A Compositional Kernel Model for Feature Learning

Feng Ruan; Keli Liu; Michael Jordan

Home/Archive/arXiv

arXiv · 2509.14158

A Compositional Kernel Model for Feature Learning

Feng Ruan·Keli Liu·Michael Jordan

View original source ↗Read source document ↗Export metadata ↓

Abstract

We study a compositional variant of kernel ridge regression in which the predictor is applied to a coordinate-wise reweighting of the inputs. Formulated as a variational problem, this model provides a tractable setting for studying feature learning in compositional architectures. From the perspective of variable selection, we show how relevant variables are recovered while noise variables are eliminated. We prove that both global minimizers and stationary points discard noise coordinates when the noise variables are Gaussian distributed. A central finding is that $\ell_1$-type kernels, such as the Laplace kernel, succeed in recovering features contributing to nonlinear effects at stationary points, whereas Gaussian kernels recover only linear ones.

Explore related subjects

cs.LG math.OC

Keep this discovery

BibTeX RIS

Feng Ruan, Keli Liu, Michael Jordan. 2026-09-01. A Compositional Kernel Model for Feature Learning. https://arxiv.org/abs/2509.14158

Cite the original work for its findings. Save a collection to share your selection of sources.

Discover connections

cs.LG subject math.OC subject

Connections use source metadata and explicit phrase matches, not verified experimental comparisons.

KEEP EXPLORING

Related papers

arXiv2026

Rates of Convergence in the Central Limit Theorem for Markov Chains, with an Application to TD Learning

We prove a non-asymptotic central limit theorem for vector-valued martingale differences using Stein's method, and use Poisson's equation to extend the result to functions of Markov Chains. We then show that these results can be applied to establish a non-asymptotic central limit theorem for Temporal Difference (TD) learning with averaging.

math.PR↗

arXiv2025

Momentum-based gradient descent methods for Lie groups

Polyak's Heavy Ball (PHB; Polyak, 1964), a.k.a. Classical Momentum, and Nesterov's Accelerated Gradient (NAG; Nesterov, 1983) are well-established momentum-descent methods for optimization. Although the latter generally outperforms the former, primarily, generalizations of PHB-like methods to nonlinear spaces have not been sufficiently explored in the literature. In this paper, we propose a generalization of NAG-like methods for Lie group optimization. This generalization is based on the variational one-to-one correspondence between classical and accelerated momentum methods (Campos et al., 2023). We provide numerical experiments for chosen retractions on the group of rotations based on the Frobenius norm and the Rosenbrock function to demonstrate the effectiveness of our proposed methods, and that align with results of the Euclidean case, that is, a faster convergence rate for NAG.

math.OC↗

arXiv2026

Sample Complexity of Linear Quadratic Regulator Without Initial Stability

Inspired by REINFORCE, we introduce a novel receding-horizon algorithm for the Linear Quadratic Regulator (LQR) problem with unknown dynamics. Unlike prior methods, our algorithm avoids reliance on two-point gradient estimates while maintaining the same order of sample complexity. Furthermore, it eliminates the restrictive requirement of starting with a stable initial policy, broadening its applicability. Beyond these improvements, we introduce a refined analysis of error propagation through the contraction of the Riccati operator under the Riemannian distance. This refinement leads to a better sample complexity and ensures improved convergence guarantees.

math.OC↗