ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1912.10784
108
29
v1v2v3 (latest)

An improper estimator with optimal excess risk in misspecified density estimation and logistic regression

23 December 2019
Jaouad Mourtada
Stéphane Gaïffas
ArXiv (abs)PDFHTML
Abstract

We introduce a procedure for conditional density estimation under logarithmic loss, which we call SMP (Sample Minmax Predictor). This estimator minimizes a new general excess risk bound for statistical learning. On standard examples, this bound scales as d/nd/nd/n with ddd the model dimension and nnn the sample size, and critically remains valid under model misspecification. Being an improper (out-of-model) procedure, SMP improves over within-model estimators such as the maximum likelihood estimator, whose excess risk degrades under misspecification. Compared to approaches reducing to the sequential problem, our bounds remove suboptimal log⁡n\log nlogn factors and can handle unbounded classes. For the Gaussian linear model, the predictions and risk bound of SMP are governed by leverage scores of covariates, nearly matching the optimal risk in the well-specified case without conditions on the noise variance or approximation error of the linear model. For logistic regression, SMP provides a non-Bayesian approach to calibration of probabilistic predictions relying on virtual samples, and can be computed by solving two logistic regressions. It achieves a non-asymptotic excess risk of O((d+B2R2)/n)O((d + B^2R^2)/n)O((d+B2R2)/n), where RRR bounds the norm of features and BBB that of the comparison parameter; by contrast, no within-model estimator can achieve better rate than min⁡(BR/n,deBR/n)\min({B R}/{\sqrt{n}}, {d e^{BR}}/{n} )min(BR/n​,deBR/n) in general. This provides a more practical alternative to Bayesian approaches, which require approximate posterior sampling, thereby partly addressing a question raised by Foster et al. (2018).

View on arXiv
Comments on this paper