SILAGE is a memory-efficient, full-gradient-free method for nonconvex optimization over nested double finite-sum problems, removing periodic full-gradient refreshes while using only O(n) memory and adapting to across- and within-group data heterogeneity.
Jun 14, 2026
A semi-asynchronous distributed optimization method (Rennala-NSGD) for stochastic optimization beyond Euclidean geometry, with convergence and wall-clock-time guarantees. Manuscript under review; preprint forthcoming.
Mar 1, 2026
We study normalized error feedback algorithms with momentum and parameter-agnostic stepsizes, eliminating the need for problem-dependent tuning while achieving competitive convergence rates.
Nov 1, 2025
We introduce Bernoulli-LoRA, a theoretical framework for randomized Low-Rank Adaptation that unifies existing approaches and provides convergence guarantees for various optimization methods.
Aug 1, 2025
We extend MARINA-P algorithm to non-smooth federated optimization, providing the first theoretical analysis with server-to-worker compression and adaptive stepsizes while achieving optimal convergence rates.
Dec 22, 2024