Optimal Transport Map Research Articles

The existence of geodesics in Wasserstein spaces over path groups and loop groups

In this work we prove the existence and uniqueness of the optimal transport map for Lp-Wasserstein distance with p>1, and particularly present an explicit expression of the optimal transport map for the case p=2. As an application, we show the existence of geodesics connecting probability measures satisfying suitable condition on path groups and loop groups.

Characterization of barycenters in the Wasserstein space by averaging optimal transport maps

This paper is concerned by the study of barycenters for random probability measures in the Wasserstein space. Using a duality argument, we give a precise characterization of the population barycenter for various parametric classes of random probability measures with compact support. In particular, we make a connection between averaging in the Wasserstein space as introduced in Agueh and Carlier [SIAM J. Math. Anal.43(2011) 904–924], and taking the expectation of optimal transport maps with respect to a fixed reference measure. We also discuss the usefulness of this approach in statistics for the analysis of deformable models in signal and image processing. In this setting, the problem of estimating a population barycenter fromnindependent and identically distributed random probability measures is also considered.

Displacement Interpolation Using Monotone Rearrangement

When approximating a function that depends on a parameter, one encounters many practical examples where linear interpolation or linear approximation with respect to the parameters prove ineffective. This is particularly true for responses from hyperbolic partial differential equations (PDEs) where linear, low-dimensional bases are difficult to construct. We propose the use of displacement interpolation where the interpolation is done on the optimal transport map between the functions at nearby parameters, to achieve an effective dimensionality reduction of hyperbolic phenomena. We further propose a multi-dimensional extension by using the intertwining property of the Radon transform. This extension is a generalization of the classical translational representation of Lax-Philips [Lax and Philips, Bull. Amer. Math. Soc. 70 (1964), pp.130--142].

Transport Map Accelerated Markov Chain Monte Carlo

We introduce a new framework for efficient sampling from complex probability distributions, using a combination of optimal transport maps and the Metropolis-Hastings rule. The core idea is to use continuous transportation to transform typical Metropolis proposal mechanisms (e.g., random walks, Langevin methods) into non-Gaussian proposal distributions that can more effectively explore the target density. Our approach adaptively constructs a lower triangular transport map-an approximation of the Knothe-Rosenblatt rearrangement-using information from previous MCMC states, via the solution of an optimization problem. This optimization problem is convex regardless of the form of the target distribution. It is solved efficiently using a Newton method that requires no gradient information from the target probability distribution; the target distribution is instead represented via samples. Sequential updates enable efficient and parallelizable adaptation of the map even for large numbers of samples. We show that this approach uses inexact or truncated maps to produce an adaptive MCMC algorithm that is ergodic for the exact target distribution. Numerical demonstrations on a range of parameter inference problems show order-of-magnitude speedups over standard MCMC techniques, measured by the number of effectively independent samples produced per target density evaluation and per unit of wallclock time.

A Texture Synthesis Model Based on Semi-Discrete Optimal Transport in Patch Space

Exemplar-based texture synthesis consists in producing new synthetic images which have the same perceptual characteristics as a given texture sample while exhibiting sufficient innovation (to avoid verbatim copy). In this paper, we propose to address this problem with a model obtained as local transformations of Gaussian random fields. The local transformations operate on $3 \times 3$ patches and are designed to solve a semi-discrete optimal transport problem in order to reimpose the patch distribution of the exemplar texture. The semi-discrete optimal transport problem is solved with a stochastic gradient algorithm, whose convergence speed is evaluated on several practical transport cases. After studying the properties of such transformed Gaussian random fields, we propose a multiscale extension of the model which aims at preserving the patch distribution of the exemplar texture at multiple scales. Experiments demonstrate that this multiscale model is able to synthesize structured textures while keeping several mathematical guarantees, and with low requirements in synthesis time and memory storage. In particular, a single patch optimal transport map is shown to be better than iterated nearest neighbor assignments in terms of statistical guarantees. Besides, once the model is estimated, the resulting synthesis algorithm is fast and highly parallel since it amounts to performing weighted nearest neighbor patch assignments at each scale.

Design of a smooth freeform illumination system for a point light source based on polar-type optimal transport mapping.

A design method is proposed to generate smooth freeform illumination optics for a point light source based on the L2 optimal transport (LOT) mapping. In this method, the LOT mapping between an assumed circular planar source and a prescribed target is first obtained by solving a polar-type LOT problem. Then, the mapping calculated for the circular source is applied for a point light source. Finally, the freeform optical surface is generated by a geometric construction method to realize the ray mapping. As examples, a series of smooth-surface freeform lenses are designed for a point light source to form uniform and complex illumination patterns on rectangular targets. The ray-tracing results show that all the designs achieve excellent performance with the light utilization efficiency η over 0.87 (Fresnel loss considered) and the relative standard deviation (RSD) of the simulated illumination distribution less than 0.051 simultaneously.

An Algorithm for Optimal Transport between a Simplex Soup and a Point Cloud

We propose a numerical method to find the optimal transport map between a measure supported on a lower-dimensional subset of $\mathbb{R}^d$ and a finitely supported measure. More precisely, the sou...

Optimal transportation for generalized Lagrangian

This paper deals with the optimal transportation for generalized Lagrangian L = L(x, u, t), and considers the following cost function: $$c\left( {x,y} \right) = \mathop {\inf }\limits_{\begin{array}{*{20}{c}} {x\left( 0 \right) = x} {x\left( 1 \right) = y} {u \in U} \end{array}} \int_0^1 {L\left( {x\left( s \right),u\left( {x\left( s \right),s} \right),s} \right)ds} ,$$ where U is a control set, and x satisfies the ordinary equation $$\dot x\left( s \right) = f\left( {x\left( s \right),u\left( {x\left( s \right),s} \right)} \right).$$ It is proved that under the condition that the initial measure μ 0 is absolutely continuous w.r.t. the Lebesgue measure, the Monge problem has a solution, and the optimal transport map just walks along the characteristic curves of the corresponding Hamilton-Jacobi equation: $$\left\{ {_{V\left( {0,x} \right) = {\phi _0}\left( x \right).}^{{V_t}\left( {t,x} \right) + \mathop {\sup }\limits_{u \in U} \left\langle {{V_x}\left( {t,x} \right),f\left( {x,u\left( {x\left( t \right),t} \right),t} \right) - L\left( {x\left( t \right),u\left( {x\left( t \right),t} \right),t} \right)} \right\rangle = 0,}} \right.$$

On coupling particle filter trajectories

Particle filters are a powerful and flexible tool for performing inference on state-space models. They involve a collection of samples evolving over time through a combination of sampling and re-sampling steps. The re-sampling step is necessary to ensure that weight degeneracy is avoided. In several situations of statistical interest, it is important to be able to compare the estimates produced by two different particle filters; consequently, being able to efficiently couple two particle filter trajectories is often of paramount importance. In this text, we propose several ways to do so. In particular, we leverage ideas from the optimal transportation literature. In general, though computing the optimal transport map is extremely computationally expensive, to deal with this, we introduce computationally tractable approximations to optimal transport couplings. We demonstrate that our resulting algorithms for coupling two particle filter trajectories often perform orders of magnitude more efficiently than more standard approaches.

Partial W2,p regularity for optimal transport maps

We prove that, in the optimal transportation problem with general costs and positive continuous densities, the potential function is always of class Wloc2,p for any p≥1 outside of a closed singular set of measure zero. We also establish global W2,p estimates when the cost is a small perturbation of the quadratic cost. The latter result is new even when the cost is exactly the quadratic cost.

Robust surface registration using optimal mass transport and Teichmüller mapping

Surface registration plays a significant role in computer vision and engineering fields. One of the most challenging problems in surface registration, however, is to obtain a unique bijective registration for surfaces with large deformations and landmarks constraints. In this work, a novel surface registration framework is proposed to tackle this problem using optimal mass transport mapping (OMT-Map) and Teichmüller mapping (T-Map). All metric surfaces with the disk topology are mapped to the planar disk using OMT-Map, which avoids huge area distortion, thus rendering our method more robust. A landmark-constrained T-Map is then computed between two planar disks such that the maximal conformality distortion is minimized while the landmarks are matched. Compared with existing surface registration methods, our method is more advantageous in enforcing the robustness by avoiding large area distortion, and producing diffeomorphisms with all landmarks matched consistently. Numerical experiments on various surfaces demonstrate the efficiency, robustness of our method.

Surface parameterization based on polar factorization

Abstract A mapping between measurable subsets of Euclidean space can be uniquely factorized to the composition of a measure-preserving mapping and an optimal transportation map, the later is also the gradient map of a convex function. This work introduces an algorithm based on variational approach to compute this type of polar factorization for mappings between planar domains and some direct applications. Our method greatly increases the flexibility for surface parameterizations by balancing between area distortion and angle distortion, and improves the accuracy and numerical stability for down steam geometric processing tasks.

Pollution Transfer as Optimal Mass Transport Problem

In this paper, we use mass transportation theory to study pollution transfer in porous media. We show the existence of a $L^2-$regular vector field defined by a $W^{1, 1}-$ optimal transport map. A sufficient condition for solvability of our model, is given by a (non homogeneous) transport equation with a source defined by a measure. The mathematical framework used, allows us to show in some specifical cases, existence of solution for a nonlinear PDE deriving from the modelling. And we end by numerical simulations.

Ray mapping approach for the efficient design of continuous freeform surfaces.

The efficient design of continuous freeform surfaces, which maps a given light source to an arbitrary target illumination pattern, remains a challenging problem and is considered here for collimated input beams. A common approach are ray-mapping methods, where first a ray mapping between the source and the irradiance distribution on the target plane is calculated and in a subsequent step the surface is constructed. The challenging aspect of this approach is to find an integrable mapping ensuring a continuous surface. Based on the law of reflection/refraction and an integrability condition, we derive a general condition for the surface and ray mapping for a collimated input beam. It is shown that in a small-angle approximation a proper mapping can be calculated via optimal mass transport - a mathematical framework for the calculation of a mapping between two positive density functions. We show that the surface can be constructed by solving a linear advection Eq. with appropriate boundary conditions. The results imply that the optimal mass transport mapping is approximately integrable over a wide range of distances between the freeform and the target plane and offer an efficient way to construct the surface by solving standard integrals. The efficiency is demonstrated by applying it to two challenging design examples, which shows the ability of the presented approach to handle target illumination patterns with steep irradiance gradients and numerous gray levels.

Volume preserving mesh parameterization based on optimal mass transportation

In order to convert a finite element mesh model to the spline representation for the purpose of isogeometric analysis, one needs to parameterize the solid. This work introduces a novel volumetric parameterization method, which guarantees to be free of volume distortion.Given a simply connected tetrahedral mesh with a single boundary surface, we first compute a harmonic map from the boundary triangle mesh to the unit sphere by non-linear heat diffusion method; then we use the surface harmonic map as the boundary condition to compute the volumetric harmonic map to parameterize the solid onto the unit solid ball; finally we compute an optimal mass transportation map from the unit solid ball with the push-forward volume element induced by the harmonic map onto itself with the Euclidean volume element. The composition of the volumetric harmonic map and the optimal mass transportation map gives an volume-preserving parameterization.The method has solid theoretic foundation, and is based on conventional algorithms in computational geometry, easy to implement. We have thoroughly tested our algorithm on many solid models in reality. The experimental results demonstrate the efficiency and efficacy of the proposed method. To the best of our knowledge, it is the first work addressing volume-preserving parameterization in the literature.

Measure controllable volumetric mesh parameterization

Volumetric parameterization is a fundamental problem in solid and physical modeling. In practice, it is highly desirable to control the volumes of the regions of interest in the parameter domain. This work introduces a novel volumetric parameterization method, which allows users to prescribe the target volumetric measure of the input solid.Given a simply connected tetrahedral mesh with a single boundary surface, we first compute a volumetric harmonic map to parameterize the solid onto the unit solid ball; then we compute an optimal mass transportation map from the unit solid ball with the push-forward volume element induced by the harmonic map onto the parameter domain with the user prescribed volumetric measure. The composition of the volumetric harmonic map and the optimal mass transportation map gives a measure controllable volumetric parameterization. Furthermore, this method can handle solids with empty voids inside.The method has solid theoretic foundation, and is based on conventional algorithms in computational geometry, and easy to implement. The experimental results demonstrate the efficiency and efficacy of the proposed method.

Freeform illumination optics construction following an optimal transport map.

We present a modified optimal transport (OT) ray-mapping approach for designing freeform illumination optics. After mapping the source intensity into a virtual irradiance distribution under stereographic projection, we employ an advanced OT map computation method with the ability to tackle nonstandard boundary conditions. Following the computed map, we construct the freeform optical surface directly from normal vectors by requiring that the chord between two adjacent points is perpendicular to the average of the two normal vectors at these two points and enforcing this relationship with a least squares method. Examples of designing freeform lenses for LED sources show that we can produce various uniform illumination patterns with high optical efficiencies.

Stability results on the smoothness of optimal transport maps with general costs

We prove some stability results concerning the smoothness of optimal transport maps with general cost functions. In particular, we show that the smoothness of optimal transport maps is an open condition with respect to the cost function and the densities. As a consequence, we obtain regularity for a large class of transport problems, where the cost does not necessarily satisfy the MTW condition.

A Multiscale Strategy for Bayesian Inference Using Transport Maps

In many inverse problems, model parameters cannot be precisely determined from observational data. Bayesian inference provides a mechanism for capturing the resulting parameter uncertainty, but typically at a high computational cost. This work introduces a multiscale decomposition that exploits conditional independence across scales, when present in certain classes of inverse problems, to decouple Bayesian inference into two stages: (1) a computationally tractable coarse-scale inference problem; and (2) a mapping of the low-dimensional coarse-scale posterior distribution into the original high-dimensional parameter space. This decomposition relies on a characterization of the non-Gaussian joint distribution of coarse- and fine-scale quantities via optimal transport maps. We demonstrate our approach on a sequence of inverse problems arising in subsurface flow, using the multiscale finite element method to discretize the steady state pressure equation. We compare the multiscale strategy with full-dimensional Markov chain Monte Carlo on a problem of moderate dimension (100 parameters) and then use it to infer a conductivity field described by over 10,000 parameters.

A Numerical Algorithm forL2Semi-Discrete Optimal Transport in 3D

This paper introduces a numerical algorithm to compute the L2 optimal transport map between two measures μ and ν, where μ derives from a density ρ defined as a piecewise linear function (supported by a tetrahedral mesh), and where ν is a sum of Dirac masses. I first give an elementary presentation of some known results on optimal transport and then observe a relation with another problem (optimal sampling). This relation gives simple arguments to study the objective functions that characterize both problems. I then propose a practical algorithm to compute the optimal transport map between a piecewise linear density and a sum of Dirac masses in 3D. In this semi-discrete setting [Aurenhammer et al., Proc. of 8th Symposium on Computational Geometry (1992) 350–357] showed that the optimal transport map is determined by the weights of a power diagram. The optimal weights are computed by minimizing a convex objective function with a quasi-Newton method. To evaluate the value and gradient of this objective function, I propose an efficient and robust algorithm, that computes at each iteration the intersection between a power diagram and the tetrahedral mesh that defines the measure μ. The numerical algorithm is experimented and evaluated on several datasets, with up to hundred thousands tetrahedra and one million Dirac masses.