Part of Advances in Neural Information Processing Systems 14 (NIPS 2001)
Dörthe Malzahn, Manfred Opper
We combine the replica approach from statistical physics with a varia- tional approach to analyze learning curves analytically. We apply the method to Gaussian process regression. As a main result we derive ap- proximative relations between empirical error measures, the generaliza- tion error and the posterior variance.