Publication
This paper considers the problem of adaptive filtering in the presence of uncertainties in the regression data. A recursive procedure is derived that is based on solving local optimization problems that attemps to alleviate the worst-case effect of data uncertainties on filter performance. The resulting procedure turns out to have similarities with leakage-based adaptive filters.
Florent Gérard Krzakala, Lenka Zdeborová, Hugo Chao Cui