B1649
Title: Statistical learning from corrupted data via robust risk minimization
Authors: Dave Zachariah - Uppsala University (Sweden) [presenting]
Abstract: A general statistical estimation/learning problem is considered, where an unknown fraction of the training data is corrupted. A discussion on the classic data corruption model is presented and the learning problem is formulated in a risk minimization framework. A computationally robust learning method is described, which only requires specifying an upper bound on the corrupted data fraction. The wide range applicability of the method is demonstrated, including regression, classification, unsupervised learning and classic parameter estimation, with state-of-the-art performance.
