English
Related papers

Related papers: A Loss Curvature Perspective on Training Instabili…

200 papers

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

While existing work in robust deep learning has focused on small pixel-level norm-based perturbations, this may not account for perturbations encountered in several real-world settings. In many such cases although test data might not be…

Computer Vision and Pattern Recognition · Computer Science 2021-04-09 Tejas Gokhale , Rushil Anirudh , Bhavya Kailkhura , Jayaraman J. Thiagarajan , Chitta Baral , Yezhou Yang

Curvature influences generalization, robustness, and how reliably neural networks respond to small input perturbations. Existing sharpness metrics are typically defined in parameter space (e.g., Hessian eigenvalues) and can be expensive,…

Machine Learning · Computer Science 2025-11-04 Jacob Poschl

Teams that have trained large Transformer-based models have reported training instabilities at large scale that did not appear when training with the same hyperparameters at smaller scales. Although the causes of such instabilities are of…

Why do neural networks trained with large learning rates for a longer time often lead to better generalization? In this paper, we delve into this question by examining the relation between training and testing loss in neural networks.…

Machine Learning · Computer Science 2024-01-23 Yinuo Ren , Chao Ma , Lexing Ying

We conjecture that the inherent difference in generalisation between adaptive and non-adaptive gradient methods in deep learning stems from the increased estimation noise in the flattest directions of the true loss surface. We demonstrate…

Machine Learning · Statistics 2022-03-17 Diego Granziol , Nicholas Baskerville

Recently, researchers observed that gradient descent for deep neural networks operates in an ``edge-of-stability'' (EoS) regime: the sharpness (maximum eigenvalue of the Hessian) is often larger than stability threshold $2/\eta$ (where…

Machine Learning · Computer Science 2023-02-22 Xingyu Zhu , Zixuan Wang , Xiang Wang , Mo Zhou , Rong Ge

We introduce a novel loss function for training deep learning architectures to perform classification. It consists in minimizing the smoothness of label signals on similarity graphs built at the output of the architecture. Equivalently, it…

Machine Learning · Computer Science 2019-05-02 Myriam Bontonou , Carlos Lassance , Ghouthi Boukli Hacene , Vincent Gripon , Jian Tang , Antonio Ortega

We focus on the robustness of neural networks for classification. To permit a fair comparison between methods to achieve robustness, we first introduce a standard based on the mensuration of a classifier's degradation. Then, we propose…

Computer Vision and Pattern Recognition · Computer Science 2021-03-23 Sadaf Gulshad , Arnold Smeulders

We provide a theoretical explanation for the effectiveness of gradient clipping in training deep neural networks. The key ingredient is a new smoothness condition derived from practical neural network training examples. We observe that…

Optimization and Control · Mathematics 2020-02-12 Jingzhao Zhang , Tianxing He , Suvrit Sra , Ali Jadbabaie

We investigate why deep neural networks suffer from loss of plasticity in deep continual learning, failing to learn new tasks without reinitializing parameters. We show that this failure is preceded by Hessian spectral collapse at new-task…

Machine Learning · Computer Science 2025-09-30 Naicheng He , Kaicheng Guo , Arjun Prakash , Saket Tiwari , Ruo Yu Tao , Tyrone Serapio , Amy Greenwald , George Konidaris

Plasticity, the ability of a neural network to quickly change its predictions in response to new information, is essential for the adaptability and robustness of deep reinforcement learning systems. Deep neural networks are known to lose…

Machine Learning · Computer Science 2023-11-28 Clare Lyle , Zeyu Zheng , Evgenii Nikishin , Bernardo Avila Pires , Razvan Pascanu , Will Dabney

The early phase of training of deep neural networks is critical for their final performance. In this work, we study how the hyperparameters of stochastic gradient descent (SGD) used in the early phase of training affect the rest of the…

Machine Learning · Computer Science 2020-02-25 Stanislaw Jastrzebski , Maciej Szymczak , Stanislav Fort , Devansh Arpit , Jacek Tabor , Kyunghyun Cho , Krzysztof Geras

Deep neural networks have shown remarkable performance across a wide range of vision-based tasks, particularly due to the availability of large-scale datasets for training and better architectures. However, data seen in the real world are…

Machine Learning · Computer Science 2018-11-26 Muhammad Usama , Dong Eui Chang

Catastrophic forgetting affects the training of neural networks, limiting their ability to learn multiple tasks sequentially. From the perspective of the well established plasticity-stability dilemma, neural networks tend to be overly…

Machine Learning · Computer Science 2020-06-15 Seyed Iman Mirzadeh , Mehrdad Farajtabar , Razvan Pascanu , Hassan Ghasemzadeh

To understand the dynamics of optimization in deep neural networks, we develop a tool to study the evolution of the entire Hessian spectrum throughout the optimization process. Using this, we study a number of hypotheses concerning…

Machine Learning · Computer Science 2019-01-30 Behrooz Ghorbani , Shankar Krishnan , Ying Xiao

We study the recently introduced stability training as a general-purpose method to increase the robustness of deep neural networks against input perturbations. In particular, we explore its use as an alternative to data augmentation and…

Machine Learning · Computer Science 2019-11-14 Jan Laermann , Wojciech Samek , Nils Strodthoff

Using backward error analysis, we compute implicit training biases in multitask and continual learning settings for neural networks trained with stochastic gradient descent. In particular, we derive modified losses that are implicitly…

Machine Learning · Statistics 2023-11-02 Benoit Dherin

Dropout is a widely utilized regularization technique in the training of neural networks, nevertheless, its underlying mechanism and its impact on achieving good generalization abilities remain poorly understood. In this work, we derive the…

Machine Learning · Computer Science 2023-05-26 Zhongwang Zhang , Yuqing Li , Tao Luo , Zhi-Qin John Xu

Recently many first and second order variants of SGD have been proposed to facilitate training of Deep Neural Networks (DNNs). A common limitation of these works stem from the fact that they use the same learning rate across all instances…

Machine Learning · Computer Science 2021-05-31 Shreyas Saxena , Nidhi Vyas , Dennis DeCoste