Gaussian approximation: From Malliavin-Stein approach to information theory

Thời gian: 09:00 đến 11:00 Ngày 04/06/2021

Địa điểm: Online qua Zoom

Báo cáo viên: TS. Diu Tran, University of Luxembourg, Luxembourg.

Tóm tắt:
We are interested in estimating the discrepancy between the distributions of F and the Gaussian vector Z by working with Lp-norms, total variation distance, relative entropy and Fisher information, when F is a d-dimensional centered random vector whose components are multiple stochastic integrals.
Under the so-called ''non-degenerate'' condition of the Malliavin derivative, the convergence in the sense of the Fisher information of F to Z is actually equivalent to convergence of only the fourth moments of each component.
Furthermore, by using the multivariate de Bruijn’s identity and the Fourth Moment Theorem, we obtain an upper bound on Fisher information, from which we deduce a list of equivalences between different forms of convergence to the Gaussian random vectors.

Meeting ID: 862 363 8574
Passcode: 123456