math.OCJul 28, 2026

Variance-Reduced Conditional Gradient Methods under Markovian Sampling for Nonconvex Composite Optimization

Authors: Zhaojun Peng

Organizations: School of Computer Science Nanjing University of Information Science and Technology

Abstract

We study stochastic composite nonconvex optimization over a compact convex set when gradient samples arrive along a single trajectory of a fixed ergodic Markov chain. Existing single-trajectory variance-reduction theory covers smooth unconstrained objectives; we address the projection-free composite setting using the generalized Frank-Wolfe gap. We propose MC-ALFCG, which combines a momentum conditional-gradient method with coupled capped multilevel Monte Carlo estimation and per-iteration clipping. The deepest nested average uses consecutive states from the same trajectory, yielding conditional bias O(τmix/T)O(τ_{\mathrm{mix}}/T) uniformly over the starting state, while coupling controls the gradient-difference second moment through the iterate displacement. Clipping enforces the pathwise bounds needed by the adaptive analysis. We reduce the Markovian recursion to its independent-sampling counterpart under σ22ΛGσ2σ^2\mapsto 2ΛG_σ^2 and L22ΛL2L^2\mapsto 2ΛL^2, where Λ=O(τmixlogT)Λ=O(τ_{\mathrm{mix}}\log T). For positive centered noise, the tuned method achieves expected sample complexity O~((τmix2Gσ+τmix5/2Gσ2)ε3+τmix5ε2)\widetilde{O}((τ_{\mathrm{mix}}^2G_σ+τ_{\mathrm{mix}}^{5/2}G_σ^2)\varepsilon^{-3}+τ_{\mathrm{mix}}^5\varepsilon^{-2}). The exactly noiseless specialization achieves O~(ε2)\widetilde{O}(\varepsilon^{-2}) with mixing-time-free constants, while a mixing-time-oblivious variant achieves O~(τmix6ε3+τmix3ε2)\widetilde{O}(τ_{\mathrm{mix}}^6\varepsilon^{-3}+τ_{\mathrm{mix}}^3\varepsilon^{-2}). All guarantees are in expectation under a fixed transition kernel. Controlled numerical studies examine dependence sensitivity, a nonconvex composite instance, and clipping behavior.

Explore similar work

May 14, 2026math.OC

Stochastic Compositional Optimization via Hybrid Momentum Frank--Wolfe

Stochastic compositional optimization minimizes objectives of the form minxXF(f(x),x)\min_{\bm{x} \in \mathcal{X}} F(\bm{f}(\bm{x}), \bm{x}), where f\bm{f} is accessible only through noisy stochastic queries. Existing methods for this problem assume that the outer function FF is continuously differentiable, which excludes many practically important applications such as robust max-of-losses, Conditional Value-at-Risk, and norm regularizers. We propose the Hybrid Momentum Stochastic Frank--Wolfe algorithm, which drops the smoothness assumption on FF. By combining a momentum-based Jacobian tracker with a Taylor-corrected function tracker, the algorithm feeds an entire stochastic linearization -- rather than a single gradient -- into a generalized linear minimization oracle. We establish an O(K1/4)\mathcal{O}(K^{-1/4}) convergence rate in the generalized Frank--Wolfe gap for non-convex objectives with LFL_F-Lipschitz outer functions, matching the optimal complexity for projection-free single-sample stochastic methods under expected smoothness. The analysis extends to heavy-tailed noise oracles with bounded rr-th moments for r(1,2]r \in (1, 2] and recovers the deterministic rates of Vladarean et al (2023) as the noise vanishes.
El Mahdi Chayti
May 12, 2026math.ST

A proximal gradient algorithm for composite log-concave sampling

We propose an algorithm to sample from composite log-concave distributions over Rd\mathbb{R}^d, i.e., densities of the form πefgπ\propto e^{-f-g}, assuming access to gradient evaluations of ff and a restricted Gaussian oracle (RGO) for gg. The latter requirement means that we can easily sample from the density RGOg,h,y(x)exp(g(x)12hyx2)\text{RGO}_{g,h,y}(x) \propto \exp(-g(x) -\frac{1}{2h}||y-x||^2), which is the sampling analogue of the proximal operator for gg. If f+gf + g is αα-strongly convex and ff is ββ-smooth, our sampler achieves ε\varepsilon error in total variation distance in O~(κdlog4(1/ε))\widetilde{\mathcal O}(κ\sqrt d \log^4(1/\varepsilon)) iterations where κ:=β/ακ:= β/α, which matches prior state-of-the-art results for the case g=0g=0. We further extend our results to cases where (1) ππ is non-log-concave but satisfies a Poincaré or log-Sobolev inequality, and (2) ff is non-smooth but Lipschitz.
Linghai Liu, Sinho Chewi
Sep 14, 2026math.OC

Projection-Free Multi-level Algorithms for Stochastic Constrained Compositional Optimization

This paper studies projection-free algorithms for stochastic constrained multi-level compositional optimization. In this context, the objective function is a nested composition of several smooth functions, and the decision set is closed and convex. Since projection onto the constraint set can be computationally expensive, we develop projection-free methods that rely on linear minimization oracles. For non-convex objectives, we propose variance-reduced projection-free algorithms and establish complexity guarantees under both the Frank-Wolfe gap and the gradient mapping criteria. We also develop momentum-based methods that achieve convergence guarantees under weaker smoothness assumptions. Additionally, by using a stage-wise design, we derive a parameter-free variant that preserves the same complexities for the Frank-Wolfe gap. Such a design can be further used to develop algorithms for convex and strongly convex functions whose rates match those of single-level projection-free counterparts. Finally, we consider finite-sum problems and derive complexities for non-convex, convex, and strongly convex objectives. Numerical experiments across multiple tasks demonstrate the effectiveness of the proposed methods.
Wei Jiang, Sifan Yang, Wenhao Yang +4