A variational finite volume scheme for Wasserstein gradient flows

被引:18
|
作者
Cances, Clement [1 ]
Gallouet, Thomas O. [2 ]
Todeschi, Gabriele [2 ]
机构
[1] Univ Lille, INRIA, CNRS, UMR 8524,Lab Paul Painleve, F-59000 Lille, France
[2] Univ Paris Dauphine, PSL Res Univ, CEREMADE, CNRS,INRIA,Project Team Mokaplan,UMR 7534, Ceremade, France
基金
欧盟地平线“2020”;
关键词
49M29; 35K65; 65M08; 65M12; PARABOLIC EQUATIONS; NUMERICAL-ANALYSIS; LAGRANGIAN SCHEME; POROUS-MEDIA; CROWD MOTION; DIFFUSION; CONVERGENCE; MODEL; DISCRETIZATION; DISTANCE;
D O I
10.1007/s00211-020-01153-9
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
We propose a variational finite volume scheme to approximate the solutions to Wasserstein gradient flows. The time discretization is based on an implicit linearization of the Wasserstein distance expressed thanks to Benamou-Brenier formula, whereas space discretization relies on upstream mobility two-point flux approximation finite volumes. The scheme is based on a first discretize then optimize approach in order to preserve the variational structure of the continuous model at the discrete level. It can be applied to a wide range of energies, guarantees non-negativity of the discrete solutions as well as decay of the energy. We show that the scheme admits a unique solution whatever the convex energy involved in the continuous problem, and we prove its convergence in the case of the linear Fokker-Planck equation with positive initial density. Numerical illustrations show that it is first order accurate in both time and space, and robust with respect to both the energy and the initial profile.
引用
收藏
页码:437 / 480
页数:44
相关论文
共 50 条
  • [1] A variational finite volume scheme for Wasserstein gradient flows
    Clément Cancès
    Thomas O. Gallouët
    Gabriele Todeschi
    Numerische Mathematik, 2020, 146 : 437 - 480
  • [2] Variational inference via Wasserstein gradient flows
    Lambert, Marc
    Chewi, Sinho
    Bach, Francis
    Bonnabel, Silvere
    Rigollet, Philippe
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022), 2022,
  • [3] FROM GEODESIC EXTRAPOLATION TO A VARIATIONAL BDF2 SCHEME FOR WASSERSTEIN GRADIENT FLOWS
    Gallouet, Thomas o.
    Natale, Andrea
    Todeschi, Gabriele
    MATHEMATICS OF COMPUTATION, 2024, 93 (350) : 2769 - 2810
  • [4] Variational Wasserstein gradient flow
    Fan, Jiaojiao
    Zhang, Qinsheng
    Taghvaei, Amirhossein
    Chen, Yongxin
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 162, 2022,
  • [5] Approximate Inference with Wasserstein Gradient Flows
    Frogner, Charlie
    Poggio, Tomaso
    INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 108, 2020, 108 : 2581 - 2589
  • [6] Policy Optimization as Wasserstein Gradient Flows
    Zhang, Ruiyi
    Chen, Changyou
    Li, Chunyuan
    Carin, Lawrence
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 80, 2018, 80
  • [7] Entropic Approximation of Wasserstein Gradient Flows
    Peyre, Gabriel
    SIAM JOURNAL ON IMAGING SCIENCES, 2015, 8 (04): : 2323 - 2351
  • [8] A VARIATIONAL ANALYSIS FOR THE MOVING FINITE ELEMENT METHOD FOR GRADIENT FLOWS*
    Xu, Xianmin
    JOURNAL OF COMPUTATIONAL MATHEMATICS, 2023, 41 (02): : 191 - 210
  • [9] TPFA Finite Volume Approximation ofWasserstein Gradient Flows
    Natale, Andrea
    Todeschi, Gabriele
    FINITE VOLUMES FOR COMPLEX APPLICATIONS IX-METHODS, THEORETICAL ASPECTS, EXAMPLES, FVCA 9, 2020, 323 : 193 - 201
  • [10] {Euclidean, metric, and Wasserstein} gradient flows: an overview
    Santambrogio, Filippo
    BULLETIN OF MATHEMATICAL SCIENCES, 2017, 7 (01) : 87 - 154