|
Welcome to the Wireless Computing Lab! We are located
in the Department
of Electrical and Computer Engineering at the University
of Toronto. We are part of the Communications
Group. We conduct research into theories and
techniques of communication and computing. Our
research promotes the convergence of network
optimization, distributed computing, machine learning,
and mobile communications, toward seamless integration
between communication and computation in networked
systems. We aim to establish theoretical foundations
and develop efficient algorithms that make large-scale
intelligent networked systems more reliable, adaptive,
and resource-efficient.
Recent research topics:
- Communication-efficient distributed learning:
gradient sparsification, adaptive bit allocation,
client sampling, decentralized learning topology
design, online distributed optimization under
communication constraints
- Wireless federated learning and edge learning:
over-the-air federated learning, analog model
aggregation, multi-antenna beamforming, power
control, device selection, multi-model federated
learning, privacy in over-the-air aggregation
- Mobile edge and cloud computing:
computation offloading, edge inference with model
partitioning, multi-resource scheduling, fair
multi-resource allocation, cloud resource
procurement and pricing
- Wireless resource management: wireless
network virtualization, multi-group multicast
beamforming, online precoding with imperfect channel
state information, device-to-device communication,
stochastic geometry analysis of heterogeneous
cellular networks
- Online optimization and algorithms: online
convex optimization with long-term constraints and
delayed feedback, distributed online mirror descent,
online matching and scheduling, dynamic regret,
competitive-ratio analysis
- Machine learning methods and theory:
differentially private training, diffusion posterior
sampling, knowledge distillation, generalization of
SGD with momentum, Bayes-optimal classification
- Machine learning for communication networks:
network traffic classification, explainable
classifiers, robustness to adversarial manipulation
- Earlier work: mobility management in ad hoc
and cellular networks, network coding in wireless
mesh and sensor networks, cooperative relaying,
video and peer-to-peer streaming, smart-grid power
balancing
|