Convergence of optimizers implies eigenvalues filtering at equilibrium

• URL: http://arxiv.org/abs/2510.09034v1
• Date: Fri, 10 Oct 2025 06:09:14 GMT
• Title: Convergence of optimizers implies eigenvalues filtering at equilibrium
• Authors: Jerome Bolte, Quoc-Tung Le, Edouard Pauwels,
• Abstract summary: We argue that different gradients effectively act as eigenvalue filters determined by their hyper parameters.<n>Inspired by these insights, we propose two novel algorithms that exhibit enhanced eigenvalue filtering.
• Score: 7.901604416781478
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Ample empirical evidence in deep neural network training suggests that a variety of optimizers tend to find nearly global optima. In this article, we adopt the reversed perspective that convergence to an arbitrary point is assumed rather than proven, focusing on the consequences of this assumption. From this viewpoint, in line with recent advances on the edge-of-stability phenomenon, we argue that different optimizers effectively act as eigenvalue filters determined by their hyperparameters. Specifically, the standard gradient descent method inherently avoids the sharpest minima, whereas Sharpness-Aware Minimization (SAM) algorithms go even further by actively favoring wider basins. Inspired by these insights, we propose two novel algorithms that exhibit enhanced eigenvalue filtering, effectively promoting wider minima. Our theoretical analysis leverages a generalized Hadamard--Perron stable manifold theorem and applies to general semialgebraic $C^2$ functions, without requiring additional non-degeneracy conditions or global Lipschitz bound assumptions. We support our conclusions with numerical experiments on feed-forward neural networks.

Related papers

• Lyapunov Stability of Stochastic Vector Optimization: Theory and Numerical Implementation [0.0]
  We use a drift--diffusion model for unconstrained vector optimization in which the drift is induced by a common descent direction.<n>We implement the resulting algorithm as an open-source Python framework for multi-objective optimization.<n> Empirical results on DTLZ2 with objective counts from three to fifteen indicate a consistent trade-off.
  arXiv  Detail & Related papers  (2026-03-04T14:04:24Z)
• Revisiting Zeroth-Order Optimization: Minimum-Variance Two-Point Estimators and Directionally Aligned Perturbations [57.179679246370114]
  We identify the distribution of random perturbations that minimizes the estimator's variance as the perturbation stepsize tends to zero.<n>Our findings reveal that such desired perturbations can align directionally with the true gradient, instead of maintaining a fixed length.
  arXiv  Detail & Related papers  (2025-10-22T19:06:39Z)
• Mirror Descent Under Generalized Smoothness [23.5387392871236]
  We introduce a new $ell*$-smoothness concept that measures the norm of Hessians in terms of a general norm and its dual.<n>We establish convergence for mirror-descent-type algorithms, matching the rates under the classic smoothness.
  arXiv  Detail & Related papers  (2025-02-02T11:23:10Z)
• Error Feedback under $(L_0,L_1)$-Smoothness: Normalization and Momentum [56.37522020675243]
  We provide the first proof of convergence for normalized error feedback algorithms across a wide range of machine learning problems. We show that due to their larger allowable stepsizes, our new normalized error feedback algorithms outperform their non-normalized counterparts on various tasks.
  arXiv  Detail & Related papers  (2024-10-22T10:19:27Z)
• Taming Nonconvex Stochastic Mirror Descent with General Bregman Divergence [25.717501580080846]
  This paper revisits the convergence of gradient Forward Mirror (SMD) in the contemporary non optimization setting. For the training, we develop provably convergent algorithms for the problem of linear networks.
  arXiv  Detail & Related papers  (2024-02-27T17:56:49Z)
• Stable Nonconvex-Nonconcave Training via Linear Interpolation [51.668052890249726]
  This paper presents a theoretical analysis of linearahead as a principled method for stabilizing (large-scale) neural network training. We argue that instabilities in the optimization process are often caused by the nonmonotonicity of the loss landscape and show how linear can help by leveraging the theory of nonexpansive operators.
  arXiv  Detail & Related papers  (2023-10-20T12:45:12Z)
• Robust Stochastic Optimization via Gradient Quantile Clipping [6.2844649973308835]
  We introduce a quant clipping strategy for Gradient Descent (SGD) We use gradient new outliers as norm clipping chains. We propose an implementation of the algorithm using Huberiles.
  arXiv  Detail & Related papers  (2023-09-29T15:24:48Z)
• The Inductive Bias of Flatness Regularization for Deep Matrix Factorization [58.851514333119255]
  This work takes the first step toward understanding the inductive bias of the minimum trace of the Hessian solutions in deep linear networks. We show that for all depth greater than one, with the standard Isometry Property (RIP) on the measurements, minimizing the trace of Hessian is approximately equivalent to minimizing the Schatten 1-norm of the corresponding end-to-end matrix parameters.
  arXiv  Detail & Related papers  (2023-06-22T23:14:57Z)
• Globally Convergent Policy Search over Dynamic Filters for Output Estimation [64.90951294952094]
  We introduce the first direct policy search algorithm convex which converges to the globally optimal $textitdynamic$ filter. We show that informativity overcomes the aforementioned degeneracy.
  arXiv  Detail & Related papers  (2022-02-23T18:06:20Z)
• Nonasymptotic theory for two-layer neural networks: Beyond the bias-variance trade-off [10.182922771556742]
  We present a nonasymptotic generalization theory for two-layer neural networks with ReLU activation function. We show that overparametrized random feature models suffer from the curse of dimensionality and thus are suboptimal.
  arXiv  Detail & Related papers  (2021-06-09T03:52:18Z)
• Robust Implicit Networks via Non-Euclidean Contractions [63.91638306025768]
  Implicit neural networks show improved accuracy and significant reduction in memory consumption. They can suffer from ill-posedness and convergence instability. This paper provides a new framework to design well-posed and robust implicit neural networks.
  arXiv  Detail & Related papers  (2021-06-06T18:05:02Z)
• Lipschitz Bounded Equilibrium Networks [3.2872586139884623]
  This paper introduces new parameterizations of equilibrium neural networks, i.e. networks defined by implicit equations. The new parameterization admits a Lipschitz bound during training via unconstrained optimization. In image classification experiments we show that the Lipschitz bounds are very accurate and improve robustness to adversarial attacks.
  arXiv  Detail & Related papers  (2020-10-05T01:00:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.