Title:Tail bounds for Multivariate Random Tensor Means

Abstract:In our recent research endeavors, we have delved into the realm of tail bounds problems concerning bivariate random tensor means. In this context, tensors are treated as finite-dimensional operators. However, the longstanding challenge of extending the concept of operator means to scenarios involving more than two variables had persisted. The primary objective of this present study is to unveil a collection of tail bounds applicable to multivariate random tensor means. These encompass the weighted arithmetic mean, weighted harmonic mean, and the Karcher mean. These bounds are derived through the utilization of Ando-Hiai's inequalities, alongside tail bounds specifically tailored for multivariate random tensor means employing reverse Ando-Hiai's inequalities, which are rooted in Kantorovich constants. Notably, our methodology involves employing the concept of deformation for operator means with multiple variables, following the principles articulated in Hiai, Seo and Wada's recent work. Additionally, our research contributes to the expansion about the Karcher mean differentiable region from the vicinity of the diagonal identity element within the Cartesian product space of positive definite tensors to the vicinity of the general element within the Cartesian product space of positive definite tensors via the application of the inverse and implicit function theorem.