Abstract
Parameter-efficient transfer learning (PETL) methods adapt large artificialneural networks to downstream tasks without fine-tuning the entire model.However, existing additive methods, such as adapters, sometimes struggle tocapture distributional shifts in intermediate feature embeddings. We propose anovel histogram-based parameter-efficient tuning (HPT) technique that capturesthe statistics of the target domain and modulates the embeddings. Experimentalresults on three downstream passive sonar datasets (ShipsEar, DeepShip, VTUAD)demonstrate that HPT outperforms conventional adapters. Notably, HPT achieves91.8% vs. 89.8% accuracy on VTUAD. Furthermore, HPT trains faster and yieldsfeature representations closer to those of fully fine-tuned models. Overall,HPT balances parameter savings and performance, providing a distribution-awarealternative to existing adapters and shows a promising direction for scalabletransfer learning in resource-constrained environments. The code is publiclyavailable:https://github.com/Advanced-Vision-and-Learning-Lab/HLAST_DeepShip_ParameterEfficient.