@article{ART003349268},
author={Min Taek Lim and Joo Hyun Lee and Jinyong Kim and Yu Rang Park},
title={Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={6},
pages={29-43}
TY - JOUR
AU - Min Taek Lim
AU - Joo Hyun Lee
AU - Jinyong Kim
AU - Yu Rang Park
TI - Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 6
PB - The Korean Society Of Computer And Information
SP - 29
EP - 43
SN - 1598-849X
AB - In Domain-Incremental Learning (DIL), models must adapt to domain shifts while mitigating catastrophic forgetting. While Pre-Trained Models (PTMs) are widely used, existing strategies often bias representations toward seen domains, failing to effectively mitigate domain shift in unseen domains and inducing representation drift. To address this, we propose a dual-encoder framework that learns domain-invariant and class-discriminative representations. A domain-specific encoder extracts prototypes, distilling stable knowledge into a trainable domain-invariant encoder. Prototype alignment and weight regularization prevent drift and preserve past knowledge. Furthermore, Representation squeezing and a Gradient Reversal Layer (GRL) form compact intra-class clusters and explicitly encourage domain-invariant learning. Extensive experiments on four benchmark datasets (Office-Home, DomainNet, CORe50, and PACS) show our framework improves unseen domain average accuracy by 1-6% across datasets. Notably, on Office-Home, it achieves the highest overall performance with 86.32% Last and 86.25% Avg accuracy. It also reaches an 83.17% Unseen Avg and successfully reduces the forgetting measure to 3.83%.
KW - Domain-Incremental Learning;Domain-Invariant;Domain Shift;Unseen Domain;Pre-trained Models
DO -
UR -
ER -
Min Taek Lim, Joo Hyun Lee, Jinyong Kim and Yu Rang Park. (2026). Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning. Journal of The Korea Society of Computer and Information, 31(6), 29-43.
Min Taek Lim, Joo Hyun Lee, Jinyong Kim and Yu Rang Park. 2026, "Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning", Journal of The Korea Society of Computer and Information, vol.31, no.6 pp.29-43.
Min Taek Lim, Joo Hyun Lee, Jinyong Kim, Yu Rang Park "Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning" Journal of The Korea Society of Computer and Information 31.6 pp.29-43 (2026) : 29.
Min Taek Lim, Joo Hyun Lee, Jinyong Kim, Yu Rang Park. Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning. 2026; 31(6), 29-43.
Min Taek Lim, Joo Hyun Lee, Jinyong Kim and Yu Rang Park. "Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning" Journal of The Korea Society of Computer and Information 31, no.6 (2026) : 29-43.
Min Taek Lim; Joo Hyun Lee; Jinyong Kim; Yu Rang Park. Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning. Journal of The Korea Society of Computer and Information, 31(6), 29-43.
Min Taek Lim; Joo Hyun Lee; Jinyong Kim; Yu Rang Park. Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning. Journal of The Korea Society of Computer and Information. 2026; 31(6) 29-43.
Min Taek Lim, Joo Hyun Lee, Jinyong Kim, Yu Rang Park. Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning. 2026; 31(6), 29-43.
Min Taek Lim, Joo Hyun Lee, Jinyong Kim and Yu Rang Park. "Mitigating Domain Shift via Domain-Invariant Representation Learning in Domain-Incremental Learning" Journal of The Korea Society of Computer and Information 31, no.6 (2026) : 29-43.
1