English
Related papers

Related papers: Does Weight Decay Enhance Training Stability?

200 papers

Curvature information -- particularly, the largest eigenvalue of the loss Hessian, known as the sharpness -- often forms the basis for learning rate tuners. However, recent work has shown that the curvature information undergoes complex…

Machine Learning · Computer Science 2024-07-09 Vincent Roulet , Atish Agarwala , Jean-Bastien Grill , Grzegorz Swirszcz , Mathieu Blondel , Fabian Pedregosa

The Edge of Stability (EoS) is a phenomenon where the sharpness (largest eigenvalue) of the Hessian approaches and then hovers near the stability threshold $2/\eta$ during gradient descent (GD) with step size $\eta$. Despite (apparently)…

Machine Learning · Computer Science 2026-05-29 Rustem Islamov , Michael Crawshaw , Jeremy Cohen , Robert Gower

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

A key challenge in building theoretical foundations for deep learning is the complex optimization dynamics of neural networks, resulting from the high-dimensional interactions between the large number of network parameters. Such non-trivial…

Machine Learning · Computer Science 2021-12-07 Mohammad Pezeshki , Amartya Mitra , Yoshua Bengio , Guillaume Lajoie

Weight decay is often used to ensure good generalization in the training practice of deep neural networks with batch normalization (BN-DNNs), where some convolution layers are invariant to weight rescaling due to the normalization. In this…

Machine Learning · Computer Science 2022-06-22 Ziquan Liu , Yufei Cui , Jia Wan , Yu Mao , Antoni B. Chan

We analyze the landscape and training dynamics of diagonal linear networks in a linear regression task, with the network parameters being perturbed by small isotropic normal noise. The addition of such noise may be interpreted as a…

Machine Learning · Computer Science 2025-03-18 Gabriel Clara , Sophie Langer , Johannes Schmidt-Hieber

Learning rate decay (lrDecay) is a \emph{de facto} technique for training modern neural networks. It starts with a large learning rate and then decays it multiple times. It is empirically observed to help both optimization and…

Machine Learning · Computer Science 2019-09-27 Kaichao You , Mingsheng Long , Jianmin Wang , Michael I. Jordan

This paper seeks to answer the question: as the (near-) orthogonality of weights is found to be a favorable property for training deep convolutional neural networks, how can we enforce it in more effective and easy-to-use ways? We develop…

Machine Learning · Computer Science 2018-10-23 Nitin Bansal , Xiaohan Chen , Zhangyang Wang

The merits of fast convergence and potentially better performance of the weight normalization family have drawn increasing attention in recent years. These methods use standardization or normalization that changes the weight…

Machine Learning · Computer Science 2019-11-15 Li Xiang , Chen Shuo , Xia Yan , Yang Jian

In the usual deep neural network optimization process, the learning rate is the most important hyper parameter, which greatly affects the final convergence effect. The purpose of learning rate is to control the stepsize and gradually reduce…

Machine Learning · Computer Science 2019-05-02 Jiakai Wei

Gradient-based optimization of neural differential equations and other parameterized dynamical systems fundamentally relies on the ability to differentiate numerical solutions with respect to model parameters. In stiff systems, it has been…

Machine Learning · Computer Science 2025-08-05 Colby Fronk , Linda Petzold

Loss of trainability refers to a phenomenon in continual learning where parameter updates no longer make progress on the optimization objective, so accuracy stalls or degrades as the learning problem changes over time. In this paper, we…

Machine Learning · Computer Science 2025-12-11 Gunbir Singh Baveja , Alex Lewandowski , Mark Schmidt

We empirically demonstrate that full-batch gradient descent on neural network training objectives typically operates in a regime we call the Edge of Stability. In this regime, the maximum eigenvalue of the training loss Hessian hovers just…

Machine Learning · Computer Science 2022-11-24 Jeremy M. Cohen , Simran Kaur , Yuanzhi Li , J. Zico Kolter , Ameet Talwalkar

Learned optimizers -- neural networks that are trained to act as optimizers -- have the potential to dramatically accelerate training of machine learning models. However, even when meta-trained across thousands of tasks at huge…

Machine Learning · Computer Science 2022-09-23 James Harrison , Luke Metz , Jascha Sohl-Dickstein

The stability and generalization of stochastic gradient-based methods provide valuable insights into understanding the algorithmic performance of machine learning models. As the main workhorse for deep learning, stochastic gradient descent…

Machine Learning · Statistics 2021-02-24 Tao Sun , Dongsheng Li , Bao Wang

We analyze gradient descent with randomly weighted data points in a linear regression model, under a generic weighting distribution. This includes various forms of stochastic gradient descent, importance sampling, but also extends to…

Machine Learning · Statistics 2025-12-12 Gabriel Clara , Yazan Mash'al

In recent years, continual learning, a prediction setting in which the problem environment may evolve over time, has become an increasingly popular research field due to the framework's gearing towards complex, non-stationary objectives.…

Machine Learning · Computer Science 2024-09-27 Max Koster , Jude Kukla

We conduct a comprehensive investigation into the dynamics of gradient descent using large-order constant step-sizes in the context of quadratic regression models. Within this framework, we reveal that the dynamics can be encapsulated by a…

Machine Learning · Computer Science 2023-10-04 Xuxing Chen , Krishnakumar Balasubramanian , Promit Ghosal , Bhavya Agrawalla

Neoteric works have shown that modern deep learning models can exhibit a sparse double descent phenomenon. Indeed, as the sparsity of the model increases, the test performance first worsens since the model is overfitting the training data;…

Machine Learning · Computer Science 2024-02-09 Victor Quétu , Enzo Tartaglione

During neural network training, the sharpness of the Hessian matrix of the training loss rises until training is on the edge of stability. As a result, even nonstochastic gradient descent does not accurately model the underlying dynamical…

Machine Learning · Statistics 2024-06-04 Mark Lowell , Catharine Kastner