Let
denote the space of all Radon measures on
with finite mass. Moreover, let
denote the space of Radon measures with finite
moment, that is,
.
Then, for
with
the
-Wasserstein distance is defined as[1]

where
denotes the set of all transport plans from
to
. Note that, if we restrict
and
to be probability measures, then the
-Wasserstein distance can be seen as a special case of the Kantorovich Problem, where
. Furthermore, if we let
denote the space of probability measures on
with finite
moment, then
defines a metric on
.
Measures with unequal mass and signed measures
The classical Wasserstein distance can be generalized for measures with unequal mass by allowing the addition and removal of mass to
and
.[1] In this case, there is unit cost
for the addition or removal of mass from both
and
, whereas the transport cost of mass between
and
stays the same with the classical Kantorovich Problem; multiplied with some rate
.
Definition.[1] For some
and
, the generalized Wasserstein distance
is given by

where
is the classical
-Wasserstein distance for measures with equal mass.
Note that, under this definition
defines a metric on
, and
is a complete metric space.[1]. Furthermore, this generalization allows one to extend the
-Wasserstein distance to the signed Radon measures as well. Let us denote the space of all signed Radon measures on
with finite mass with
.
Definition.[2] For some
, the generalized Wasserstein distance for signed measures
is given by

where
are any measures in
such that
and
.
Moreover, if we let
, then
is a normed vector space. However, as opposed to the completeness of
,
fails to be a Banach space.[2]
Duality
As a special case of the Kantorovich Dual Problem when
,
-Wasserstein metric has the following dual characterization.
Theorem.[3] Let Lip
denote the space of all Lipschitz functions on
, and let
. Then,
.
In a similar spirit, this duality result can be extended for generalized
-Wasserstein distances
and
as well, where
and
are taken as 1. For measures with unequal mass, when the additional constraint
is imposed on the test functions, it holds that[4]
.
In terms of the generalized 1-Wasserstein metric for signed measures, when we relax the compact support condition on the test functions we obtain the equivalent duality characterization as follows.[2]
.
References
- ↑ 1.0 1.1 1.2 1.3 Piccoli, B., Rossi, F. Generalized Wasserstein Distance and its Application to Transport Equations with Source. Arch Rational Mech Anal 211, 335–358 (2014). https://doi.org/10.1007/s00205-013-0669-x
- ↑ 2.0 2.1 2.2 Piccoli, B., Rossi, F., Tournus M. A Wasserstein norm for signed measures, with application to nonlocal transport equation with source term. arXiv:1910.05105 (2019).
- ↑ C. Villani, Topics in Optimal Transportation, Chapter 1.
- ↑ Piccoli, B., Rossi, F. Piccoli, B., Rossi, F. On Properties of the Generalized Wasserstein Distance. Arch Rational Mech Anal 222, 1339–1365 (2016). https://doi.org/10.1007/s00205-016-1026-7