English

Scalable Spike-and-Slab

Computation 2022-06-28 v2 Machine Learning Methodology Machine Learning

Abstract

Spike-and-slab priors are commonly used for Bayesian variable selection, due to their interpretability and favorable statistical properties. However, existing samplers for spike-and-slab posteriors incur prohibitive computational costs when the number of variables is large. In this article, we propose Scalable Spike-and-Slab (S3S^3), a scalable Gibbs sampling implementation for high-dimensional Bayesian regression with the continuous spike-and-slab prior of George and McCulloch (1993). For a dataset with nn observations and pp covariates, S3S^3 has order max{n2pt,np}\max\{ n^2 p_t, np \} computational cost at iteration tt where ptp_t never exceeds the number of covariates switching spike-and-slab states between iterations tt and t1t-1 of the Markov chain. This improves upon the order n2pn^2 p per-iteration cost of state-of-the-art implementations as, typically, ptp_t is substantially smaller than pp. We apply S3S^3 on synthetic and real-world datasets, demonstrating orders of magnitude speed-ups over existing exact samplers and significant gains in inferential quality over approximate samplers with comparable cost.

Keywords

Cite

@article{arxiv.2204.01668,
  title  = {Scalable Spike-and-Slab},
  author = {Niloy Biswas and Lester Mackey and Xiao-Li Meng},
  journal= {arXiv preprint arXiv:2204.01668},
  year   = {2022}
}

Comments

Accepted to ICML 2022. Open-source software in Python and R available at https://github.com/niloyb/ScaleSpikeSlab

R2 v1 2026-06-24T10:37:21.591Z