18
71

Fast Rates for General Unbounded Loss Functions: from ERM to Generalized Bayes

Abstract

We present new excess risk bounds for general unbounded loss functions including log loss and squared loss, where the distribution of the losses may be heavy-tailed. The bounds hold for general estimators, but they are optimized when applied to η\eta-generalized Bayesian, MDL, and empirical risk minimization estimators. In the case of log loss, the bounds imply convergence rates for generalized Bayesian inference under misspecification in terms of a generalization of the Hellinger metric as long as the learning rate η\eta is set correctly. For general loss functions, our bounds rely on two separate conditions: the vv-GRIP (generalized reversed information projection) conditions, which control the lower tail of the excess loss; and the newly introduced witness condition, which controls the upper tail. The parameter vv in the vv-GRIP conditions determines the achievable rate and is akin to the exponent in the Tsybakov margin condition and the Bernstein condition for bounded losses, which the vv-GRIP conditions generalize; favorable vv in combination with small model complexity leads to O~(1/n)\tilde{O}(1/n) rates. The witness condition allows us to connect the excess risk to an "annealed" version thereof, by which we generalize several previous results connecting Hellinger and R\ényi divergence to KL divergence.

View on arXiv
Comments on this paper