Resource Efficient Distributed Learning in Large Scale Systems
Date and time: 30 May 2023, 13:00 – 14:00 CEST (UTC +2)
Speaker: Sindri Magnússon, Associate Professor, Department of Computer and Systems Sciences, Stockholm University
Title: Resource Efficient Distributed Learning in Large Scale Systems
Where: Digital Futures hub, Osquars Backe 5, floor 2 at KTH main campus
Meeting ID: 695 6088 7455
Moderator: Xuyang Wu, Digital Future postdoc
Administrator: Beatrice Vincenzi, email@example.com
Abstract: Distributed machine learning algorithms are becoming increasingly important for training large-scale models and for operating large-scale systems. For example, deep neural networks are now typically trained on multiple CPUs or GPUs. Moreover, data is increasingly being collected and processed in IoT networks (e.g., smartphones, home appliances, and wireless sensors). Building efficient distributed algorithms for these systems introduces new challenges as it shifts the hardness from the computation to the coordination of the compute nodes. Moreover, real-world networks have various constraints, e.g., in terms of limited energy/communication resources, safety or other physical limits, that must be met during distributed training and execution.
In this talk, I give a broad overview of our ongoing work on resource-efficient and safe distributed learning and its application in cyber-physical systems such as smart grids, water distribution networks, and smart transportation.
Bio: Sindri Magnússon is an Associate Professor at the Department of Computer and Systems Science, Stockholm University, Sweden. He received a B.Sc. degree in Mathematics from the University of Iceland, Reykjavík, Iceland, in 2011, the Master’s degree in Applied Mathematics (Optimization and Systems Theory) from KTH Royal Institute of Technology, Stockholm, Sweden, in 2013, and a PhD in Electrical Engineering from the same institution, in 2017. He was a postdoctoral researcher from 2018-2019 at Harvard University, Cambridge, MA and a visiting PhD student at Harvard University for 9 months in 2015 and 2016. His research interests include distributed optimization, machine learning and data-driven decision-making, both theory and applications in complex networks.
Link to speaker’s webpage: https://sites.google.com/view/sindri-magnusson
Link to LinkedIn account: https://www.linkedin.com/in/sindri-magn%C3%BAsson-616b7025/