Bayesian learning via neural Schrödinger-Föllmer flows

DOI10.1007/S11222-022-10172-5MaRDI QIDQ2104005zbMATH OpenOpenAlexFDO

Authors Francisco J. Vargas, Andrius Ovsianas, David Fernandes, Neil D. Lawrence, Nikolas Nüsken, M. Girolami

Publication date 9 December 2022

Published in Statistics and Computing (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2111.10510

zbMATH Keywords

Bayesian inferencestochastic controlBayesian deep learningSchrödinger bridge problemFöllmer drift

Mathematics Subject Classification ID

Computational methods for problems pertaining to statistics (62-08) Bayesian inference (62F15) Artificial neural networks and deep learning (68T07)

Abstract: In this work we explore a new framework for approximate Bayesian inference in large datasets based on stochastic control (i.e. Schr"odinger bridges). We advocate stochastic control as a finite time and low variance alternative to popular steady-state methods such as stochastic gradient Langevin dynamics (SGLD). Furthermore, we discuss and adapt the existing theoretical guarantees of this framework and establish connections to already existing VI routines in SDE-based models.

Recommendations

• Bayesian Inference Based on Stationary Fokker-Planck Sampling
• Manifold stochastic dynamics for Bayesian learning
• Stein variational gradient descent with local approximations
• scientific article; zbMATH DE number 6860839
• scientific article

Cites work

• scientific article; zbMATH DE number 6377992 (Why is no real title available?)
• scientific article; zbMATH DE number 646825 (Why is no real title available?)
• scientific article; zbMATH DE number 3005682 (Why is no real title available?)
• 10.1162/jmlr.2003.3.4-5.993
• A stochastic control approach to reciprocal diffusion processes
• A variational representation for certain functionals of Brownian motion
• Automatic differentiation of algorithms
• Bayesian Estimation of Linear Mixtures Using the Normal Compositional Model. Application to Hyperspectral Imagery
• Bayesian canonical correlation analysis
• Data assimilation: the Schrödinger perspective
• Existence of strong solutions for Itô's stochastic equations via approximations
• Exponential convergence of Langevin distributions and their discrete approximations
• From the Schrödinger problem to the Monge-Kantorovich problem
• Handbook of Markov Chain Monte Carlo
• MCMC using Hamiltonian dynamics
• Monte Carlo sampling methods using Markov chains and their applications
• Reconciling modern machine-learning practice and the classical bias-variance trade-off
• Riemann manifold Langevin and Hamiltonian Monte Carlo methods. With discussion and authors' reply
• Sequential Monte Carlo Methods in Practice
• Solving high-dimensional Hamilton-Jacobi-Bellman PDEs using neural networks: perspectives from the theory of controlled diffusions and measures on path space
• Stochastic control and nonequilibrium thermodynamical systems
• The <scp>Data‐Driven</scp> Schrödinger Bridge
• Variational Inference for Stochastic Differential Equations

Cited in

(8)

• Global Optimization via Schrödinger–Föllmer Diffusion
• Hybrid deterministic-stochastic gradient Langevin dynamics for Bayesian learning
• Control of neural transport for normalising flows
• Deterministic particle flows for stochastic nonlinear systems
• Optimal control of underdamped systems: an analytic approach
• Variational inference in neural functional prior using normalizing flows: application to differential equation and operator learning problems
• Manifold stochastic dynamics for Bayesian learning
• Diffusion Schrödinger bridges for Bayesian computation

Describes a project that uses

Uses Software

• GitHub
• PyTorch
• CIFAR
• Delve Datasets
• STRUCTURE
• MNIST

This page was built for publication: Bayesian learning via neural Schrödinger-Föllmer flows

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2104005)