Accelerating diffusions

DOI10.1214/105051605000000025zbMATH Open1069.60065arXivmath/0505245OpenAlexW3037124231MaRDI QIDQ558681FDOQ558681

Authors: Jianyong Qiao, Sumit K. Garg 

Publication date: 13 July 2005

Published in: The Annals of Applied Probability (Search for Journal in Brave)

Abstract: Let U be a given function defined on R^d and pi(x) be a density function proportional to exp -U(x). The following diffusion X(t) is often used to sample from pi(x), dX(t)=-

abla U(X(t)) dt+sqrt2 dW(t),qquad X(0)=x_0. To accelerate the convergence, a family of diffusions with pi(x) as their common equilibrium is considered, dX(t)=�igl(- abla U(X(t))+C(X(t))�igr) dt+sqrt2 dW(t),qquad X(0)=x_0. Let L_C be the corresponding infinitesimal generator. The spectral gap of L_C in L^2(pi) (lambda (C)), and the convergence exponent of X(t) to pi in variational norm ( ho(C)), are used to describe the convergence rate, where lambda(C)= Sup{real part of mudvtxmu is in the spectrum of L_C, mu is not zero}, {-2.8cm} ho(C) = Inf�iggl{ hodvtxint | p(t,x,y) -pi(y)| dy le g(x) e^{ ho t}�iggr}.Roughly speaking, L_C is a perturbation of the self-adjoint L_0 by an antisymmetric operator Ccdot

abla, where C is weighted divergence free. We prove that lambda (C)le lambda (0) and equality holds only in some rare situations. Furthermore, ho(C)le lambda (C) and equality holds for C=0. In other words, adding an extra drift, C(x), accelerates convergence. Related problems are also discussed.

Full work available at URL: https://arxiv.org/abs/math/0505245

Recommendations

• Accelerating Gaussian diffusions
• Langevin-type models. I: Diffusions with given stationary distributions and their discretizations
• Accelerating planar Ornstein-Uhlenbeck diffusion with suitable drift
• Attaining the optimal Gaussian diffusion acceleration
• On how to use drift to push the spectral gap of a diffusion on \(S^{2}\) to infinity

zbMATH Keywords

ergodicityMCMCconvergence ratespectrumaccelerationspectral gapvariational normMonte Carlo Markov process

Mathematics Subject Classification ID

Markov semigroups and applications to diffusion processes (47D07) Diffusion processes (60J60) General topics in linear spectral theory for PDEs (35P05) Acceleration of convergence in numerical analysis (65B99)

Cites Work

• Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images
• Title not available (Why is that?)
• General state space Markov chains and MCMC algorithms
• Rates of convergence of the Hastings and Metropolis algorithms
• Perturbation theory for linear operators.
• Title not available (Why is that?)
• One-parameter semigroups of positive operators
• ON REGULARITY OF TRANSITION PROBABILITIES AND INVARIANT MEASURES OF SINGULAR DIFFUSIONS UNDER MINIMAL CONDITIONS
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Weak Poincaré inequalities and \(L^2\)-convergence rates of Markov semigroups
• Title not available (Why is that?)
• Accelerating Gaussian diffusions
• Pointwise estimates and quasilinear parabolic equations
• Optimal spectral structure of reversible stochastic matrices, Monte Carlo methods and the simulation of Markov random fields
• On the convergence of the Markov chain simulation method
• Title not available (Why is that?)
• On rates of convergence of stochastic relaxation for Gaussian and non- Gaussian distributions
• Title not available (Why is that?)
• Comparing sweep strategies for stochastic relaxation
• On the geometrical convergence of Gibbs sampler in \(\mathbb R^d\)
• Existence of the spectral gap for elliptic operators
• Convergence properties of the Gibbs sampler for perturbations of Gaussians
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)

Cited In (53)

• Characterising the nonequilibrium stationary states of Ornstein-Uhlenbeck processes
• Optimal variance reduction for Markov chain Monte Carlo
• The behavior of the spectral gap under growing drift
• Dirichlet eigenvalue problems of irreversible Langevin diffusion
• The speed of relaxation for diffusion with drift satisfying exponential decay of correlations
• Accelerating reversible Markov chains
• A trajectorial interpretation of the dissipations of entropy and Fisher information for stochastic differential equations
• On some mixing times for nonreversible finite Markov chains
• Analysis of multiscale integrators for multiple attractors and irreversible Langevin samplers
• On hitting time, mixing time and geometric interpretations of Metropolis-Hastings reversiblizations
• Accelerating diffusion by incompressible drift on the two-dimensional torus
• Projected Wasserstein Gradient Descent for High-Dimensional Bayesian Inference
• Curl flux induced drift in stochastic differential equations in the zero-mass limit
• Generalisation of the Eyring-Kramers transition rate formula to irreversible diffusion processes
• Constructing sampling schemes via coupling: Markov semigroups and optimal transport
• Accelerating diffusion on compact Riemannian surfaces by incompressible drift
• Variance reduction using nonreversible Langevin samplers
• Title not available (Why is that?)
• Optimal non-reversible linear drift for the convergence to equilibrium of a diffusion
• Large deviations of empirical measures of diffusions in weighted topologies
• Variance reduction for diffusions
• Long‐Time Behaviour of Langevin Algorithms with Time‐dependent Energy Function
• On the asymptotic variance of reversible Markov chain without cycles
• Using perturbed underdamped Langevin dynamics to efficiently sample from probability distributions
• Non-reversible metastable diffusions with Gibbs invariant measure. I: Eyring-Kramers formula
• Accelerating Gaussian diffusions
• Improving the convergence of reversible samplers
• Partial differential equations and stochastic methods in molecular dynamics
• Acceleration of convergence to equilibrium in Markov chains by breaking detailed balance
• Irreversible samplers from jump and continuous Markov processes
• On multiple acceleration of reversible Markov chain
• Non-reversible metastable diffusions with Gibbs invariant measure. II: Markov chain convergence
• A function space HMC algorithm with second order Langevin diffusion limit
• Accelerating planar Ornstein-Uhlenbeck diffusion with suitable drift
• Large deviations for the empirical measure of the zig-zag process
• Attaining the optimal Gaussian diffusion acceleration
• On the convergence time of some non-reversible Markov chain Monte Carlo methods
• Geometry-informed irreversible perturbations for accelerated convergence of Langevin dynamics
• On how to use drift to push the spectral gap of a diffusion on \(S^{2}\) to infinity
• Markov chain Monte Carlo and irreversibility
• Optimal scaling of the MALA algorithm with irreversible proposals for Gaussian targets
• Stochastic gradient descent and fast relaxation to thermodynamic equilibrium: a stochastic control approach
• Heat kernels for generalized Ornstein-Uhlenbeck operators
• Ensemble preconditioning for Markov chain Monte Carlo simulation
• Accelerated convergence to equilibrium and reduced asymptotic variance for Langevin dynamics using Stratonovich perturbations
• A note on the asymptotic variance of drift accelerated diffusions
• Title not available (Why is that?)
• Accelerating Brownian motion on \(N\)-torus
• The entropy production of stationary diffusions
• Filtering dynamical systems using observations of statistics
• Wasserstein contraction and Poincaré inequalities for elliptic diffusions with high diffusivity
• The free energy principle made simpler but not too simple
• A large deviation principle for the empirical measures of Metropolis-Hastings chains

This page was built for publication: Accelerating diffusions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q558681)