Adaptive networks
Policies that keep learning when channels, users, rewards, or capacities change over time.
NetworksElectrical Engineering · University of Southern California
Mevan Wijewardena
PhD Candidate in Electrical Engineering University of Southern California
Theory is most interesting when it changes what a system can do.
I am a Ph.D. candidate in Electrical Engineering at the University of Southern California, advised by Professor Michael J. Neely. My research develops mathematically rigorous algorithms for optimization and sequential decision-making under uncertainty, drawing on stochastic optimization, stochastic control, bandit learning, online learning, and game theory. I am particularly interested in non-stationarity, fairness, and information asymmetry.
Before joining USC, I earned a First Class Honours bachelor’s degree in Electronic and Telecommunication Engineering from the University of Moratuwa, Sri Lanka. I enjoy turning mathematical structure into algorithms that are provable, adaptive, and useful in real networked systems.
Research interests
- Stochastic optimization
- Online learning
- Stochastic control
- Game theory
Research / focus
Making uncertainty mathematically manageable.
I build algorithms for networked systems where decisions must be made before the environment is fully known.
Bandits & partial feedback
Online methods that learn useful decisions from success/failure signals and limited observations.
LearningStochastic control
Queueing, drift-plus-penalty, and utility optimization with performance guarantees.
ControlGames & fairness
Robust decisions for multi-agent systems with asymmetric information and competing incentives.
Multi-agentSelected work / recent
Research with proofs—and a path to practice.
Online Multi-player Resource-Sharing Games with Bandit Feedback
This paper explores a multi-player resource sharing game with a fair-reward allocation model and extends it to an online scenario.
Adaptive Algorithms for Automatic Link Selection in Multiple Access with Link Failures
This paper is about developing adaptive algorithms for utility maximization with bandit feedback. The paper also looks at models with matching constraints.
Automatic Link Selection in Multi-Channel Multiple Access with Link Failures
This paper focuses on the problem of automatic link selection in multi-channel multiple access control using bandit feedback.
Asymptotic Behaviour of the Connection Probability of Cell Partitioned Networks
This write up focusses on establshing the asymptotic connectiveity probability of cell partitioned networks as a function of the user density.
Experience / trajectory
Researcher, educator, problem solver.
From mathematical olympiads to adaptive network algorithms, the throughline is the same: find structure in difficult problems and make it useful.
Full curriculum vitae2022—Present
Education & research
PhD in Electrical Engineering
University of Southern California · Los Angeles
Advised by Professor Michael J. Neely. Research in adaptive network optimization, stochastic control, online learning, and multi-agent systems.
2023—Present
Teaching
Teaching Assistant
USC · Electrical and Computer Engineering
Stochastic processes, financial engineering mathematics, computer networks, computing principles, and applied linear algebra.
Teaching record2021—2022
Academic experience
Lecturer on Contract
University of Moratuwa · Sri Lanka
Taught and designed laboratory work across communications, random processes, electronics, and computer organization.
2016—2021
Education
BSc Engineering, First Class Honours
University of Moratuwa · Rank 5 of 101
Electronic and Telecommunication Engineering, with research spanning wireless security, HCI, and non-intrusive load monitoring.
Selected distinctions
A lifelong practice of solving hard problems.
International Mathematical Olympiad
Bronze medals in 2012, 2013, 2014, and 2016.
APMO Silver Medal
Asia Pacific Mathematical Olympiad.
IEEExtreme
World rank 52 and country rank 1.
SLMO Gold
Four Sri Lanka Mathematical Olympiad gold medals.
Beyond the papers
Strong systems need range.
Contact / collaborate