English
Related papers

Related papers: Do Deep Networks Forget Initialization? A Forgetti…

200 papers

Transformer-based Large Language Models (LLMs) have revolutionized Natural Language Processing by demonstrating exceptional performance across diverse tasks. This study investigates the impact of the parameter initialization scale on the…

Computation and Language · Computer Science 2025-05-22 Junjie Yao , Zhongwang Zhang , Zhi-Qin John Xu

We identify and study two common failure modes for early training in deep ReLU nets. For each we give a rigorous proof of when it occurs and how to avoid it, for fully connected and residual architectures. The first failure mode,…

Machine Learning · Statistics 2018-11-14 Boris Hanin , David Rolnick

In matrix sensing, we first numerically identify the sensitivity to the initialization rank as a new limitation of the implicit bias of gradient flow. We will partially quantify this phenomenon mathematically, where we establish that the…

Information Theory · Computer Science 2021-06-08 Armin Eftekhari , Konstantinos Zygalakis

This paper revisits the so-called vanishing gradient phenomenon, which commonly occurs in deep randomly initialized neural networks. Leveraging an in-depth analysis of neural chains, we first show that vanishing gradients cannot be…

Machine Learning · Computer Science 2021-06-08 Antonio Orvieto , Jonas Kohler , Dario Pavllo , Thomas Hofmann , Aurelien Lucchi

Transformers have shown impressive capabilities across various tasks, but their performance on compositional problems remains a topic of debate. In this work, we investigate the mechanisms of how transformers behave on unseen compositional…

Machine Learning · Computer Science 2025-01-14 Zhongwang Zhang , Pengxiao Lin , Zhiwei Wang , Yaoyu Zhang , Zhi-Qin John Xu

Initialization of parameters in deep neural networks has been shown to have a big impact on the performance of the networks (Mishkin & Matas, 2015). The initialization scheme devised by He et al, allowed convolution activations to carry a…

Machine Learning · Computer Science 2017-02-28 Armen Aghajanyan

Transformers underpin modern large language models (LLMs) and are commonly assumed to be behaviorally unstructured at random initialization, with all meaningful preferences emerging only through large-scale training. We challenge this…

Machine Learning · Statistics 2026-02-06 Siquan Li , Yao Tong , Haonan Wang , Tianyang Hu

Innovations in neural architectures have fostered significant breakthroughs in language modeling and computer vision. Unfortunately, novel architectures often result in challenging hyper-parameter choices and training instability if the…

Machine Learning · Computer Science 2021-11-25 Chen Zhu , Renkun Ni , Zheng Xu , Kezhi Kong , W. Ronny Huang , Tom Goldstein

People use rich prior knowledge about the world in order to efficiently learn new concepts. These priors - also known as "inductive biases" - pertain to the space of internal models considered by a learner, and they help the learner make…

Computation and Language · Computer Science 2018-06-20 Reuben Feinman , Brenden M. Lake

Proper weight initialization prior to training has historically been one of the key factors that helped kick off the deep learning revolution. Initialization is even more crucial in "reservoir computing", where the weights of a readout…

Machine Learning · Computer Science 2026-05-12 Tommaso Fioratti , Riccardo Marcaccioli , Francesco Casola

The approximation and convergence properties of implicit neural representations (INRs) are known to be highly sensitive to parameter initialization strategies. While several data-driven initialization methods demonstrate significant…

Computer Vision and Pattern Recognition · Computer Science 2026-04-01 Kushal Vyas , Alper Kayabasi , Daniel Kim , Vishwanath Saragadam , Ashok Veeraraghavan , Guha Balakrishnan

Continual (or "incremental") learning approaches are employed when additional knowledge or tasks need to be learned from subsequent batches or from streaming data. However these approaches are typically adversary agnostic, i.e., they do not…

Machine Learning · Computer Science 2021-02-17 Muhammad Umer , Robi Polikar

Modern deep learning science often assumes that neural networks learn from a fixed data distribution. However, many practically important learning problems involve data distributions that change throughout training. How does such…

Machine Learning · Computer Science 2026-05-19 Afiq Abdillah Effiezal Aswadi , Oliver Britton , Ross Baker , Matthew Farrugia-Roberts

The utilization of residual learning has become widespread in deep and scalable neural nets. However, the fundamental principles that contribute to the success of residual learning remain elusive, thus hindering effective training of plain…

Computer Vision and Pattern Recognition · Computer Science 2024-02-14 Tunhou Zhang , Feng Yan , Hai Li , Yiran Chen

Selective forgetting or removing information from deep neural networks (DNNs) is essential for continual learning and is challenging in controlling the DNNs. Such forgetting is crucial also in a practical sense since the deployed DNNs may…

Machine Learning · Statistics 2021-01-01 Tomohiro Hayase , Suguru Yasutomi , Takashi Katoh

Despite their massive success, training successful deep neural networks still largely relies on experimentally choosing an architecture, hyper-parameters, initialization, and training mechanism. In this work, we focus on determining the…

Machine Learning · Computer Science 2021-11-29 Umangi Jain , Harish G. Ramaswamy

A leading hypothesis for the surprising generalization of neural networks is that the dynamics of gradient descent bias the model towards simple solutions, by searching through the solution space in an incremental order of complexity. We…

Machine Learning · Computer Science 2020-01-01 Daniel Gissin , Shai Shalev-Shwartz , Amit Daniely

We study the interplay between memorization and generalization of overparameterized networks in the extreme case of a single training example and an identity-mapping task. We examine fully-connected and convolutional networks (FCN and CNN),…

Machine Learning · Statistics 2020-01-10 Chiyuan Zhang , Samy Bengio , Moritz Hardt , Michael C. Mozer , Yoram Singer

Neural networks trained with stochastic gradient descent (SGD) starting from different random initialisations typically find functionally very similar solutions, raising the question of whether there are meaningful differences between…

Sequential learning in physical networks is hindered by catastrophic forgetting, where training a new task erases solutions to earlier ones. We show that we can significantly enhance memory of previous tasks by introducing a hard threshold…

Disordered Systems and Neural Networks · Physics 2025-12-04 Purba Chatterjee , Marcelo Guzman , Andrea J. Liu
‹ Prev 1 4 5 6 7 8 10 Next ›