Seminar - Distributed Processing
School of Engineering and Computer Science Seminar
Speaker: Thomas Sherson (Delft University of Technology)
Time: Monday 2nd October 2017 at 03:10 PM - 04:00 PM
Location: Cotton Club, Cotton 350
Over the last few decades, methods of parallel and distributed computation have become essential tools in a wide range of applications such as machine learning, wireless sensor network processing and big data signal processing. In this vein, in this presentation we highlight the role that distributed convex optimisation can play in these efforts. We further demonstrate how a number of distributed optimisation algorithms from within the literature can be derived via their relation to monotone operator theory including the likes of the proximal gradient algorithm, the alternating method of multipliers (ADMM) as well as more recent approaches such as the primal dual method of multipliers (PDMM). Finally we highlight some of the practical benefits of these methods (resilience to packet loss, asynchronous operation, the effect of quantisation on convergence, etc) as well as some open questions for future research.