A Value Theory of Meta-Learning Algorithms

Book One

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 21

AAAI / SIGART Doctoral Consortium

We use game theory to analyze meta-learning algorithms. The objective of meta-learning is to determine which algorithm to apply on a given task. This is an instance of a more general problem that consists of allocating knowledge consumers to learning producers. Solving this general problem in the field of meta-learning yields solutions for related fields such as information retrieval and recommender systems.