Probability Seminar
Department of Mathematical Sciences |
|
Shige PengShandong University, China
Title: Why Gaussian distributions are widely applied under distribution
uncertainties.
Why are Gaussian distributions actually widely used in our real
world in which probability and statistic model uncertainties are clearly
strong? In this talk we will present the author's result of a new central
limit theorem: under the upper (or robust, sublinear) expectation over the
uncertainty subset of probabilities, we still have the corresponding central
limit theorem (CLM). This result tells us that in many important situations
we can still use the classical Gaussian distribution to calculate E[u(X)].
But the choice of the variance sensitively depends on whether the
above function u is convex or concave. If u is neither convex nor concave
then the situation becomes more subtle and a new calculation method is
introduced for this G-normal distribution. We also present the dynamic
(thus infinite-dimensional) counterpart of the problem which gives us a new
type of Brownian motion, called G-Brownian motion, and its corresponding
Ito calculus.
©2008, Department of Mathematical Sciences
Last Modified:
November 18, 2008