Foundations of multi-agent learning