English

Algorithms for Learning Sparse Additive Models with Interactions in High Dimensions

Machine Learning 2017-05-09 v3 Information Theory math.IT Numerical Analysis Machine Learning

Abstract

A function f:RdRf: \mathbb{R}^d \rightarrow \mathbb{R} is a Sparse Additive Model (SPAM), if it is of the form f(x)=lSϕl(xl)f(\mathbf{x}) = \sum_{l \in \mathcal{S}}\phi_{l}(x_l) where S[d]\mathcal{S} \subset [d], Sd|\mathcal{S}| \ll d. Assuming ϕ\phi's, S\mathcal{S} to be unknown, there exists extensive work for estimating ff from its samples. In this work, we consider a generalized version of SPAMs, that also allows for the presence of a sparse number of second order interaction terms. For some S1[d],S2([d]2)\mathcal{S}_1 \subset [d], \mathcal{S}_2 \subset {[d] \choose 2}, with S1d,S2d2|\mathcal{S}_1| \ll d, |\mathcal{S}_2| \ll d^2, the function ff is now assumed to be of the form: pS1ϕp(xp)+(l,l)S2ϕ(l,l)(xl,xl)\sum_{p \in \mathcal{S}_1}\phi_{p} (x_p) + \sum_{(l,l^{\prime}) \in \mathcal{S}_2}\phi_{(l,l^{\prime})} (x_l,x_{l^{\prime}}). Assuming we have the freedom to query ff anywhere in its domain, we derive efficient algorithms that provably recover S1,S2\mathcal{S}_1,\mathcal{S}_2 with finite sample bounds. Our analysis covers the noiseless setting where exact samples of ff are obtained, and also extends to the noisy setting where the queries are corrupted with noise. For the noisy setting in particular, we consider two noise models namely: i.i.d Gaussian noise and arbitrary but bounded noise. Our main methods for identification of S2\mathcal{S}_2 essentially rely on estimation of sparse Hessian matrices, for which we provide two novel compressed sensing based schemes. Once S1,S2\mathcal{S}_1, \mathcal{S}_2 are known, we show how the individual components ϕp\phi_p, ϕ(l,l)\phi_{(l,l^{\prime})} can be estimated via additional queries of ff, with uniform error bounds. Lastly, we provide simulation results on synthetic data that validate our theoretical findings.

Keywords

Cite

@article{arxiv.1605.00609,
  title  = {Algorithms for Learning Sparse Additive Models with Interactions in High Dimensions},
  author = {Hemant Tyagi and Anastasios Kyrillidis and Bernd Gärtner and Andreas Krause},
  journal= {arXiv preprint arXiv:1605.00609},
  year   = {2017}
}

Comments

To appear in Information and Inference: A Journal of the IMA. Made following changes after review process: (a) Corrected typos throughout the text. (b) Corrected choice of sampling distribution in Section 5, see eqs. (5.2), (5.3). (c) More detailed comparison with existing work in Section 8. (d) Added Section B in appendix on roots of cubic equation

R2 v1 2026-06-22T13:47:00.139Z