Fast Spawn\&Prune (FS\&P): Global convergence of stochastic conic particle gradient descent via birth/death process

TL;DR

This paper introduces FS&P, a stochastic algorithm combining conic particle gradient descent with birth-death processes, providing the first global convergence guarantees for such methods in sparse regression.

Contribution

It develops a novel FS&P algorithm with theoretical guarantees of global convergence and explicit convergence rates, addressing local minima issues in non-convex sparse regression.

Findings

First theoretical guarantee of global convergence for this class of algorithms.

Explicit convergence rates for excess risk depending on iterations and dimension.

Sample complexity bounds and a horizon-free variant are derived.

Abstract

We investigate the global optimization of the objective function arising in continuous sparse regression, specifically the Beurling LASSO (BLASSO), over the space of measures. While Conic Particle Gradient Descent (CPGD) methods are computationally efficient, they may become trapped in local minima due to the non-convexity of the parameterization. To overcome this limitation, we introduce Fast Spawn\&Prune (FS\&P), a stochastic algorithm that extends FastPart introduced in De Castro et al. (2025) and combines CPGD with a birth-death process. The birth mechanism ensures asymptotic global exploration by introducing particles in regions where first-order optimality conditions are violated, while the death process preserves computational efficiency by pruning non-informative particles. We provide the first theoretical guarantee of global convergence for this class of discrete-time stochastic algorithms, without requiring exponentially large initializations. Furthermore, we derive explicit convergence rates for the excess risk, which scale as $\mathcal{O}\big(\left(\log K / K\right)^{\frac{1}{2(2+d)}}\big)$, where $K$ denotes the number of iterations and d the dimension of the domain, thereby quantifying the trade-off between global exploration and local refinement. Moreover, the sample complexity is $\mathcal{O}\big(N^{-\frac{1}{4(2+d)}}\big)$ (up to logarithmic factors). We also propose a horizon-free variant that does not require prior knowledge of the iteration budget.