Generalized Orlicz spaces and Wasserstein distances for convex-concave scale functions

Abstract: Given a strictly increasing, continuous function

v a r t h e t a : R_{+} o R_{+}

, based on the cost functional

i n t_{X i m e s X} v a r t h e t a (d (x, y)), d q (x, y)

, we define the

L^{v} a r t h e t a

-Wasserstein distance

W_{v} a r t h e t a (m u, u)

between probability measures

m u, u

on some metric space

(X, d)

. The function

v a r t h e t a

will be assumed to admit a representation

v a r t h e t a = p h i c i r c p s i

as a composition of a convex and a concave function

p h i

and

p s i

, resp. Besides convex functions and concave functions this includes all

m a t h c a l C^{2}

functions. For such functions

v a r t h e t a

we extend the concept of Orlicz spaces, defining the metric space

L^{v} a r t h e t a (X, m)

of measurable functions

f : X o R

such that, for instance,

d_vartheta(f,g)le1quadLongleftrightarrowquad int_Xvartheta(|f(x)-g(x)|),dmu(x)le1.

Recommendations

Cites Work

Cited In (6)

This page was built for publication: Generalized Orlicz spaces and Wasserstein distances for convex-concave scale functions