English
Related papers

Related papers: On progressive sharpening, flat minima and general…

200 papers

Generalizing to out-of-distribution (OOD) data or unseen domain, termed OOD generalization, still lacks appropriate theoretical guarantees. Canonical OOD bounds focus on different distance measurements between source and target domains but…

Machine Learning · Computer Science 2024-03-12 Yingtian Zou , Kenji Kawaguchi , Yingnan Liu , Jiashuo Liu , Mong-Li Lee , Wynne Hsu

While stochastic gradient descent (SGD) and variants have been surprisingly successful for training deep nets, several aspects of the optimization dynamics and generalization are still not well understood. In this paper, we present new…

Machine Learning · Computer Science 2019-07-26 Xinyan Li , Qilong Gu , Yingxue Zhou , Tiancong Chen , Arindam Banerjee

Deep neural networks are known to suffer from exploding or vanishing gradients as depth increases, a phenomenon closely tied to the spectral behavior of the input-output Jacobian. Prior work has identified critical initialization schemes…

Machine Learning · Computer Science 2025-11-25 Benjamin Dadoun , Soufiane Hayou , Hanan Salam , Mohamed El Amine Seddik , Pierre Youssef

Machine learning techniques enhanced by noisy intermediate-scale quantum (NISQ) devices and especially variational quantum circuits (VQC) have recently attracted much interest and have already been benchmarked for certain problems. Inspired…

Quantum Physics · Physics 2021-03-24 Patrick Huembeli , Alexandre Dauphin

Classical assumptions like strong convexity and Lipschitz smoothness often fail to capture the nature of deep learning optimization problems, which are typically non-convex and non-smooth, making traditional analyses less applicable. This…

Machine Learning · Computer Science 2025-05-01 Binchuan Qi , Wei Gong , Li Li

When several models have similar training scores, classical model selection heuristics follow Occam's razor and advise choosing the ones with least capacity. Yet, modern practice with large neural networks has often led to situations where…

Machine Learning · Computer Science 2022-11-29 Luis Sa-Couto , Jose Miguel Ramos , Andreas Wichert

The recent theoretical analysis of deep neural networks in their infinite-width limits has deepened our understanding of initialisation, feature learning, and training of those networks, and brought new practical techniques for finding…

Machine Learning · Computer Science 2024-08-23 Taeyoung Kim , Hongseok Yang

Deep neural networks have been demonstrated to achieve phenomenal success in many domains, and yet their inner mechanisms are not well understood. In this paper, we investigate the curvature of image manifolds, i.e., the manifold deviation…

Machine Learning · Computer Science 2023-11-17 Ilya Kaufman , Omri Azencot

We introduce a general theoretical framework, designed for the study of gradient optimisation of deep neural networks, that encompasses ubiquitous architecture choices including batch normalisation, weight normalisation and skip…

Machine Learning · Computer Science 2023-12-05 Lachlan Ewen MacDonald , Jack Valmadre , Hemanth Saratchandran , Simon Lucey

We analyze the training dynamics for deep linear networks using a new metric - layer imbalance - which defines the flatness of a solution. We demonstrate that different regularization methods, such as weight decay or noise data…

Machine Learning · Computer Science 2020-07-21 Boris Ginsburg

Recent advances in deep learning optimization have unveiled two intriguing phenomena under large learning rates: Edge of Stability (EoS) and Progressive Sharpening (PS), challenging classical Gradient Descent (GD) analyses. Current research…

Machine Learning · Computer Science 2025-03-05 Liming Liu , Zixuan Zhang , Simon Du , Tuo Zhao

Understanding the curvature evolution of the loss landscape is fundamental to analyzing the training dynamics of neural networks. The most commonly studied measure, Hessian sharpness ($\lambda_{\max}^H$) -- the largest eigenvalue of the…

Stochastic Gradient Descent (SGD) introduces anisotropic noise that is correlated with the local curvature of the loss landscape, thereby biasing optimization toward flat minima. Prior work often assumes an equivalence between the Fisher…

Machine Learning · Computer Science 2026-02-06 Yikuan Zhang , Ning Yang , Yuhai Tu

One of the main challenges in modern deep learning is to understand why such over-parameterized models perform so well when trained on finite data. A way to analyze this generalization concept is through the properties of the associated…

Machine Learning · Computer Science 2023-07-11 Alison Pouplin , Hrittik Roy , Sidak Pal Singh , Georgios Arvanitidis

Despite extensive study, the significance of sharpness -- the trace of the loss Hessian at local minima -- remains unclear. We investigate an alternative perspective: how sharpness relates to the geometric structure of neural…

Machine Learning · Computer Science 2026-02-24 Shirui Chen , Stefano Recanatesi , Eric Shea-Brown

Decentralized optimization has become a fundamental tool for large-scale learning systems; however, most existing methods rely on the classical Lipschitz smoothness assumption, which is often violated in problems with rapidly varying…

Optimization and Control · Mathematics 2026-01-08 Yanan Bo , Yongqiang Wang

Sharpness-Aware Minimization (SAM) is a recent training method that relies on worst-case weight perturbations which significantly improves generalization in various settings. We argue that the existing justifications for the success of SAM…

Machine Learning · Computer Science 2022-06-14 Maksym Andriushchenko , Nicolas Flammarion

We take a Bayesian perspective to illustrate a connection between training speed and the marginal likelihood in linear models. This provides two major insights: first, that a measure of a model's training speed can be used to estimate its…

Machine Learning · Computer Science 2020-10-28 Clare Lyle , Lisa Schut , Binxin Ru , Yarin Gal , Mark van der Wilk

The properties of flat minima in the empirical risk landscape of neural networks have been debated for some time. Increasing evidence suggests they possess better generalization capabilities with respect to sharp ones. First, we discuss…

We propose a quadratic penalty method for continual learning of neural networks that contain batch normalization (BN) layers. The Hessian of a loss function represents the curvature of the quadratic penalty function, and a…

Machine Learning · Computer Science 2020-04-17 Janghyeon Lee , Hyeong Gwon Hong , Donggyu Joo , Junmo Kim