Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Xu, Zhi-Qin John, Zhang, Yaoyu, Zhou, Zhangchen
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2504.09484
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918439827800064
author Xu, Zhi-Qin John
Zhang, Yaoyu
Zhou, Zhangchen
author_facet Xu, Zhi-Qin John
Zhang, Yaoyu
Zhou, Zhangchen
contents In this paper, we provide an overview of a common phenomenon, condensation, observed during the nonlinear training of neural networks: During the nonlinear training of neural networks, neurons in the same layer tend to condense into groups with similar outputs. Empirical observations suggest that the number of condensed clusters of neurons in the same layer typically increases monotonically as training progresses. Neural networks with small weight initializations or Dropout optimization can facilitate this condensation process. We also examine the underlying mechanisms of condensation from the perspectives of training dynamics and the structure of the loss landscape. The condensation phenomenon offers valuable insights into the generalization abilities of neural networks and correlates to stronger reasoning abilities in transformer-based language models.
format Preprint
id arxiv_https___arxiv_org_abs_2504_09484
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle An overview of condensation phenomenon in deep learning
Xu, Zhi-Qin John
Zhang, Yaoyu
Zhou, Zhangchen
Machine Learning
In this paper, we provide an overview of a common phenomenon, condensation, observed during the nonlinear training of neural networks: During the nonlinear training of neural networks, neurons in the same layer tend to condense into groups with similar outputs. Empirical observations suggest that the number of condensed clusters of neurons in the same layer typically increases monotonically as training progresses. Neural networks with small weight initializations or Dropout optimization can facilitate this condensation process. We also examine the underlying mechanisms of condensation from the perspectives of training dynamics and the structure of the loss landscape. The condensation phenomenon offers valuable insights into the generalization abilities of neural networks and correlates to stronger reasoning abilities in transformer-based language models.
title An overview of condensation phenomenon in deep learning
topic Machine Learning
url https://arxiv.org/abs/2504.09484