Keynote Speakers
Hear from leaders shaping federated learning and intelligent computing—spanning theory, systems, healthcare, privacy, and real-world deployments.
Peter Richtárik — Keynote Details
Communication-Efficient Optimization for Federated Learning (tentative)
Federated systems face fundamental trade-offs between communication cost, statistical efficiency, and robustness to client heterogeneity. This talk surveys recent advances in randomized and variance-reduced methods, compressed and quantized updates, and error-feedback mechanisms that enable provably faster convergence with partial participation and non-IID data. We discuss lower bounds, tight complexity results, and practical algorithms that bridge theory and system realities, highlighting open problems in personalization and adaptivity.
Peter Richtárik is a professor at KAUST whose work spans optimization and large-scale machine learning, with contributions to distributed and randomized algorithms widely adopted in ML systems.
Title and fine-grained abstract TBA.
Holger Roth — Keynote Details
Federated Foundation Models for Medical Imaging (tentative)
Clinical imaging pipelines are constrained by label scarcity, stringent privacy, and site variability. We present a practical recipe for federated pretraining and adaptation of imaging foundation models, combining self-supervised objectives, secure aggregation, and differential privacy. Case studies demonstrate robustness to domain shift and variability in acquisition protocols, and we outline integration patterns with clinical workflows and validation frameworks.
Holger Roth is a Principal Applied Research Scientist at NVIDIA focusing on deep learning for medical imaging. He has collaborated closely with clinicians and academics to build models for radiological applications and holds a Ph.D. from UCL.
Title and fine-grained abstract TBA.
Zheng Xu — Keynote Details
Federated Learning in the Wild: Privacy, Scale, and Personalization (tentative)
Operating FL at global scale introduces non-IID data, churn, resource variability, and regulatory constraints. This talk covers hierarchical aggregation, adaptive client selection, and on-device personalization under privacy budgets. We share lessons from large deployments, discuss auditability and policy alignment, and highlight open challenges in measurement, fairness, and robust evaluation.
Zheng Xu is a research scientist at Google working on federated learning and privacy. He earned his Ph.D. in optimization and machine learning from the University of Maryland and has co-organized community workshops across ICML/ICLR/KDD.
Title and fine-grained abstract TBA.
Be there for the keynotes
Secure your seat and engage directly with the speakers during Q&A.