Guardado en:
Detalles Bibliográficos
Autores principales: Chooi, Jay, Cong, Kevin, Li, Russell, Sun, Lillian
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2511.07843
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909897740779520
author Chooi, Jay
Cong, Kevin
Li, Russell
Sun, Lillian
author_facet Chooi, Jay
Cong, Kevin
Li, Russell
Sun, Lillian
contents As deep learning methods increasingly utilize sensitive data on a widespread scale, differential privacy (DP) offers formal guarantees to protect against information leakage during model training. A significant challenge remains in implementing DP optimizers that retain strong performance while preserving privacy. Recent advances introduced ever more efficient optimizers, with AdamW being a popular choice for training deep learning models because of strong empirical performance. We study \emph{DP-AdamW} and introduce \emph{DP-AdamW-BC}, a differentially private variant of the AdamW optimizer with DP bias correction for the second moment estimator. We start by showing theoretical results for privacy and convergence guarantees of DP-AdamW and DP-AdamW-BC. Then, we empirically analyze the behavior of both optimizers across multiple privacy budgets ($ε= 1, 3, 7$). We find that DP-AdamW outperforms existing state-of-the-art differentially private optimizers like DP-SGD, DP-Adam, and DP-AdamBC, scoring over 15\% higher on text classification, up to 5\% higher on image classification, and consistently 1\% higher on graph node classification. Moreover, we empirically show that incorporating bias correction in DP-AdamW (DP-AdamW-BC) consistently decreases accuracy, in contrast to the improvement of DP-AdamBC improvement over DP-Adam.
format Preprint
id arxiv_https___arxiv_org_abs_2511_07843
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DP-AdamW: Investigating Decoupled Weight Decay and Bias Correction in Private Deep Learning
Chooi, Jay
Cong, Kevin
Li, Russell
Sun, Lillian
Machine Learning
As deep learning methods increasingly utilize sensitive data on a widespread scale, differential privacy (DP) offers formal guarantees to protect against information leakage during model training. A significant challenge remains in implementing DP optimizers that retain strong performance while preserving privacy. Recent advances introduced ever more efficient optimizers, with AdamW being a popular choice for training deep learning models because of strong empirical performance. We study \emph{DP-AdamW} and introduce \emph{DP-AdamW-BC}, a differentially private variant of the AdamW optimizer with DP bias correction for the second moment estimator. We start by showing theoretical results for privacy and convergence guarantees of DP-AdamW and DP-AdamW-BC. Then, we empirically analyze the behavior of both optimizers across multiple privacy budgets ($ε= 1, 3, 7$). We find that DP-AdamW outperforms existing state-of-the-art differentially private optimizers like DP-SGD, DP-Adam, and DP-AdamBC, scoring over 15\% higher on text classification, up to 5\% higher on image classification, and consistently 1\% higher on graph node classification. Moreover, we empirically show that incorporating bias correction in DP-AdamW (DP-AdamW-BC) consistently decreases accuracy, in contrast to the improvement of DP-AdamBC improvement over DP-Adam.
title DP-AdamW: Investigating Decoupled Weight Decay and Bias Correction in Private Deep Learning
topic Machine Learning
url https://arxiv.org/abs/2511.07843