Computer Science > Machine Learning
[Submitted on 2 Apr 2025]
Title:Enlightenment Period Improving DNN Performance
View PDF HTML (experimental)Abstract:In the early stage of deep neural network training, the loss decreases rapidly before gradually leveling off. Extensive research has shown that during this stage, the model parameters undergo significant changes and their distribution is largely established. Existing studies suggest that the introduction of noise during early training can degrade model performance. We identify a critical "enlightenment period" encompassing up to the first 4% of the training cycle (1--20 epochs for 500-epoch training schedules), a phase characterized by intense parameter fluctuations and heightened noise sensitivity. Our findings reveal that strategically reducing noise during this brief phase--by disabling data augmentation techniques such as Mixup or removing high-loss samples--leads to statistically significant improvements in model performance. This work opens new avenues for exploring the relationship between the enlightenment period and network training dynamics across diverse model architectures and tasks.
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
IArxiv Recommender
(What is IArxiv?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.