Adan Adaptive Nesterov Momentum Algorithm for Faster Optimizing Deep Models

Adan : Adaptive Nesterov Momentum Algorithm for Faster Optimizing Deep Models

In deep learning, different kinds of deep networks typically need different optimizers, which have to be chosen after multiple trials, making the training process inefficient. To relieve this issue and consistently improve the model training speed across deep networks, we propose the ADAptive Nester...

Description complète

Détails bibliographiques
Publié dans:IEEE transactions on pattern analysis and machine intelligence. - 1979. - 46(2024), 12 vom: 04. Dez., Seite 9508-9520
Auteur principal: Xie, Xingyu (Auteur)
Autres auteurs: Zhou, Pan, Li, Huan, Lin, Zhouchen, Yan, Shuicheng
Format: Article en ligne
Langue:English
Publié: 2024
Accès à la collection:IEEE transactions on pattern analysis and machine intelligence
Sujets:Journal Article
LEADER 01000caa a22002652c 4500
001 NLM374515654
003 DE-627
005 20250306091743.0
007 cr uuu---uuuuu
008 240705s2024 xx |||||o 00| ||eng c
024 7 |a 10.1109/TPAMI.2024.3423382  |2 doi 
028 5 2 |a pubmed25n1247.xml 
035 |a (DE-627)NLM374515654 
035 |a (NLM)38963744 
040 |a DE-627  |b ger  |c DE-627  |e rakwb 
041 |a eng 
100 1 |a Xie, Xingyu  |e verfasserin  |4 aut 
245 1 0 |a Adan  |b Adaptive Nesterov Momentum Algorithm for Faster Optimizing Deep Models 
264 1 |c 2024 
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 08.11.2024 
500 |a published: Print-Electronic 
500 |a Citation Status PubMed-not-MEDLINE 
520 |a In deep learning, different kinds of deep networks typically need different optimizers, which have to be chosen after multiple trials, making the training process inefficient. To relieve this issue and consistently improve the model training speed across deep networks, we propose the ADAptive Nesterov momentum algorithm, Adan for short. Adan first reformulates the vanilla Nesterov acceleration to develop a new Nesterov momentum estimation (NME) method, which avoids the extra overhead of computing gradient at the extrapolation point. Then Adan adopts NME to estimate the gradient's first- and second-order moments in adaptive gradient algorithms for convergence acceleration. Besides, we prove that Adan finds an ϵ-approximate first-order stationary point within O(ϵ-3.5) stochastic gradient complexity on the non-convex stochastic problems (e.g., deep learning problems), matching the best-known lower bound. Extensive experimental results show that Adan consistently surpasses the corresponding SoTA optimizers on vision, language, and RL tasks and sets new SoTAs for many popular networks and frameworks, e.g., ResNet, ConvNext, ViT, Swin, MAE, DETR, GPT-2, Transformer-XL, and BERT. More surprisingly, Adan can use half of the training cost (epochs) of SoTA optimizers to achieve higher or comparable performance on ViT, GPT-2, MAE, etc, and also shows great tolerance to a large range of minibatch size, e.g., from 1 k to 32 k 
650 4 |a Journal Article 
700 1 |a Zhou, Pan  |e verfasserin  |4 aut 
700 1 |a Li, Huan  |e verfasserin  |4 aut 
700 1 |a Lin, Zhouchen  |e verfasserin  |4 aut 
700 1 |a Yan, Shuicheng  |e verfasserin  |4 aut 
773 0 8 |i Enthalten in  |t IEEE transactions on pattern analysis and machine intelligence  |d 1979  |g 46(2024), 12 vom: 04. Dez., Seite 9508-9520  |w (DE-627)NLM098212257  |x 1939-3539  |7 nnas 
773 1 8 |g volume:46  |g year:2024  |g number:12  |g day:04  |g month:12  |g pages:9508-9520 
856 4 0 |u http://dx.doi.org/10.1109/TPAMI.2024.3423382  |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 46  |j 2024  |e 12  |b 04  |c 12  |h 9508-9520