Global Optimization via Schrödinger–Föllmer Diffusion

DOI10.1137/22M1477738MaRDI QIDQ6057791zbMATH OpenOpenAlexFDO

Authors Yin Dai, Yu Ling Jiao, Lican Kang, Xiliang Lu, Jerry Zhijian Yang

Publication date 26 October 2023

Published in SIAM Journal on Control and Optimization (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2111.00402

zbMATH Keywords

global optimizationstochastic approximationSchrödinger-Föllmer diffusion

Mathematics Subject Classification ID

Computational methods for problems pertaining to probability theory (60-08) Nonconvex programming, global optimization (90C26)

Abstract: We study the problem of finding global minimizers of $V(x):mathbb{R}^{d}ightarrowmathbbR$ approximately via sampling from a probability distribution $m{u}_{sigma}$ with density $p_{sigma}(x)=dfracexp(-V(x)/sigma)in{t}_{mathbb{R}^d}exp(-V(y)/sigma)dy$ with respect to the Lebesgue measure for $sigmain(0,1]$ small enough. We analyze a sampler based on the Euler-Maruyama discretization of the Schr{"o}dinger-F{"o}llmer diffusion processes with stochastic approximation under appropriate assumptions on the step size $s$ and the potential $V$. We prove that the output of the proposed sampler is an approximate global minimizer of $V(x)$ with high probability at cost of sampling $mathcalO(d^3)$ standard normal random variables. Numerical studies illustrate the effectiveness of the proposed method and its superiority to the Langevin method.

Recommendations

• Diffusions for Global Optimization
• Diffusion for Global Optimization in $\mathbb{R}^n $
• Convergence analysis of a global optimization algorithm using stochastic differential equations
• Global optimization and stochastic differential equations
• Publication:4509896

Cites work

• scientific article; zbMATH DE number 3896050 (Why is no real title available?)
• scientific article; zbMATH DE number 3932149 (Why is no real title available?)
• scientific article; zbMATH DE number 4080537 (Why is no real title available?)
• scientific article; zbMATH DE number 50805 (Why is no real title available?)
• scientific article; zbMATH DE number 1153603 (Why is no real title available?)
• scientific article; zbMATH DE number 3438144 (Why is no real title available?)
• scientific article; zbMATH DE number 3005682 (Why is no real title available?)
• scientific article; zbMATH DE number 2150787 (Why is no real title available?)
• A Generalization of Laplaces's Method
• A Relationship Between Arbitrary Positive Matrices and Doubly Stochastic Matrices
• A consensus-based global optimization method for high dimensional machine learning problems
• A stochastic control approach to reciprocal diffusion processes
• A survey of the Schrödinger problem and some of its connections with optimal transport
• Accelerated information gradient flow
• Analysis and geometry of Markov diffusion operators
• Asymptotic probabilities and differential equations
• Asymptotics of the spectral gap with applications to the theory of simulated annealing
• Bayesian learning via neural Schrödinger-Föllmer flows
• Computational optimal transport. With applications to data sciences
• Convergence of Langevin MCMC in KL-divergence
• Convergence rates for annealing diffusion processes
• Coupling and convergence for Hamiltonian Monte Carlo
• Diffusion for Global Optimization in $\mathbb{R}^n $
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• Improved bounds for discretization of Langevin diffusions: near-optimal rates without convexity
• Interacting Langevin diffusions: gradient structure and ensemble Kalman sampler
• Laplace's method revisited: Weak convergence of probability measures
• Log-Sobolev inequalities: different roles of Ric and Hess
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• Nonasymptotic estimates for stochastic gradient Langevin dynamics under local conditions in nonconvex optimization
• On Stochastic Gradient Langevin Dynamics with Dependent Data Streams: The Fully Nonconvex Case
• On stochastic gradient Langevin dynamics with dependent data streams in the logconcave case
• Poincaré and logarithmic Sobolev inequalities by decomposition of the energy landscape
• Rate of convergence for ergodic continuous Markov processes: Lyapunov versus Poincaré
• Representation formula for the entropy and functional inequalities
• Sampling can be faster than optimization
• Simulation and Inference for Stochastic Processes with YUIMA
• Smooth minimization of non-smooth functions
• Stability of the logarithmic Sobolev inequality via the Föllmer process
• Stein's lemma for elliptical random vectors
• Stochastic control liaisons. Richard Sinkhorn meets Gaspard Monge on a Schrödinger bridge
• Stochastic processes and applications. Diffusion processes, the Fokker-Planck and Langevin equations
• The Markov processes of Schr�dinger
• Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
• Unbiased estimation using a class of diffusion processes
• User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient

Cited in

(2)

• Weak approximation of Schrödinger-Föllmer diffusion
• Global optimization using diffusion perturbations with large noise intensity

This page was built for publication: Global Optimization via Schrödinger–Föllmer Diffusion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6057791)