English
Related papers

Related papers: Training Instabilities Induce Flatness Bias in Gra…

200 papers

Recent findings (e.g., arXiv:2103.00065) demonstrate that modern neural networks trained by full-batch gradient descent typically enter a regime called Edge of Stability (EOS). In this regime, the sharpness, i.e., the maximum Hessian…

Machine Learning · Computer Science 2022-11-04 Zhouzi Li , Zixuan Wang , Jian Li

Empirically, modern deep learning training often occurs at the Edge of Stability (EoS), where the sharpness of the loss exceeds the threshold below which classical convergence analysis applies. Despite recent progress, existing theoretical…

Machine Learning · Computer Science 2026-04-14 Eric Gan

Hessian based measures of flatness, such as the trace, Frobenius and spectral norms, have been argued, used and shown to relate to generalisation. In this paper we demonstrate that for feed forward neural networks under the cross entropy…

Machine Learning · Statistics 2020-06-17 Diego Granziol

We discover restrained numerical instabilities in current training practices of deep networks with stochastic gradient descent (SGD), and its variants. We show numerical error (on the order of the smallest floating point bit and thus the…

Machine Learning · Computer Science 2024-06-13 Yuxin Sun , Dong Lao , Ganesh Sundaramoorthi , Anthony Yezzi

Recent research shows that when Gradient Descent (GD) is applied to neural networks, the loss almost never decreases monotonically. Instead, the loss oscillates as gradient descent converges to its ''Edge of Stability'' (EoS). Here, we find…

Machine Learning · Computer Science 2023-05-23 Itai Kreisler , Mor Shpigel Nacson , Daniel Soudry , Yair Carmon

In deep learning, it is common to use more network parameters than training points. In such scenarioof over-parameterization, there are usually multiple networks that achieve zero training error so that thetraining algorithm induces an…

Machine Learning · Computer Science 2023-08-22 Hung-Hsu Chou , Carsten Gieshoff , Johannes Maly , Holger Rauhut

The deep learning recipe of casting real-world problems as mathematical optimisation and tackling the optimisation by training deep neural networks using gradient-based optimisation has undoubtedly proven to be a fruitful one. The…

Machine Learning · Statistics 2023-10-24 Mihaela Claudia Rosca

We study the learning performance of gradient descent when the empirical risk is weakly convex, namely, the smallest negative eigenvalue of the empirical risk's Hessian is bounded in magnitude. By showing that this eigenvalue can control…

Machine Learning · Statistics 2021-06-02 Dominic Richards , Mike Rabbat

Understanding the dynamics of optimization in deep learning is increasingly important as models scale. While stochastic gradient descent (SGD) and its variants reliably find solutions that generalize well, the mechanisms driving this…

Machine Learning · Computer Science 2026-04-07 Wei-Kai Chang , Rajiv Khanna

The phenomenon that stochastic gradient descent (SGD) favors flat minima has played a critical role in understanding the implicit regularization of SGD. In this paper, we provide an explanation of this striking phenomenon by relating the…

Machine Learning · Statistics 2022-10-18 Lei Wu , Mingze Wang , Weijie Su

Recent research has observed that in machine learning optimization, gradient descent (GD) often operates at the edge of stability (EoS) [Cohen, et al., 2021], where the stepsizes are set to be large, resulting in non-monotonic losses…

Machine Learning · Computer Science 2023-10-17 Jingfeng Wu , Vladimir Braverman , Jason D. Lee

Full-batch gradient descent on neural networks drives the largest Hessian eigenvalue to the threshold $2/\eta$, where $\eta$ is the learning rate. This phenomenon, the Edge of Stability, has resisted a unified explanation: existing accounts…

Machine Learning · Computer Science 2026-04-23 Elon Litman

Recent findings by Cohen et al., 2021, demonstrate that when training neural networks using full-batch gradient descent with a step size of $\eta$, the largest eigenvalue $\lambda_{\max}$ of the full-batch Hessian consistently stabilizes…

Machine Learning · Computer Science 2025-12-30 Arseniy Andreyev , Pierfrancesco Beneventano

Stochastic Gradient Descent (SGD) based training of neural networks with a large learning rate or a small batch-size typically ends in well-generalizing, flat regions of the weight space, as indicated by small eigenvalues of the Hessian of…

Machine Learning · Statistics 2019-12-24 Stanisław Jastrzębski , Zachary Kenton , Nicolas Ballas , Asja Fischer , Yoshua Bengio , Amos Storkey

Cohen et al. (2021) empirically study the evolution of the largest eigenvalue of the loss Hessian, also known as sharpness, along the gradient descent (GD) trajectory and observe the Edge of Stability (EoS) phenomenon. The sharpness…

Machine Learning · Computer Science 2023-10-27 Minhak Song , Chulhee Yun

Stochastic gradient descent (SGD) is central to deep learning, yet the dynamical origin of its preference for flatter, more generalizable solutions remains unclear. Here, by analyzing SGD learning dynamics, we identify a nonequilibrium…

Machine Learning · Computer Science 2026-01-19 Ning Yang , Yikuan Zhang , Qi Ouyang , Chao Tang , Yuhai Tu

It is important to understand how the popular regularization method dropout helps the neural network training find a good generalization solution. In this work, we show that the training with dropout finds the neural network with a flatter…

Machine Learning · Computer Science 2022-05-24 Zhongwang Zhang , Hanxu Zhou , Zhi-Qin John Xu

Stochastic Gradient Descent (SGD) and its variants are mainstream methods for training deep networks in practice. SGD is known to find a flat minimum that often generalizes well. However, it is mathematically unclear how deep learning can…

Machine Learning · Computer Science 2021-01-18 Zeke Xie , Issei Sato , Masashi Sugiyama

Recent empirical advances show that training deep models with large learning rate often improves generalization performance. However, theoretical justifications on the benefits of large learning rate are highly limited, due to challenges in…

Machine Learning · Computer Science 2022-03-01 Yuqing Wang , Minshuo Chen , Tuo Zhao , Molei Tao

Recent empirical and theoretical work has shown that the dynamics of the large eigenvalues of the training loss Hessian have some remarkably robust features across models and datasets in the full batch regime. There is often an early period…

Machine Learning · Computer Science 2025-02-04 Atish Agarwala , Jeffrey Pennington