FoCM 2014 conference
Workshop C3 - Learning Theory
December 20, 16:00 ~ 16:30 - Room B23
Learning Theory and Adaptive Partitioning in High Dimensions
Peter Binev
University of South Carolina, USA - binev@mailbox.sc.edu
Adaptive partitioning has been one of the methods of choice for several problems from nonlinear approximation theory. A typical challenge in applying this approach to learning theory is the increased complexity of the high-dimensional realization of the adaptive algorithms. We discuss paradigms like sparse occupancy and decorated trees that are designed to alleviate the difficulties related to high dimensions and tuned to certain learning theory setups.