Delving into the Training Dynamics for Image Classification

In recent years, there has been an increase in exploring and applying the training dynamics (TD) of deep neural networks (DNNs). Current studies typically rely on quite limited TD quantities and apply their sequences to understand or aid training. This study investigates how to create more effective...

Description complète

Détails bibliographiques
Publié dans:IEEE transactions on image processing : a publication of the IEEE Signal Processing Society. - 1992. - PP(2025) vom: 13. Okt.
Auteur principal: Li, Mengyang (Auteur)
Autres auteurs: Zhou, Xiaoling, Wu, Ou
Format: Article en ligne
Langue:English
Publié: 2025
Accès à la collection:IEEE transactions on image processing : a publication of the IEEE Signal Processing Society
Sujets:Journal Article
LEADER 01000naa a22002652c 4500
001 NLM393991040
003 DE-627
005 20251015232840.0
007 cr uuu---uuuuu
008 251015s2025 xx |||||o 00| ||eng c
024 7 |a 10.1109/TIP.2025.3618395  |2 doi 
028 5 2 |a pubmed25n1599.xml 
035 |a (DE-627)NLM393991040 
035 |a (NLM)41082428 
040 |a DE-627  |b ger  |c DE-627  |e rakwb 
041 |a eng 
100 1 |a Li, Mengyang  |e verfasserin  |4 aut 
245 1 0 |a Delving into the Training Dynamics for Image Classification 
264 1 |c 2025 
336 |a Text  |b txt  |2 rdacontent 
337 |a ƒaComputermedien  |b c  |2 rdamedia 
338 |a ƒa Online-Ressource  |b cr  |2 rdacarrier 
500 |a Date Revised 13.10.2025 
500 |a published: Print-Electronic 
500 |a Citation Status Publisher 
520 |a In recent years, there has been an increase in exploring and applying the training dynamics (TD) of deep neural networks (DNNs). Current studies typically rely on quite limited TD quantities and apply their sequences to understand or aid training. This study investigates how to create more effective TD representations, and then apply them to improve the training process of real learning tasks. Specifically, first, an epoch-wise vector comprising 142-dimensional TD quantities, such as loss, is extracted for each sample. Second, a new learning strategy with both self-supervised and supervised learning is designed to learn the deep TD representation of each sample on 200 typical image classification tasks. Third, two novel methods for noisy label detection and imbalance learning, respectively, are presented based on deep TD representations. Our study reveals that neighborhoods and logits are the most important TD quantities, unlike the traditional research that focuses on loss and margin. Moreover, our method based on deep TD representations achieves better performance and demonstrates that high-level TD quantities can facilitate understanding model training, leading to improvements in practical learning tasks, such as noisy label detection and imbalance learning. All the codes are available at https://github.com/limengyang1992/TD_Exploring 
650 4 |a Journal Article 
700 1 |a Zhou, Xiaoling  |e verfasserin  |4 aut 
700 1 |a Wu, Ou  |e verfasserin  |4 aut 
773 0 8 |i Enthalten in  |t IEEE transactions on image processing : a publication of the IEEE Signal Processing Society  |d 1992  |g PP(2025) vom: 13. Okt.  |w (DE-627)NLM09821456X  |x 1941-0042  |7 nnas 
773 1 8 |g volume:PP  |g year:2025  |g day:13  |g month:10 
856 4 0 |u http://dx.doi.org/10.1109/TIP.2025.3618395  |3 Volltext 
912 |a GBV_USEFLAG_A 
912 |a SYSFLAG_A 
912 |a GBV_NLM 
912 |a GBV_ILN_350 
951 |a AR 
952 |d PP  |j 2025  |b 13  |c 10