Skip to content

Nonsmooth Optimization via Orthogonalized Momentum

Sep 2026 · 1 citation · 77 references
Mathematics Computer Science

Abstract

Modern real application problems involve matrix-valued parameters, yet conventional optimizers treat them as vectors, thereby motivating matrix-aware methods that exploit input-output geometry, such as Muon which orthogonalizes the momentum matrices before parameter updates. Its empirical success raises a conceptual question: can orthogonalized momentum remain effective beyond smooth optimization? This paper studies this question for locally Lipschitz functions using a generalized derivative framework compatible with backpropagation. Our first contribution is to identify a key limitation: for every fixed momentum factor $\beta\in[0,1)$, Muon can fail to approach the global optimal solution of a convex Lipschitz objective from almost every initialization, when step sizes adapt to the full gradient history. The failure can occur even along bounded iterates. Our example is inspired by the one of Parshakova et al. which only covers $\beta\in[0,\frac{1}{2})$. Then, we show that the obstruction lies in fixed momentum rather than orthogonalization. Indeed, when the momentum factor is adaptive and approaches 1 together with a vanishing step size, Muon recovers asymptotic convergence for nonconvex nonsmooth optimization under the boundedness and regularity conditions. Moreover, we propose MAGD, which combines orthogonalized momentum with gradient, weighted based on their relative progress. MAGD retains asymptotic convergence in nonconvex settings and achieves an $O(\min\{m,n\}\epsilon^{-2})$ rate in convex settings. A lower bound shows the optimal dimension dependence. Experiments on synthetic problems, image classification, and LLM pretraining show MAGD is a simple and practical alternative to Muon. Together, our results characterize when orthogonalized momentum fails without smoothness and how it can be made reliable and we hope that the analysis may be useful more broadly.

View source

Similar papers

#machine learning Preprint Oct 2026

Optimal Algorithms for Nonsmooth Nonconvex Stochastic Optimization with Constraints

We present new algorithms for stochastic optimization problems in which the objective is neither smooth nor convex, subject to a convex constraint set. Our algorithms return a $(\delta,\varepsilon)$-Goldstein stationary point, as recently defined for constrained problems by Liu et al. (2024), with improved complexities...

G. Buzaglo, Guy Kornowski · 0 citations
Preprint Aug 2026

On Nonsmooth and Relatively Weakly Convex Minimization

A model-based minimization method is developed under a relative Lipschitz condition and a framework with convergence guarantees is extended to the setting where the distance generating function and its gradient are accessible only through a stochastic oracle.

Chen Jiang, Jia-Wen Bi, Shu-Zhong Zhang · 0 citations
#machine learning Preprint Oct 2026

Optimal Momentum Methods for Stochastic Multilevel Compositional Optimization

A momentum-based estimator with mini-batches is employed to track the function values of each level, which are subsequently used to construct momentum gradient estimators, and a batch-free method that incorporates a first-order approximation and a clipping technique for function value estimation is developed.

Wei Jiang, Rui Yan, Si-Fan Yang et al. · 0 citations
Preprint Sep 2026

Projected Subgradient Methods for a Class of Nonsmooth and Nonconvex Optimization Problems

We investigate the optimization problem of minimizing a nonsmooth function that satisfies a nonsmooth version of the descent lemma over a nonempty and closed but not necessarily convex set. The objective function belongs to the class of upper-$\mathcal{C}^2$ functions, whereas the constraints may promote a sparse or lo...

Christian Kanzow, Jannis Krüger, Leo Lehmann · 0 citations
Preprint Sep 2026

Constrained Nonconvex Stochastic Optimization with One Projection

Constrained nonconvex optimization has seen increasing application in modern machine learning, such as safe LLM alignment/finetuning, transfer learning and rank-constrained continual learning. The commonly used approach is Projected SGD which requires projection at every iteration. However, projection onto a functional...

Yu-Yang Deng, Mohammadreza M. Kalan, Eitan J. Neugut et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.