English
Related papers

Related papers: Loss Spike in Training Neural Networks

200 papers

Supervised training of deep neural nets typically relies on minimizing cross-entropy. However, in many domains, we are interested in performing well on metrics specific to the application. In this paper we propose a direct loss minimization…

Machine Learning · Computer Science 2016-06-03 Yang Song , Alexander G. Schwing , Richard S. Zemel , Raquel Urtasun

When an online learning algorithm is used to estimate the unknown parameters of a model, the signals interacting with the parameter estimates should not decay too quickly for the optimal values to be discovered correctly. This requirement…

Machine Learning · Computer Science 2019-11-05 Kamil Nar , S. Shankar Sastry

Deep neural network training often exhibits highly anisotropic loss geometry, where a few sharp dominant Hessian directions coexist with a large flatter bulk. Gradients tend to align disproportionately with these dominant directions,…

Machine Learning · Computer Science 2026-05-28 Tolga Dimlioglu , Kristi Topollai , Anna Choromanska

Using established principles from Statistics and Information Theory, we show that invariance to nuisance factors in a deep neural network is equivalent to information minimality of the learned representation, and that stacking layers and…

Machine Learning · Computer Science 2018-06-29 Alessandro Achille , Stefano Soatto

We introduce a model for neural scaling laws under sparse activations. In the model, test loss is often dominated by rare coordinates that are never observed in the training input. This mechanism induces a novel bottleneck absent from dense…

Machine Learning · Statistics 2026-05-25 John Sous , Michael Winer

The first part of this paper studies the evolution of gradient flow for homogeneous neural networks near a class of saddle points exhibiting a sparsity structure. The choice of these saddle points is motivated from previous works on…

Machine Learning · Computer Science 2025-09-16 Akshay Kumar , Jarvis Haupt

`Double descent' delineates the generalization behaviour of models depending on the regime they belong to: under- or over-parameterized. The current theoretical understanding behind the occurrence of this phenomenon is primarily based on…

Machine Learning · Statistics 2022-03-15 Sidak Pal Singh , Aurelien Lucchi , Thomas Hofmann , Bernhard Schölkopf

When random label noise is added to a training dataset, the prediction error of a neural network on a label-noise-free test dataset initially improves during early training but eventually deteriorates, following a U-shaped dependence on…

Machine Learning · Computer Science 2023-06-06 Chaoyue Liu , Amirhesam Abedsoltan , Mikhail Belkin

Flat regions of the neural network loss landscape have long been hypothesized to correlate with better generalization properties. A closely related but distinct problem is training models that are robust to internal perturbations to their…

Machine Learning · Computer Science 2026-02-10 Philip Jacobson , Ben Feinberg , Suhas Kumar , Sapan Agarwal , T. Patrick Xiao , Christopher Bennett

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

We study the effect of mini-batching on the loss landscape of deep neural networks using spiked, field-dependent random matrix theory. We demonstrate that the magnitude of the extremal values of the batch Hessian are larger than those of…

Machine Learning · Statistics 2021-11-08 Diego Granziol , Stefan Zohren , Stephen Roberts

Neural networks can be trained to solve regression problems by using gradient-based methods to minimize the square loss. However, practitioners often prefer to reformulate regression as a classification problem, observing that training on…

Machine Learning · Computer Science 2023-03-02 Lawrence Stewart , Francis Bach , Quentin Berthet , Jean-Philippe Vert

Large loss spikes in stochastic gradient descent are studied through a rigorous large-deviations analysis for a shallow, fully connected network in the NTK scaling. In contrast to full-batch gradient descent, the catapult phase is shown to…

Machine Learning · Computer Science 2026-04-14 Benjamin Gess , Daniel Heydecker

In this paper, we study the sharpness of a deep learning (DL) loss landscape around local minima in order to reveal systematic mechanisms underlying the generalization abilities of DL models. Our analysis is performed across varying network…

Machine Learning · Computer Science 2022-02-07 Devansh Bisla , Jing Wang , Anna Choromanska

In this paper, we describe a phenomenon, which we named "super-convergence", where neural networks can be trained an order of magnitude faster than with standard training methods. The existence of super-convergence is relevant to…

Machine Learning · Computer Science 2018-05-18 Leslie N. Smith , Nicholay Topin

Recently, there has been a surge in interest in developing optimization algorithms for overparameterized models as achieving generalization is believed to require algorithms with suitable biases. This interest centers on minimizing…

Machine Learning · Computer Science 2026-02-05 Behrooz Tahmasebi , Ashkan Soleymani , Dara Bahri , Stefanie Jegelka , Patrick Jaillet

We explore the loss landscape of fully-connected and convolutional neural networks using random, low-dimensional hyperplanes and hyperspheres. Evaluating the Hessian, $H$, of the loss function on these hypersurfaces, we observe 1) an…

Machine Learning · Computer Science 2018-11-13 Stanislav Fort , Adam Scherlis

In this paper, we advance the understanding of neural network training dynamics by examining the intricate interplay of various factors introduced by weight parameters in the initialization process. Motivated by the foundational work of Luo…

Machine Learning · Computer Science 2024-04-09 Yuqing Li , Tao Luo , Qixuan Zhou

Explainability of neural network prediction is essential to understand feature importance and gain interpretable insight into neural network performance. However, explanations of neural network outcomes are mostly limited to visualization,…

Machine Learning · Computer Science 2023-07-13 Arnab Neelim Mazumder , Niall Lyons , Ashutosh Pandey , Avik Santra , Tinoosh Mohsenin

The training of over-parameterized neural networks has received much study in recent literature. An important consideration is the regularization of over-parameterized networks due to their highly nonconvex and nonlinear geometry. In this…

Machine Learning · Computer Science 2024-09-24 Hongyang R. Zhang , Dongyue Li , Haotian Ju