Transactions on Power Systems, Vol. 30, No. 1, pp. 263-273, 2015.
[2] N. Chen, C. W. Tan and T. Quek, Electric Vehicle Charging in Smart Grid: Optimality and Valley-Filling Algorithms, IEEE Journal of Selected Topics in Signal Processing, Vol. 8, No. 6, pp. 1073-1083, 2014.
[3] P.-D. Yu, C. W. Tan and H.-L. Fu, Averting Cascading Failures in Networked Infrastructures: Poset-Constrained Graph Algorithms, IEEE Journal of Selected Topics in Signal Processing, Vol. 12, No. 4, pp. 733-748, 2018.

(13) Coding Theory and Fundamental Limits

What is the best any scheme could do, and what does it cost to reach it? For distributed storage in cloud systems, we characterized the rate region of a class of secure asymmetric multilevel diversity coding systems, where data is recoverable from any sufficient subset of nodes while leaking nothing to a limited eavesdropper [1]. The same idea applies to computation: treating stragglers and failures as erasures gives optimal scheduling of coded tasks [2]. The figure shows the residue-number form of this principle, where a message stored as residues modulo coprime moduli is recovered from any sufficient subset by the Chinese remainder theorem, the algebraic structure that CRT decoding shares with linear network coding. In high-mobility wireless, the structure of Alamouti space-time codes yields low-complexity multiuser detection [3], and for interference channels we established the maximum achievable sum rate under successive decoding [4].

[1] C. Li, X. Guang, C. W. Tan and R. W. Yeung, Fundamental Limits on a Class of Secure Asymmetric Multilevel Diversity Coding Systems, IEEE Journal on Selected Areas in Communications, Vol. 36, No. 4, pp. 737-747, 2018.
[2] C. Li, Y. Zhang and C. W. Tan, Fault-Tolerant Computation Meets Network Coding: Optimal Scheduling in Parallel Computing, IEEE Transactions on Communications, Vol. 71, No. 7, pp. 3847-3860, 2023 (open-source code).
[3] C. W. Tan and A. R. Calderbank, Multiuser Detection of Alamouti Signals, IEEE Transactions on Communications, Vol. 57, No. 7, pp. 2080-2089, 2009.
[4] Y. Zhao, C. W. Tan, A. S. Avestimehr, S. N. Diggavi and G. J. Pottie, On the Maximum Achievable Sum-rate with Successive Decoding in Interference Channels, IEEE Transactions on Information Theory, Vol. 58, No. 6, pp. 3798-3820, 2012.