Model Selection as a Multiple Testing Procedure: Improving Akaike's Information Criterion
Authors:
Preprint
Abstract
By interpreting the model selection problem as a multiple hypothesis testing task, we propose a modification of Akaike's Information Criterion that avoids overfitting, even when the sample size is small. We call this correction an over-penalization procedure. As a proof of concept, we show nonasymptotic optimality of our procedure for histogram selection in density estimation, by establishing sharp oracle inequalities for the Kullback-Leibler divergence. A strong feature of our theoretical results is that they include the estimation of unbounded log-densities. To do so, we prove several analytical and probabilistic lemmas that are of independent interest. In an experimental study, we also demonstrate state-of-the-art performance of our over-penalization procedure for bin size selection.Keywords
All Related Versions
This publication has 34 references indexed in Scilit:
- Learning from MOM’s principles: Le Cam’s approachStochastic Processes and their Applications, 2019
- On Concentration for (Regularized) Empirical Risk MinimizationSankhya A, 2017
- A high-dimensional Wilks phenomenonProbability Theory and Related Fields, 2010
- Estimating the intensity of a random measure by histogram type estimatorsProbability Theory and Related Fields, 2008
- Model selection via testing: an alternative to (penalized) maximum likelihood estimatorsAnnales de l'Institut Henri Poincaré, Probabilités et Statistiques, 2006
- How many bins should be put in a regular histogramESAIM: Probability and Statistics, 2006
- Estimation of equifrequency histogramsStatistics & Probability Letters, 2002
- A Bennett concentration inequality and its application to suprema of empirical processesComptes Rendus Mathématique, 2002
- Risk bounds for model selection via penalizationProbability Theory and Related Fields, 1999
- Model selection for least absolute deviations regression in small samplesStatistics & Probability Letters, 1990