English
Related papers

Related papers: Saddle-to-Saddle Dynamics in Diagonal Linear Netwo…

200 papers

Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training…

Optimization and Control · Mathematics 2020-06-23 Lenaic Chizat , Francis Bach

In this paper, we unroll the dynamics of the dual ascent (DA) algorithm in two coupled graph neural networks (GNNs) to solve constrained optimization problems. The two networks interact with each other at the layer level to find a saddle…

Machine Learning · Computer Science 2026-02-03 Samar Hadou , Alejandro Ribeiro

Diagonal linear networks (DLNs) are a tractable model that captures several nontrivial behaviors in neural network training, such as initialization-dependent solutions and incremental learning. These phenomena are typically studied in…

Machine Learning · Statistics 2026-03-16 Sota Nishiyama , Masaaki Imaizumi

A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks have made a theory of learning dynamics elusive. In this work, we…

Continual learning, the ability of a model to adapt to an ongoing sequence of tasks without forgetting earlier ones, is a central goal of artificial intelligence. To better understand its underlying mechanisms, we study the limitations of…

Machine Learning · Statistics 2026-04-21 Hossein Taheri , Avishek Ghosh , Arya Mazumdar

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

We study the training dynamics of a shallow neural network with quadratic activation functions and quadratic cost in a teacher-student setup. In line with previous works on the same neural architecture, the optimization is performed…

Optimization and Control · Mathematics 2023-11-08 Simon Martin , Francis Bach , Giulio Biroli

Deep neural networks are usually trained with stochastic gradient descent (SGD), which minimizes objective function using very rough approximations of gradient, only averaging to the real gradient. Standard approaches like momentum or ADAM…

Machine Learning · Computer Science 2023-03-14 Jarek Duda

The process of training an artificial neural network involves iteratively adapting its parameters so as to minimize the error of the network's prediction, when confronted with a learning task. This iterative change can be naturally…

Machine Learning · Computer Science 2024-04-10 Kaloyan Danovski , Miguel C. Soriano , Lucas Lacasa

We study the gradient descent (GD) dynamics of a depth-2 linear neural network with a single input and output. We show that GD converges at an explicit linear rate to a global minimum of the training loss, even with a large stepsize --…

Machine Learning · Computer Science 2025-01-22 Pierfrancesco Beneventano , Blake Woodworth

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

Machine learning problems such as neural network training, tensor decomposition, and matrix factorization, require local minimization of a nonconvex function. This local minimization is challenged by the presence of saddle points, of which…

Optimization and Control · Mathematics 2018-07-23 Santiago Paternain , Aryan Mokhtari , Alejandro Ribeiro

Adversarial neural networks solve many important problems in data science, but are notoriously difficult to train. These difficulties come from the fact that optimal weights for adversarial nets correspond to saddle points, and not…

Machine Learning · Computer Science 2018-02-12 Abhay Yadav , Sohil Shah , Zheng Xu , David Jacobs , Tom Goldstein

Recent focus on robustness to adversarial attacks for deep neural networks produced a large variety of algorithms for training robust models. Most of the effective algorithms involve solving the min-max optimization problem for training…

Machine Learning · Computer Science 2021-03-03 Yasaman Esfandiari , Aditya Balu , Keivan Ebrahimi , Umesh Vaidya , Nicola Elia , Soumik Sarkar

Residual networks have shown great success and become indispensable in recent deep neural network models. In this work, we aim to re-investigate the training process of residual networks from a novel social psychology perspective of…

Machine Learning · Computer Science 2023-05-05 Peng Ye , Tong He , Shengji Tang , Baopu Li , Tao Chen , Lei Bai , Wanli Ouyang

Satisfaction of the strict saddle property has become a standard assumption in non-convex optimization, and it ensures that many first-order optimization algorithms will almost always escape saddle points. However, functions exist in…

Optimization and Control · Mathematics 2022-08-23 Matthew Ubl , Kasra Yazdani , Matthew T. Hale

Real-world datasets exhibit imbalances of varying types and degrees. Several techniques based on re-weighting and margin adjustment of loss are often used to enhance the performance of neural networks, particularly on minority classes. In…

Machine Learning · Computer Science 2022-12-29 Harsh Rangwani , Sumukh K Aithal , Mayank Mishra , R. Venkatesh Babu

We reconsider the stochastic (sub)gradient approach to the unconstrained primal L1-SVM optimization. We observe that if the learning rate is inversely proportional to the number of steps, i.e., the number of times any training pattern is…

Machine Learning · Computer Science 2014-01-28 Constantinos Panagiotakopoulos , Petroula Tsampouka

Characterizing and understanding the dynamics of stochastic gradient descent (SGD) around saddle points remains an open problem. We first show that saddle points in neural networks can be divided into two types, among which the Type-II…

Machine Learning · Computer Science 2024-07-03 Liu Ziyin , Botao Li , Tomer Galanti , Masahito Ueda

The loss landscapes of neural networks contain minima and saddle points that may be connected in flat regions or appear in isolation. We identify and characterize a special structure in the loss landscape: channels along which the loss…

Machine Learning · Computer Science 2026-05-11 Flavio Martinelli , Alexander Van Meegen , Berfin Şimşek , Wulfram Gerstner , Johanni Brea