Domain-Agnostic Feature Modulation for Semi-Supervised Domain Generalization

Venuri Amarasinghe, Kalinga Bandara, Isun Randila, Asini Jayakody, Chamuditha Jayanga Galappaththige, Ranga Rodrigo · 2025

Read Domain-Agnostic Feature Modulation for Semi-Supervised Domain Generalization by Venuri Amarasinghe, Kalinga Bandara, Isun Randila, Asini Jayakody, Chamuditha Jayanga Galappaththige, Ranga Rodrigo (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-20897.txt.

Description

Venuri Amarasinghe and co-authors tackle learning that must generalize to new domains with few labels and no domain tags, in a 2025 paper. Semi-supervised domain generalization uses a small labeled set plus unlabeled data. Most methods pseudo-label the unlabeled data and assume domain labels are known.

Domain shifts add noise, making pseudo-labels inconsistent and hurting performance. The authors propose a feature modulation strategy that boosts class-relevant features and suppresses domain-specific ones. Features shift toward averaged class representations. The method targets the harder setting where domain labels for unlabeled data are absent during training, a common real-world case.

Catalog record

CREATORVenuri Amarasinghe, Kalinga Bandara, Isun Randila, Asini Jayakody, Chamuditha Jayanga Galappaththige, Ranga Rodrigo
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-20897

Browse the full library · About the Tower of Babel