-
Optimal quantization with branched optimal transport distances
Authors:
Paul Pegon,
Mircea Petrache
Abstract:
We consider the problem of optimal approximation of a target measure by an atomic measure with $N$ atoms, in branched optimal transport distance. This is a new branched transport version of optimal quantization problems. New difficulties arise, since in classical semi-discrete optimal transport with Wasserstein distance, the interfaces between cells associated with neighboring atoms have Voronoi s…
▽ More
We consider the problem of optimal approximation of a target measure by an atomic measure with $N$ atoms, in branched optimal transport distance. This is a new branched transport version of optimal quantization problems. New difficulties arise, since in classical semi-discrete optimal transport with Wasserstein distance, the interfaces between cells associated with neighboring atoms have Voronoi structure and satisfy an explicit description. This description is missing for our problem, in which the cell interfaces are thought to have fractal boundary. We study the asymptotic behaviour of optimal quantizers for absolutely continuous measures as the number $N$ of atoms grows to infinity. We compute the limit distribution of the corresponding point clouds and show in particular a branched transport version of Zador's theorem. Moreover, we establish uniformity bounds of optimal quantizers in terms of separation distance and covering radius of the atoms, when the measure is $d$-Ahlfors regular. A crucial technical tool is the uniform in $N$ Hölder regularity of the landscape function, a branched transport analog to Kantorovich potentials in classical optimal transport.
△ Less
Submitted 30 March, 2025; v1 submitted 15 September, 2023;
originally announced September 2023.
-
Convergence rate of entropy-regularized multi-marginal optimal transport costs
Authors:
Luca Nenna,
Paul Pegon
Abstract:
We investigate the convergence rate of multi-marginal optimal transport costs that are regularized with the Boltzmann-Shannon entropy, as the noise parameter $\varepsilon$ tends to $0$. We establish lower and upper bounds on the difference with the unregularized cost of the form $C\varepsilon\log(1/\varepsilon)+O(\varepsilon)$ for some explicit dimensional constants $C$ depending on the marginals…
▽ More
We investigate the convergence rate of multi-marginal optimal transport costs that are regularized with the Boltzmann-Shannon entropy, as the noise parameter $\varepsilon$ tends to $0$. We establish lower and upper bounds on the difference with the unregularized cost of the form $C\varepsilon\log(1/\varepsilon)+O(\varepsilon)$ for some explicit dimensional constants $C$ depending on the marginals and on the ground cost, but not on the optimal transport plans themselves. Upper bounds are obtained for Lipschitz costs or locally semi-concave costs for a finer estimate, and lower bounds for $\mathscr{C}^2$ costs satisfying some signature condition on the mixed second derivatives that may include degenerate costs, thus generalizing results previously in the two marginals case and for non-degenerate costs. We obtain in particular matching bounds in some typical situations where the optimal plan is deterministic.
△ Less
Submitted 9 April, 2024; v1 submitted 6 July, 2023;
originally announced July 2023.
-
Convergence rate of general entropic optimal transport costs
Authors:
Guillaume Carlier,
Paul Pegon,
Luca Tamanini
Abstract:
We investigate the convergence rate of the optimal entropic cost $v_\varepsilon$ to the optimal transport cost as the noise parameter $\varepsilon \downarrow 0$. We show that for a large class of cost functions $c$ on $\mathbb{R}^d\times \mathbb{R}^d$ (for which optimal plans are not necessarily unique or induced by a transport map) and compactly supported and $L^{\infty}$ marginals, one has…
▽ More
We investigate the convergence rate of the optimal entropic cost $v_\varepsilon$ to the optimal transport cost as the noise parameter $\varepsilon \downarrow 0$. We show that for a large class of cost functions $c$ on $\mathbb{R}^d\times \mathbb{R}^d$ (for which optimal plans are not necessarily unique or induced by a transport map) and compactly supported and $L^{\infty}$ marginals, one has $v_\varepsilon-v_0= \frac{d}{2} \varepsilon \log(1/\varepsilon)+ O(\varepsilon)$. Upper bounds are obtained by a block approximation strategy and an integral variant of Alexandrov's theorem. Under an infinitesimal twist condition on $c$, i.e. invertibility of $\nabla_{xy}^2 c(x,y)$, we get the lower bound by establishing a quadratic detachment of the duality gap in $d$ dimensions thanks to Minty's trick.
△ Less
Submitted 7 June, 2022;
originally announced June 2022.
-
Mass concentration in rescaled first order integral functionals
Authors:
Antonin Monteil,
Paul Pegon
Abstract:
We consider first order local minimization problems of the form $\min \int_{\mathbb{R}^N}f(u,\nabla u)$ under a mass constraint $\int_{\mathbb{R}^N}u=m$. We prove that the minimal energy function $H(m)$ is always concave, and that relevant rescalings of the energy, depending on a small parameter $\varepsilon$, $Γ$-converge towards the $H$-mass, defined for atomic measures $\sum_i m_iδ_{x_i}$ as…
▽ More
We consider first order local minimization problems of the form $\min \int_{\mathbb{R}^N}f(u,\nabla u)$ under a mass constraint $\int_{\mathbb{R}^N}u=m$. We prove that the minimal energy function $H(m)$ is always concave, and that relevant rescalings of the energy, depending on a small parameter $\varepsilon$, $Γ$-converge towards the $H$-mass, defined for atomic measures $\sum_i m_iδ_{x_i}$ as $\sum_i H(m_i)$. We also consider Lagrangians depending on $\varepsilon$, as well as space-inhomogeneous Lagrangians and $H$-masses. Our result holds under mild assumptions on $f$, and covers in particular $α$-masses in any dimension $N\geq 2$ for exponents $α$ above a critical threshold, and all concave $H$-masses in dimension $N=1$. Our result yields in particular the concentration of Cahn-Hilliard fluids into droplets, and is related to the approximation of branched transport by elliptic energies.
△ Less
Submitted 22 February, 2024; v1 submitted 2 March, 2022;
originally announced March 2022.
-
Stability of optimal traffic plans in the irrigation problem
Authors:
Maria Colombo,
Antonio de Rosa,
Andrea Marchese,
Paul Pegon,
Antoine Prouff
Abstract:
We prove the stability of optimal traffic plans in branched transport. In particular, we show that any limit of optimal traffic plans is optimal as well. This is the Lagrangian counterpart of the recent Eulerian version proved in [CDM19a].
We prove the stability of optimal traffic plans in branched transport. In particular, we show that any limit of optimal traffic plans is optimal as well. This is the Lagrangian counterpart of the recent Eulerian version proved in [CDM19a].
△ Less
Submitted 26 March, 2020;
originally announced March 2020.
-
A fractal shape optimization problem in branched transport
Authors:
Paul Pegon,
Filippo Santambrogio,
Qinglan Xia
Abstract:
We investigate the following question: what is the set of unit volume which can be best irrigated starting from a single source at the origin, in the sense of branched transport? We may formulate this question as a shape optimization problem and prove existence of solutions, which can be considered as a sort of "unit ball" for branched transport. We establish some elementary properties of optimize…
▽ More
We investigate the following question: what is the set of unit volume which can be best irrigated starting from a single source at the origin, in the sense of branched transport? We may formulate this question as a shape optimization problem and prove existence of solutions, which can be considered as a sort of "unit ball" for branched transport. We establish some elementary properties of optimizers and describe these optimal sets A as sublevel sets of a so-called landscape function which is now classical in branched transport. We prove $β$-H{ö}lder regularity of the landscape function, allowing us to get an upper bound on the Minkowski dimension of the boundary: dim $\partial$A $\le$ d -- $β$ (where $β$ := d($α$ -- (1 -- 1/d)) $\in$ (0, 1) is a relevant exponent in branched transport, associated with the exponent $α$ > 1 -- 1/d appearing in the cost). We are not able to prove the upper bound, but we conjecture that $\partial$A is of non-integer dimension d -- $β$. Finally, we make an attempt to compute numerically an optimal shape, using an adaptation of the phase-field approximation of branched transport introduced some years ago by Oudet and the second author.
△ Less
Submitted 29 December, 2017; v1 submitted 5 September, 2017;
originally announced September 2017.
-
On the Lagrangian branched transport model and the equivalence with its Eulerian formulation
Authors:
Paul Pegon
Abstract:
First we present two classical models of Branched Transport: the Lagrangian model introduced by Bernot, Caselles, Morel, Maddalena, Solimini, and the Eulerian model introduced by Xia. An emphasis is put on the Lagrangian model, for which we give a complete proof of existence of minimizers in a --hopefully-- simplified manner. We also treat in detail some $σ$-finiteness and rectifiability issues to…
▽ More
First we present two classical models of Branched Transport: the Lagrangian model introduced by Bernot, Caselles, Morel, Maddalena, Solimini, and the Eulerian model introduced by Xia. An emphasis is put on the Lagrangian model, for which we give a complete proof of existence of minimizers in a --hopefully-- simplified manner. We also treat in detail some $σ$-finiteness and rectifiability issues to yield rigorously the energy formula connecting the irrigation cost I$α$ to the Gilbert Energy E$α$. Our main purpose is to use this energy formula and exploit a Smirnov decomposition of vector flows, which was proved via the Dacorogna-Moser approach by Santambrogio, to establish the equivalence between the Lagrangian and Eulerian models.
△ Less
Submitted 5 September, 2017;
originally announced September 2017.
-
A kinetic selection principle for curl-free vector fields of unit norm
Authors:
Pierre Bochard,
Paul Pegon
Abstract:
This article is devoted to the generalization of results obtained in 2002 by Jabin, Otto and Perthame. In their article they proved that planar vector fields taking value into the unit sphere of the euclidean norm and satisfying a given kinetic equation are locally Lipschitz. Here, we study the same question replacing the unit sphere of the euclidean norm by the unit sphere of \emph{any} norm. Und…
▽ More
This article is devoted to the generalization of results obtained in 2002 by Jabin, Otto and Perthame. In their article they proved that planar vector fields taking value into the unit sphere of the euclidean norm and satisfying a given kinetic equation are locally Lipschitz. Here, we study the same question replacing the unit sphere of the euclidean norm by the unit sphere of \emph{any} norm. Under natural asumptions on the norm, namely smoothness and a qualitative convexity property, that is to be of power type $p$, we prove that planar vector fields taking value into the unit sphere of such a norm and satisfying a certain kinetic equation are locally $\frac{1}{p-1}$-Hölder continuous. Furthermore we completely describe the behaviour of such a vector field around singular points as a \emph{vortex} associated to the norm. As our kinetic equation implies for the vector field to be curl-free, this can be seen as a selection principle for curl-free vector fields valued in spheres of general norms which rules out line-like singularities.
△ Less
Submitted 11 January, 2017;
originally announced January 2017.
-
Full characterization of optimal transport plans for concave costs
Authors:
Paul Pegon,
Davide Piazzoli,
Filippo Santambrogio
Abstract:
This paper slightly improves a classical result by Gangbo and McCann (1996) about the structure of optimal transport plans for costs that are concave functions of the Euclidean distance. Since the main difficulty for proving the existence of an optimal map comes from the possible singularity of the cost at $0$, everything is quite easy if the supports of the two measures are disjoint; Gangbo and M…
▽ More
This paper slightly improves a classical result by Gangbo and McCann (1996) about the structure of optimal transport plans for costs that are concave functions of the Euclidean distance. Since the main difficulty for proving the existence of an optimal map comes from the possible singularity of the cost at $0$, everything is quite easy if the supports of the two measures are disjoint; Gangbo and McCann proved the result under the assumption $μ(\spt(ν))=0$; in this paper we replace this assumption with the fact that the two measures are singular to each other. In this case it is possible to prove the existence of an optimal transport map, provided the starting measure $μ$ does not give mass to small sets (i.e. $(d\!-\!1)-$rectifiable sets). When the measures are not singular the optimal transport plan decomposes into two parts, one concentrated on the diagonal and the other being a transport map between mutually singular measures.
△ Less
Submitted 16 September, 2014; v1 submitted 14 November, 2013;
originally announced November 2013.
-
Optimal transportation with an oscillation-type cost: the one-dimensional case
Authors:
Didier Lesesvre,
Paul Pegon,
Filippo Santambrogio
Abstract:
The main result of this paper is the existence of an optimal transport map $T$ between two given measures $μ$ and $ν$, for a cost which considers the maximal oscillation of $T$ at scale $δ$, given by $ω_δ(T):=\sup_{|x-y|<δ}|T(x)-T(y)|$. The minimization of this criterion finds applications in the field of privacy-respectful data transmission. The existence proof unfortunately only works in dimensi…
▽ More
The main result of this paper is the existence of an optimal transport map $T$ between two given measures $μ$ and $ν$, for a cost which considers the maximal oscillation of $T$ at scale $δ$, given by $ω_δ(T):=\sup_{|x-y|<δ}|T(x)-T(y)|$. The minimization of this criterion finds applications in the field of privacy-respectful data transmission. The existence proof unfortunately only works in dimension one and is based on some monotonicity considerations.
△ Less
Submitted 2 October, 2012;
originally announced October 2012.
