A theorem on majorizing measures (Q858981): Difference between revisions

Let \((T,d)\) be a metric space, \(\varphi:\mathbb{R}_+\to \mathbb{R}\) a Young function. The main result of the paper states that is the probability measure \(m\) on \(T\), \(\text{supp}\, m=T\), is majorizing with respect to \(d\) and \(\varphi\), that is, \[ S:=\sum_{x\in T} \int_0^{D(T)} \varphi^{-1}\left( \frac{1}{m(B(x,\varepsilon))}\right)\, d\varepsilon <\infty, \] where \(D(T)=\text{diam}(T)\), then \[ E\, \sup_{s,t\in T} | X(s)-X(t)| \leq 32 S, \] for each separable stochastic process \(X(t)\), \(t\in T\), satisfying \[ E\varphi \left( \frac{| X(s)-X(t)| }{d(s,t)}\right)\leq 1\quad \text{for all \(s,t\in T\), \(s\neq t\)}. \] This result, although obtained by a different method, generalizes for all Young functions the Theorem 4.6 from \textit{M. Talagrand} [Ann. Probab. 18, No. 1, 1--49 (1990; Zbl 0703.60033)], which was proved under the doubling condition on the Young functions: \[ \varphi(2t)\leq C\varphi(t)\quad \text{for all \(t\geq 0\)}. \]