A blob method for inhomogeneous diffusion with applications to multi-agent control and sampling
From MaRDI portal
Publication:6133439
DOI10.1090/MCOM/3841zbMATH Open1525.35200arXiv2202.12927OpenAlexW4380591085MaRDI QIDQ6133439FDOQ6133439
Authors: Katy Craig, Karthik Elamvazhuthi, Olga Turanova 
Publication date: 18 August 2023
Published in: Mathematics of Computation (Search for Journal in Brave)
Abstract: As a counterpoint to classical stochastic particle methods for linear diffusion equations, we develop a deterministic particle method for the weighted porous medium equation (WPME) and prove its convergence on bounded time intervals. This generalizes related work on blob methods for unweighted porous medium equations. From a numerical analysis perspective, our method has several advantages: it is meshfree, preserves the gradient flow structure of the underlying PDE, converges in arbitrary dimension, and captures the correct asymptotic behavior in simulations. That our method succeeds in capturing the long time behavior of WPME is significant from the perspective of related problems in quantization. Just as the Fokker-Planck equation provides a way to quantize a probability measure by evolving an empirical measure according to stochastic Langevin dynamics so that the empirical measure flows toward , our particle method provides a way to quantize according to deterministic particle dynamics approximating WMPE. In this way, our method has natural applications to multi-agent coverage algorithms and sampling probability measures. A specific case of our method corresponds exactly to confined mean-field dynamics of training a two-layer neural network for a radial basis function activation function. From this perspective, our convergence result shows that, in the overparametrized regime and as the variance of the radial basis functions goes to zero, the continuum limit is given by WPME. This generalizes previous results, which considered the case of a uniform data distribution, to the more general inhomogeneous setting. As a consequence of our convergence result, we identify conditions on the target function and data distribution for which convexity of the energy landscape emerges in the continuum limit.
Full work available at URL: https://arxiv.org/abs/2202.12927
Recommendations
Artificial neural networks and deep learning (68T07) Asymptotic behavior of solutions to PDEs (35B40) PDEs in connection with fluid mechanics (35Q35) PDEs in connection with statistical mechanics (35Q82) Fokker-Planck equations (35Q84) PDEs with randomness, stochastic partial differential equations (35R60) Stability and convergence of numerical methods for initial value and initial-boundary value problems involving PDEs (65M12) Flows in porous media; filtration; seepage (76S05) Interacting particle systems in time-dependent statistical mechanics (82C22) Stochastic methods (Fokker-Planck, Langevin, etc.) applied to problems in time-dependent statistical mechanics (82C31) PDEs in connection with statistics (35Q62) PDEs in connection with computer science (35Q68) Multi-agent systems (93A16)
Cites Work
- Measure theory. Vol. I and II
- Optimal transport for applied mathematicians. Calculus of variations, PDEs, and modeling
- Gradient flows in metric spaces and in the space of probability measures
- Introduction to nonparametric estimation
- Asymptotics for transportation cost in high dimensions
- Foundations of quantization for probability distributions
- THE GEOMETRY OF DISSIPATIVE EVOLUTION EQUATIONS: THE POROUS MEDIUM EQUATION
- Numerical simulation of diffusive and aggregation phenomena in nonlinear continuity equations by evolving diffeomorphisms
- A Family of Nonlinear Fourth Order Equations of Gradient Flow Type
- Porous media equations with two weights: smoothing and decay properties of energy solutions via Poincaré inequalities
- The Variational Formulation of the Fokker--Planck Equation
- Title not available (Why is that?)
- Aggregation-diffusion equations: dynamics, asymptotics, and singular limits
- The Speed of Mean Glivenko-Cantelli Convergence
- Gamma-convergence of gradient flows on Hilbert and metric spaces and applications
- Convergence of a variational Lagrangian scheme for a nonlinear drift diffusion equation
- Distributed control of robotic networks: a mathematical approach to motion coordination algorithms.
- A finite-volume method for nonlinear nonlocal equations with a gradient flow structure
- Rigorous derivation of nonlinear scalar conservation laws from follow-the-leader type models via many particle limit
- A blob method for the aggregation equation
- A finite volume scheme for nonlinear degenerate parabolic equations
- A mixed finite element method for nonlinear diffusion equations
- A deterministic particle method for nonlinear diffusion equations
- Diffeomorphisms and Nonlinear Heat Flows
- Title not available (Why is that?)
- Title not available (Why is that?)
- Lectures on optimal transport
- Sharp asymptotic and finite-sample rates of convergence of empirical measures in Wasserstein distance
- Optimotaxis: A Stochastic Multi-agent Optimization Procedure with Point Measurements
- Convergence of a particle method for diffusive gradient flows in one dimension
- Variational particle schemes for the porous medium equation and for the system of isentropic Euler equations
- Numerical study of a particle method for gradient flows
- Discretization of functionals involving the Monge-Ampère operator
- Nonconvex gradient flow in the Wasserstein metric and applications to constrained nonlocal interactions
- Large systems of interacting particles and the porous medium equation
- \(L^q\)-functional inequalities and weighted porous media equations
- A discontinuous Galerkin method for nonlinear parabolic equations and gradient flow problems with interaction potentials
- Numerical simulation of nonlinear continuity equations by evolving diffeomorphisms
- Scaling Limit of the Stein Variational Gradient Descent: The Mean Field Regime
- The mathematical theories of diffusion: nonlinear and fractional diffusion
- A mean field view of the landscape of two-layer neural networks
- Machine learning from a continuous viewpoint. I
- A gradient flow approach to quantization of measures
- The role of a strong confining potential in a nonlinear Fokker-Planck equation
- Asymptotic optimality of the triangular lattice for a class of optimal location problems
- The nonlocal-interaction equation near attracting manifolds
- An invitation to optimal transport, Wasserstein distances, and gradient flows
- A blob method for diffusion
- Asymptotic analysis for a very fast diffusion equation arising from the 1D quantization problem
- Coupling and convergence for Hamiltonian Monte Carlo
- Centroidal Power Diagrams, Lloyd's Algorithm, and Applications to Optimal Location Problems
- Convergent Lagrangian discretization for drift-diffusion with nonlocal aggregation
- Quantifying Robotic Swarm Coverage
- Analysis of a two-layer neural network via displacement convexity
- Mean field analysis of neural networks: a law of large numbers
- Deterministic particle approximation of aggregation-diffusion equations on unbounded domains
- Primal dual methods for Wasserstein gradient flows
- Trainability and Accuracy of Artificial Neural Networks: An Interacting Particle System Approach
- A gradient flow perspective on the quantization problem
- Porous medium equation and cross-diffusion systems as limit of nonlocal interaction
- Convergence of a Lagrangian discretization for barotropic fluids and porous media flow
Cited In (1)
This page was built for publication: A blob method for inhomogeneous diffusion with applications to multi-agent control and sampling
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6133439)
