I read the article "Komlós Theorem for Unbounded Random Sets" by G. KRUPA (MSN), but I did not understand the difference between:
Measurable multifunction integrably bounded,
Measurable multifunction integrable
(see p. 239).
I read the article "Komlós Theorem for Unbounded Random Sets" by G. KRUPA (MSN), but I did not understand the difference between:
Measurable multifunction integrably bounded,
Measurable multifunction integrable
(see p. 239).
Let $F:\Omega\to 2^X$ be a measurable multifunction with closed values in the separable Banach space $X$. It seems to be implicitly assumed that $F$ jas nonempty values.
$F$ is integrably bounded if the function $\omega\mapsto \sup_{x\in F(\omega)}\|x\|$ is integrable.
$F$ is integrable if the function $\omega\mapsto \inf_{x\in F(\omega)}\|x\|=d(0,F(\omega))$ is integrable.
The point of these definitions might be better understood in terms of measurable selections. By the Kuratowski-Ryll-Nardzewski measurable selection theorem, $F$ admits measurable selections. If $F$ is integrably bounded, all selections will be integrable and dominated by a single integrable function. If $F$ is integrable, at least one selection will be integrable.