Thursday,
September 16, 12:00 p.m. PSA 311
Guohei Song
School of Mathematical and Statistical Sciences
Regularization in Reproducing Kernel Banach Spaces
Abstract
Reproducing kernel Hilbert spaces (RKHS) are Hilbert spaces on
which point evaluation functionals are continuous. It is known that the
regularization problem (also known as penalized least squares) in a
(possibly infinite-dimensional) RKHS admits a minimizer in a
finite-dimensional subspace. We try to generate this property to Banach
spaces. The setup of Reproducing kernel Banach Spaces (RKBS) will be
presented. A necessary and sufficient condition for the validity of this
property in RKBS will also be investigated.
