The Kantorovich problem [1] is one of the basic minimization problems in optimal transport. It is named after Russian mathematician and Nobel Laureate Leonid Kantorovich.
Kantorovich Problem. The Kantorovich Problem (like the Monge Problem) can be visualized as distributing piles of sand into holes (of equal volume) in an optimal fashion.
[2]
Introduction
There are two basic problems in optimal transport the Monge problem and the Kantorovich problem. In contrast to the Monge problem, The Kantorovich problem allows a non-empty minimization set, a convex constraint set, and a convex effort functional. The Kantrovich problem admits a dual because it is a linear minimization problem with convex constraints.
Shipping problem
Suppose there is a merchant who is attempting to ship their items from one place to another. They can hire trucks at some cost
for each unit of merchandise which is shipped from point
to point
. Now the shipper is approached by a mathematician, who claims that prices can be set such that they align with the shipper's financial interests. This would be achieved by setting the price
and
such that the sum of
and
is always less than the cost
. This may even involve setting negative prices in certain cases. However, it can be shown that the shipper will spend almost as much as they would have if they had opted for the original pricing method.
Kantorovich Optimal Transport Problem
Consider the basic premises of optimal mass transportation. Consider probability spaces
and
. Let
be a nonnegative measurable function on
. The Kantorovich problem is the following:
This is on the convex set
which is also nonempty. Note
if and only if
is a nonnegative measure which satisfies:
for all measurable subsets
of
and
of
. This definition implies that
is a probability measure. Another way to say this is that
if and only if it is a nonnegative measure on
such that, for all measurable functions
or equivalently
(2)
There are some topological assumptions that can be made on the measure spaces
and
. When
and
are Polish spaces (i.e. completely metrizable and separable spaces), and
are Borel probability measures, it is sufficient to impose the expression above for
only [3].
In addition if
and
are locally compact. then one can even be content with imposing (2) for
.[4] Note that
is the space of bounded continuous functions on
and
the space of continuous functions going to 0 at infinity, i.e. those continuous functions
such that for any
there is a compact set
satisfying
note that
This possibility to restrict the class of test functions to the narrower space
when
and
are locally compact is due to Riesz' theorem, which identifies the space
of Borel measures having finite total variation on
with the topological dual of
.[5]
Kantorovich Duality
Since the Kantorovich problem is a linear minimization problem with convex constraints it admits a dual. Kantorovich expressed this in 1942, where he considered the case in which the cost function may be conceived of as a distance: function is a distance:
.
Theorem
Let
and
be Polish spaces, let
and
and let
be a lower
semi-continuous cost function.
Whenever
and
define
Define
to be the set of all Borel probability measures
on
such that for all measurable subsets
and
.
and define
to be the set of all measurable functions
satisfying
for d
-almost all
-almost all
(
outside of
negligible set.
Then
If the definition of
to those functions
that are continuous and bounded, then the supremum on the right hand side of (3) is not affected.