Abstract: We study a natural Wasserstein gradient flow on manifolds of probability distributions with discrete sample spaces. We derive the Riemannian structure for the probability simplex from the dynamical formulation of the Wasserstein distance on a weighted graph. We pull back the geometric structure to the parameter space of any given probability model, which allows us to define a natural gradient flow there. In contrast to the natural Fisher-Rao gradient, the natural Wasserstein gradient incorporates a ground metric on sample space. We illustrate the analysis of elementary exponential family examples and demonstrate an application of the Wasserstein natural gradient to maximum likelihood estimation.
Recommendations
Cites work
- scientific article; zbMATH DE number 3897897 (Why is no real title available?)
- scientific article; zbMATH DE number 3761167 (Why is no real title available?)
- scientific article; zbMATH DE number 3894218 (Why is no real title available?)
- A computational fluid mechanics solution to the Monge-Kantorovich mass transfer problem
- A gradient structure for reaction–diffusion systems and for energy-drift-diffusion systems
- An Extended Cencov Characterization of the Information Metric
- Axiomatic Geometry of Conditional Models
- Computations of optimal transport distance with Fisher information regularization
- Constrained steepest descent in the 2-Wasserstein metric
- Entropy dissipation of Fokker-Planck equations on graphs
- Fokker-Planck equations for a free energy functional or Markov process on a graph
- Geodesics of minimal length in the set of probability measures on graphs
- Geometry of matrix decompositions seen through optimal transport and information geometry
- Gradient flows of the entropy for finite Markov chains
- Information geometry
- Information geometry and its applications
- Information geometry connecting Wasserstein distance and Kullback-Leibler divergence via the entropy-relaxed transportation problem
- Information geometry of Wasserstein divergence
- Natural gradient flow in the mixture geometry of a discrete exponential family
- On the Fisher metric of conditional probability polytopes
- Optimal Transport
- Ricci curvature for parametric statistics via optimal transport
- Some geometric calculations on Wasserstein space
- THE GEOMETRY OF DISSIPATIVE EVOLUTION EQUATIONS: THE POROUS MEDIUM EQUATION
- The Density Manifold and Configuration Space Quantization
- Towards the geometry of estimation of distribution algorithms based on the exponential family
- Wasserstein geometry of Gaussian measures
Cited in
(33)- Wasserstein gradients for the temporal evolution of probability distributions
- Quantum statistical learning via quantum Wasserstein natural gradient
- Information geometry of physics-informed statistical manifolds and its use in data assimilation
- Gradient flow of the stochastic relaxation on a generic exponential family
- Optimal transport natural gradient for statistical manifolds with continuous sample space
- Wasserstein proximal of GANs
- High order spatial discretization for variational time implicit schemes: Wasserstein gradient flows and reaction-diffusion systems
- When optimal transport meets information geometry
- Invariance properties of the natural gradient in overparametrised systems
- Efficient Natural Gradient Descent Methods for Large-Scale PDE-Based Optimization Problems
- Fisher information regularization schemes for Wasserstein gradient flows
- Accelerated information gradient flow
- Natural gradient for combined loss using wavelets
- Analysis of asymptotic escape of strict saddle sets in manifold optimization
- Pseudo-Riemannian geometry encodes information geometry in optimal transport
- Transport information geometry: Riemannian calculus on probability simplex
- Weyl geometric approach to the gradient-flow equations in information geometry
- Wasserstein information matrix
- Information geometry of smooth densities on the Gaussian space: Poincaré inequalities
- Interacting Langevin diffusions: gradient structure and ensemble Kalman sampler
- Hessian transport gradient flows
- Neural parametric Fokker-Planck equation
- Hessian metric via transport information geometry
- Ricci curvature for parametric statistics via optimal transport
- The information geometry of mirror descent
- Affine statistical bundle modeled on a Gaussian Orlicz-Sobolev space
- Geometry and convergence of natural policy gradient methods
- Affine natural proximal learning
- Transport information Bregman divergences
- Online natural gradient as a Kalman filter
- Mirror descent algorithms for minimizing interacting free energy
- Lagrangian and Hamiltonian dynamics for probabilities on the statistical bundle
- Sparse optimization on measures with over-parameterized gradient descent
This page was built for publication: Natural gradient via optimal transport
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1713655)
