Supervised learning: no loss no cry

Richard Nock, Aditya Krishna Menon

Abstract: Supervised learning requires the specification of a loss function to minimise. While the theory of admissible losses from both a computational and statistical perspective is well-developed, these o...