Kernel-Based Learning of Conditional Utility Functions



: Pahikkala T

: 2013

: European Conference on Operational Research (EURO 2013)

: http://www.euro-online.org/conf/euro26/treat_abstract?paperid=16575


In this work, we consider the problem of learning cardinal utility functions with a condition. The condition being, for example, a query object given at prediction time, the learned utility function assigns to it the cardinal utility values of a set of target objects, also given at prediction time. On one hand, we analyze the universal approximation properties of kernel-based regression algorithms on this task, and on the other hand, we study how the generalization properties of the algorithms can be improved via incorporation of certain invariances inherent to the considered task.



Last updated on 2024-26-11 at 12:16