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. 2310.18860
20
3

Bayes beats Cross Validation: Efficient and Accurate Ridge Regression via Expectation Maximization

29 October 2023
Shu Yu Tew
Mario Boley
Daniel F. Schmidt
    UQCV
ArXivPDFHTML
Abstract

We present a novel method for tuning the regularization hyper-parameter, λ\lambdaλ, of a ridge regression that is faster to compute than leave-one-out cross-validation (LOOCV) while yielding estimates of the regression parameters of equal, or particularly in the setting of sparse covariates, superior quality to those obtained by minimising the LOOCV risk. The LOOCV risk can suffer from multiple and bad local minima for finite nnn and thus requires the specification of a set of candidate λ\lambdaλ, which can fail to provide good solutions. In contrast, we show that the proposed method is guaranteed to find a unique optimal solution for large enough nnn, under relatively mild conditions, without requiring the specification of any difficult to determine hyper-parameters. This is based on a Bayesian formulation of ridge regression that we prove to have a unimodal posterior for large enough nnn, allowing for both the optimal λ\lambdaλ and the regression coefficients to be jointly learned within an iterative expectation maximization (EM) procedure. Importantly, we show that by utilizing an appropriate preprocessing step, a single iteration of the main EM loop can be implemented in O(min⁡(n,p))O(\min(n, p))O(min(n,p)) operations, for input data with nnn rows and ppp columns. In contrast, evaluating a single value of λ\lambdaλ using fast LOOCV costs O(nmin⁡(n,p))O(n \min(n, p))O(nmin(n,p)) operations when using the same preprocessing. This advantage amounts to an asymptotic improvement of a factor of lll for lll candidate values for λ\lambdaλ (in the regime q,p∈O(n)q, p \in O(\sqrt{n})q,p∈O(n​) where qqq is the number of regression targets).

View on arXiv
Comments on this paper