Optimal Transport in One Dimension

In this article, we explore the optimal transport problem on the real line along with some examples.

Linear Cost Example

For this example, consider the cost function $c(x,y)=L(x-y)$ along with a given linear map $A:\mathbb {R} ^{d}\rightarrow \mathbb {R}$. Moreover, if let $\gamma$ be any transport plan, then by direct computation we see that

                                                                                $\int L(x-y)d\gamma =\int L(x)d\gamma -\int L(y)d\gamma =\int L(x)d\mu -\int L(y)d\nu$

which suggests that this result only depends on the marginals of $\gamma$ (wherein $\mu$ and $\nu$ are compactly supported probability measures). In fact, in such cases, every transport plan/map is optimal.

Distance Cost Example

Consider the cost function $c(x,y)=|x-y|$ along with probability measures (on $\mathbb {R}$) $\mu$ and $\nu$. Then, for any $(x,y)\in spt(\mu )\times spt(\nu )$ we see that $c(x,y)=y-x$, which then immediately puts us back in the linear cost position, so any transport map/plan is also optimal for such costs.

Book Shifting Example

Consider the cost function $c(x,y)=|x-y|$ along with $\mu ={\frac {1}{2}}\lambda _{[0,2]}$ and $\nu ={\frac {1}{2}}\lambda _{[1,3]}$ (where $\lambda$ is the one-dimensional Lebesgue measure). A (monotone) transport plan that rearranges $\mu$ to look like $\nu$ is given by $T_{0}(x)=x+1$ and its corresponding cost is

                                                                                               $M(T_{0})=\int |T_{0}(x)-x|d\mu \equiv 1$.

Furthermore, notice that the piecewise map $T_{1}$ given by $T_{1}(x)=x+2$ (for $x\leq 1$) and $T_{1}(x)=x$ (for $x>1$) satisfies $T_{1}\#\mu =\nu$, i.e. $T_{1}$ is a transport map from $\mu$ to $\nu$; moreover, the corresponding cost is

                                                                                          $M(T_{1})=\int |T_{1}(x)-x|d\mu ={\frac {1}{2}}\int _{0}^{2}dx\equiv 1$