研究者の方へ
談話会
講演者:Jin Feng 氏 (Department of Mathematics, The University of Kansas)
時間: 2016年 7月 8日 (金) 16:30~17:30(16:00からtea)
場所: 談話会室
題目: A metric nature of the space of probability measures and an example of Hamilton-Jacobi equation on quotient spaces
概要:
A Hamilton-Jacobi equation in the space of probability measures will be
introduced as a continuum limit for deterministically interacting
particles evolving according to the Newtonian laws.
Such equation has been considered before. The well-posedness theory was
open unless each individual particle only live in one space dimension.
We give a well-posedness result for general dimensions by introducing
different new arguments. A key observation is that the space of
probability measures has hidden symmetries. We can profit greatly
by viewing such space as an infinite dimensional quotient space
with the quotient structure expressed using the Wasserstein metric.
The quotient structure observation lead us to consider some fine aspects
of the optimal transportation calculus that connect with
the probabilistic coupling ideas. By using a geometric tangent cone
concept to characterized the tangent and co-tangent space and
redefine the PDEs, we develops the well-posedness. The cone can be
identified with a subset of Markov transition kernels.
Its use is critical when a probability measure charge positive mass
on small sets (i.e. the phenomenon of condensation).
The talk is based on my joint work with Luigi Ambrosio.
問合せ先:津田塾大学数学科・情報科学科事務室(suujiアットマークtsuda.ac.jp)