Decentralized Bayesian learning with Metropolis-adjusted Hamiltonian Monte Carlo

DOI10.1007/S10994-023-06345-6arXiv2107.07211OpenAlexW3180796377MaRDI QIDQ6134345FDOQ6134345

Authors: Vyacheslav Kungurtsev, Adam D. Cobb, Tara Javidi, Brian A. Jalaian 

Publication date: 22 August 2023

Published in: Machine Learning (Search for Journal in Brave)

Abstract: Federated learning performed by a decentralized networks of agents is becoming increasingly important with the prevalence of embedded software on autonomous devices. Bayesian approaches to learning benefit from offering more information as to the uncertainty of a random quantity, and Langevin and Hamiltonian methods are effective at realizing sampling from an uncertain distribution with large parameter dimensions. Such methods have only recently appeared in the decentralized setting, and either exclusively use stochastic gradient Langevin and Hamiltonian Monte Carlo approaches that require a diminishing stepsize to asymptotically sample from the posterior and are known in practice to characterize uncertainty less faithfully than constant step-size methods with a Metropolis adjustment, or assume strong convexity properties of the potential function. We present the first approach to incorporating constant stepsize Metropolis-adjusted HMC in the decentralized sampling framework, show theoretical guarantees for consensus and probability distance to the posterior stationary distribution, and demonstrate their effectiveness numerically on standard real world problems, including decentralized learning of neural networks which is known to be highly non-convex.

Full work available at URL: https://arxiv.org/abs/2107.07211

zbMATH Keywords

Monte CarloHMCfederated learningdecentralized learning

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Cites Work

• Exponential convergence of Langevin distributions and their discrete approximations
• Hedonic housing prices and the demand for clean air
• Consistency and fluctuations for stochastic gradient Langevin dynamics
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• Coupling and convergence for Hamiltonian Monte Carlo
• Distributed stochastic gradient tracking methods
• Title not available (Why is that?)
• Balancing Communication and Computation in Distributed Optimization
• Stochastic gradient Hamiltonian Monte Carlo for non-convex learning
• Title not available (Why is that?)
• Global Convergence of Stochastic Gradient Hamiltonian Monte Carlo for Nonconvex Stochastic Optimization: Nonasymptotic Performance Bounds and Momentum-Based Acceleration

Cited In (1)

• Distributed event-triggered unadjusted Langevin algorithm for Bayesian learning

This page was built for publication: Decentralized Bayesian learning with Metropolis-adjusted Hamiltonian Monte Carlo

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6134345)