Oleksandr Zadorozhnyi (Potsdam)
In the talk I present a Bernstein type inequality for Banach-valued random sums under weak-dependency assumption of general kind on the variables and smoothness assumption on the underlying Banach norm. In the succeeding part this concentration inequality will be used in asymptotical regime to derive risk upper bounds for the family of spectral regularization methods for reproducing kernel decision rules, when trained on a sample coming from a tau-mixing process.
This talk is based on a joint work with Gilles Blanchard.
