Sample Complexity of Representation Learning for Sparse and Related Data Models
Matthias Seibert · mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) · 2019
This thesis investigates the sample complexity of representation learning algorithms for sparse and related signal models. Two frameworks are developed for establishing bounds on the generalization error and sample complexity for a general class of learned representation models based on dictionaries. The proposed bounding schemes are presented in a self-contained manner and can be used to analyze a wide variety of learning scenarios. Both methods are then applied to a selection of learning algorithms.