Section: New Results
Finite-sample Analysis of M-estimators using Self-concordance
In , we demonstrate how self-concordance of the loss allows to obtain asymptotically optimal rates for -estimators in finite-sample regimes. We consider two classes of losses: (i) self-concordant losses, i.e., whose third derivative is uniformly bounded with the power of the second; (ii) pseudo self-concordant losses, for which the power is removed. These classes contain some losses arising in generalized linear models, including the logistic loss; in addition, the second class includes some common pseudo-Huber losses. Our results consist in establishing the critical sample size sufficient to reach the asymptotically optimal excess risk in both cases. Denoting the parameter dimension, and the effective dimension taking into account possible model misspecification, we find the critical sample size to be for the first class of losses, and for the second class, where is the problem-dependent parameter that characterizes the risk curvature at the best predictor . In contrast to the existing results, we only impose local assumptions on the data distribution, assuming that the calibrated design, i.e., the design scaled with the square root of the second derivative of the loss, is subgaussian at the best predictor. Moreover, we obtain the improved bounds on the critical sample size, scaling near-linearly in , under the extra assumption that the calibrated design is subgaussian in the Dikin ellipsoid of . Motivated by these findings, we construct canonically self-concordant analogues of the Huber and logistic losses with improved statistical properties. Finally, we extend some of the above results to -penalized -estimators in high-dimensional setups.