Talk Title: Statistical inference over networks: Decentralized optimization meets high-dimensional statistics
Abstract: The interest in solving large-scale statistical learning problems within decentralized networks is rapidly increasing, particularly in systems where data is distributed across network nodes without centralized coordination, often referred to as "mesh" networks. Inference from massive datasets presents critical challenges at the intersection of computational and statistical sciences, particularly ensuring the quality of the performed analytic when computational resources, like time and communication, are constrained. While the trade-offs between statistical accuracy and computational efficiency have been extensively studied in centralized settings, our comprehension in the context of mesh networks remains underdeveloped. Specifically, (i) distributed schemes that perform effectively in classical low-dimensional regimes can fail in high-dimensional scenarios, and (ii) existing convergence analyses may not accurately predict algorithmic behavior, with empirical tests often contradicting theoretical findings. This discrepancy arises primarily because most decentralized algorithms have been developed and analyzed primarily from an optimization perspective, neglecting the statistical aspects. This talk will present novel analyses and decentralized algorithm designs through various vignettes from high-dimensional statistical inference, aiming to integrate statistical considerations into decentralized optimization. By adopting this new perspective, some long-standing myths in the literature of distributed optimization will be demystified.
Bio: Gesualdo Scutari is the Pedro and Barbara Granadillo Professor in the School of Industrial Engineering and Electrical and Computer Engineering at Purdue University, West Lafayette, IN, USA. His research interests include continuous (distributed, stochastic) optimization, equilibrium programming, and their applications to signal processing and statistical learning. Among others, he was a recipient of the 2013 NSF CAREER Award, the 2015 IEEE Signal Processing Society Young Author Best Paper Award, and the 2020 IEEE Signal Processing Society Best Paper Award. He serves as an IEEE Signal Processing Distinguish Lecturer (2023-2024). He served on the editorial broad of several IEEE journals, and he is currently an Associate Editor of SIAM Journal on Optimization. He is a fellow of IEEE.
