Revisiting the Effects of Stochasticity for Hamiltonian Samplers

06/30/2021
∙
by   Giulio Franzese, et al.
∙
0
∙

We revisit the theoretical properties of Hamiltonian stochastic differential equations (SDES) for Bayesian posterior sampling, and we study the two types of errors that arise from numerical SDE simulation: the discretization error and the error due to noisy gradient estimates in the context of data subsampling. Our main result is a novel analysis for the effect of mini-batches through the lens of differential operator splitting, revising previous literature results. The stochastic component of a Hamiltonian SDE is decoupled from the gradient noise, for which we make no normality assumptions. This leads to the identification of a convergence bottleneck: when considering mini-batches, the best achievable error rate is 𝒪(η^2), with η being the integrator step size. Our theoretical results are supported by an empirical study on a variety of regression and classification tasks for Bayesian neural networks.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
∙ 03/29/2018

Stochastic Gradient Hamiltonian Monte Carlo with Variance Reduction for Bayesian Inference

Gradient-based Monte Carlo sampling algorithms, like Langevin dynamics a...
research
∙ 10/14/2020

Symplectic method for Hamiltonian stochastic differential equations with multiplicative Lévy noise in the sense of Marcus

A class of Hamiltonian stochastic differential equations with multiplica...
research
∙ 02/15/2021

Some approximation results for mild solutions of stochastic fractional order evolution equations driven by Gaussian noise

We investigate the quality of space approximation of a class of stochast...
research
∙ 07/05/2019

Stochastic modified equations for symplectic methods applied to rough Hamiltonian systems based on the Wong--Zakai approximation

We investigate the stochastic modified equation which plays an important...
research
∙ 03/21/2023

Universal Approximation Property of Hamiltonian Deep Neural Networks

This paper investigates the universal approximation capabilities of Hami...
research
∙ 06/08/2020

Structure-preserving discretization of port-Hamiltonian plate models

Methods for discretizing port-Hamiltonian systems are of interest both f...
research
∙ 07/21/2021

Differentiable Annealed Importance Sampling and the Perils of Gradient Noise

Annealed importance sampling (AIS) and related algorithms are highly eff...

Please sign up or login with your details

Forgot password? Click here to reset