Discrete-time gradient flows and law of large numbers in Alexandrov spaces (Q745561): Difference between revisions

From MaRDI portal
Importer (talk | contribs)
Changed an Item
ReferenceBot (talk | contribs)
Changed an Item
 
Property / cites work
 
Property / cites work: Riemannian $L^{p}$ center of mass: Existence, uniqueness, and convexity / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q5441008 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Barycenters of measures transported by stochastic flows / rank
 
Normal rank
Property / cites work
 
Property / cites work: Stochastic algorithms for computing means of probability measures / rank
 
Normal rank
Property / cites work
 
Property / cites work: The proximal point algorithm in metric spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Computing Medians and Means in Hadamard Spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Finite termination of the proximal point method for convex functions on Hadamard manifolds / rank
 
Normal rank
Property / cites work
 
Property / cites work: Incremental proximal methods for large scale convex optimization / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q4257216 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Positive definite matrices / rank
 
Normal rank
Property / cites work
 
Property / cites work: Riemannian geometry and matrix geometric means / rank
 
Normal rank
Property / cites work
 
Property / cites work: Produits infinis de resolvantes / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q2731895 / rank
 
Normal rank
Property / cites work
 
Property / cites work: CAT\((k)\)-spaces, weak convergence and fixed points / rank
 
Normal rank
Property / cites work
 
Property / cites work: Proximal Point Algorithm On Riemannian Manifolds / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q2839722 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Equilibrium maps between metric spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Convex functionals and generalized harmonic maps into spaces of non positive curvature / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q4380377 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Monotonic properties of the least squares mean / rank
 
Normal rank
Property / cites work
 
Property / cites work: Riemannian center of mass and mollifier smoothing / rank
 
Normal rank
Property / cites work
 
Property / cites work: Probability, Convexity, and Harmonic Maps with Small Image I: Uniqueness and Fine Existence / rank
 
Normal rank
Property / cites work
 
Property / cites work: Convexity and the Hemisphere / rank
 
Normal rank
Property / cites work
 
Property / cites work: Random LSC Functions: An Ergodic Theorem / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3654935 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Monotone vector fields and the proximal point algorithm on Hadamard manifolds / rank
 
Normal rank
Property / cites work
 
Property / cites work: Weighted deterministic walks for the least squares mean on Hadamard spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Open map theorem for metric spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Gradient flows on nonpositively curved metric spaces and harmonic maps / rank
 
Normal rank
Property / cites work
 
Property / cites work: Means and Averaging in the Group of Rotations / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q2752037 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Incremental Subgradient Methods for Nondifferentiable Optimization / rank
 
Normal rank
Property / cites work
 
Property / cites work: Convexities of metric spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Gradient flows on Wasserstein spaces over compact Alexandrov spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Barycenters in Alexandrov spaces of curvature bounded below / rank
 
Normal rank
Property / cites work
 
Property / cites work: Semiconcave functions in Alexandrov's geometry / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q4235027 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Variational Analysis / rank
 
Normal rank
Property / cites work
 
Property / cites work: Nonlinear Markov operators associated with symmetric Markov kernels and energy minimizing maps between singular spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Nonlinear Markov operators, discrete heat flow, and harmonic maps between singular spaces / rank
 
Normal rank
Property / cites work
 
Property / cites work: Nonlinear martingale theory for processes with values in metric spaces of nonpositive curvature / rank
 
Normal rank
Property / cites work
 
Property / cites work: A semigroup approach to harmonic maps / rank
 
Normal rank
Property / cites work
 
Property / cites work: On the geometry of metric measure spaces. I / rank
 
Normal rank

Latest revision as of 21:12, 10 July 2024

scientific article
Language Label Description Also known as
English
Discrete-time gradient flows and law of large numbers in Alexandrov spaces
scientific article

    Statements

    Discrete-time gradient flows and law of large numbers in Alexandrov spaces (English)
    0 references
    0 references
    0 references
    14 October 2015
    0 references
    Let \(\left( M,\rho\right) \) be an Alexandrov space \(\left( M,\rho\right) \) of curvature \(\leq K\) (in which minimal geodesics depend continuously on their endpoints, i.e, \(\left( M,\rho\right) \) is an Alexandrov \(\Re_{K}\) domain also known as \(\text{CAT}\left( K\right) \) space) or an Alexandrov space of curvature \(\geq K^{\prime}\). The authors study discrete-time gradient flows for finding a point of minimum of a convex function \(f:M\rightarrow(-\infty,+\infty]\). The discrete-time gradient flow is constructed by means of the (metric) resolvent \(J_{\lambda }^{f}:M\rightarrow M\) by setting \(x_{k+1}=J_{\lambda}^{f}\left( x_{k}\right) \), \(k\in \mathbb{N}\) where \(\lambda>0\). In the case of curvature bounded above, the familiar Moreau-Yosida resolvent is used: \(J_{\lambda}^{f}\left( x\right) =\text{argmin}_{y\in M}\left\{ f\left( y\right) +\frac{1}{2\lambda }d^{2}\left( x,y\right) \right\} \); for \(K=0\), see [\textit{J. Jost}, Comment. Math. Helv. 70, No. 4, 659--673 (1995; Zbl 0852.58022)]. In the case of the lower curvature bound, \(J_{\lambda}^{f}\left( x\right) =\text{gexp}\left( \lambda\nabla\left( -f\right) \left( x\right) \right) \) where \(\text{gexp}\) is the gradient exponential map (Section 4). In Section 5, the authors generalize the result from [\textit{M. Bačák}, Isr. J. Math. 194, Part B, 689--701 (2013; Zbl 1278.49039)] by proving the following theorem for complete \(\Re_{K}\) domains: If \(f:M\rightarrow(-\infty,+\infty]\) is a convex, lower semi-continuous function, \(G\subseteq M\) is a closed, geodesically convex set containing a sublevel set of \(f\) satisfying \(\text{diam}\left( G\right) <\pi/2\sqrt{K}\) for positive \(K\), \(\left( \lambda_{k}\right) _{k=1}^{\infty}\) is a sequence of positive numbers such that \(\sum _{k=1}^{\infty}\lambda_{k}=+\infty\), \(x_{0}\in G\) and \(x_{k}=J_{\lambda_{k} }^{f}\left( x_{k-1}\right) \), \(k\in\mathbb{N}\), then \(\lim_{k\rightarrow\infty}f\left( x_{k}\right) =\inf_{y\in G}f\left( y\right) \). The authors also consider the case when \(f=\sum _{k=0}^{n}f_{k}\) where \(f_{k}\) are convex and Lipschitz in a locally compact Alexandrov space of curvature either \(\leq K\) or \(\geq K^{\prime}\) \ and prove that the proximal point algorithm produces the sequence converging to a point of minima of \(f\) when it exists (Theorem 5.5). In Section 6, the authors present results related to stochastic discrete-time gradient flow for a convex infinite (integral form) combination of convex functions and generalize Sturm's law of large numbers to Alexandrov spaces of curvature either \(\leq K\) or \(\geq K^{\prime}\) for arbitrary \(K\) and \(K^{\prime}\in\mathbb{R}\).
    0 references
    Alexandrov spaces
    0 references
    convex functions
    0 references
    resolvent
    0 references
    law of large numbers
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references