Loshchilov and Hutter: decoupled weight decay for adaptive optimizers

What changed

Separated weight shrinkage from the adaptive gradient transformation in AdamW, clarifying its difference from adding an L2 penalty and reporting improved image-classification generalization.

Scientific scope

Separated weight shrinkage from the adaptive gradient transformation in AdamW, clarifying its difference from adding an L2 penalty and reporting improved image-classification generalization. Introduced as a 2017 preprint, with the cited revision published at ICLR 2019. Weight decay predates this work. Benefits are evaluated settings, not guaranteed superiority or complete hyperparameter independence.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Introduced as a 2017 preprint, with the cited revision published at ICLR 2019. Weight decay predates this work. Benefits are evaluated settings, not guaranteed superiority or complete hyperparameter independence.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Decoupled Weight Decay Regularization
• Loshchilov and Hutter: decoupled weight decay for adaptive optimizers — author/proceedings record

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction