Quantcast
Viewing all articles
Browse latest Browse all 9295

Clarification about continuity of cost function

I've come across the statement that the following function, referred to as Coulomb cost in optimal transport, is lower semicontinuous: $c : \mathbb{R}^{Nd} \rightarrow \mathbb{R}, c(x_1,.., x_N) = \sum_{1 \leq i < j \leq N}\frac{1}{|x_i - x_j|}$ (for example, this paper: https://flore.unifi.it/bitstream/2158/1070973/1/m2an150051.pdf, Remark 3.14).

It is assumed to be $\infty$ when $x_j = x_j$ for two indices $i \neq j.$ The function is continuous when $x_i \neq x_j, \forall 1 \leq i < j \leq N,$ so I'm wondering about the pathological case where it is $\infty$. If we take $N=2$ for simplicity, for every sequence $(x_n, y_n) \rightarrow (x,x), \liminf_{n \rightarrow \infty} c(x_n, y_n) = \infty = c(x,x),$ because $c(x_n, y_n)$ diverges. My confusion: Why is this specifically stated as lower semicontinuous and not simply continuous? We would also get $\limsup_{n \rightarrow \infty} c(x_n,y_n) \leq \infty = c(x,x),$ (actually "$\limsup_{n \rightarrow \infty} c(x_n,y_n) = \infty = c(x,x)$"), so why not simply continuous?


Viewing all articles
Browse latest Browse all 9295

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>