Erwan Le Pennec

PAC-Bayesian aggregation and NL-Means

joint work with J. Salmon and L. Montuelle

How to combine estimators to obtain a better one? This is a goal of estimator aggregation. We study here the case of PAC-Bayesian aggregation in which the weights of each estimator depends exponentially on an estimate of its risk. The first serie of work studies the NL-Means estimate, an image estimate based on patch recombination, under this angle. The second considers general projection (and smoothed projection) estimators.

By default, only communications in english are shown. Using the language filter on the right side, you can also see the french ones.