The Hoffmann-Jørgensen inequality in metric semigroups (Q682261): Difference between revisions

The authors present a generalization of the Hoffmann-Jørgensen inequality for independent random variables in the setting of metric semigroups. Let \((\mathcal{G},d_\mathcal{G})\) be a separable metric semigroup with \(z_0,z_1\in\mathcal{G}\) fixed, let \(X_1,\ldots,X_n\in L^0(\Omega,\mathcal{G})\) be independent random variables, and let the integers \(k,n_1,\ldots,n_k\in\mathbb{N}\) and scalars \(s,t_1,\ldots,t_k\in[0,\infty)\) be fixed. If we define \(S_j=X_1\cdots X_j\) and \(U_n=\max_{1\leq j\leq n}d_\mathcal{G}(z_1,z_0S_j)\), then the main result of the present paper is an upper bound on \[ \mathbb{P}\left(U_n>(2n_1-1)t_1+2\sum_{i=2}^kn_it_i+\left(\sum_{i=1}^kn_i-1\right)s\right)\,, \] for \(\sum_{i=1}^kn_i\leq n+1\). This upper bound unifies several different versions of the Hoffmann-Jørgensen inequality available in the literature, provides improved bounds over other versions available, and demonstrates that such an inequality holds not only in the Banach space setting, but also more generally in metric semigroups (including, for example, normed linear spaces and certain Lie groups).