Domain Generalization Using Large Pretrained Models With Mixture-of-Adapters
Keywords: domain generalization, distribution shift, fine-tuning, transfer learning
TL;DR: We analyze performance on PEFT methods in domain generalization tasks for vision, and proposes Mixture-of-Adapter method.
Abstract: Learning a robust vision model despite large distribution shift situations is an important task for model deployment in real-world settings. Especially, domain generalization (DG) algorithm aims to maintain the performance of a trained model on different distributions which were not seen during training. One of the most effective method has been leveraging the already learned rich knowledge of large pretrained models. However, naively tuning large models to DG tasks is practically infeasible due to memory limitations, extensive time requirements for training, and the risk of learned knowledge deterioration. Parameter-efficient fine-tuning (PEFT) methods have been used to reduce the high computational cost during training and efficiently adapt large models to downstream tasks. In this work for the first time we find that the use of adapters in PEFT methods not only reduce high computational cost during training but also serve as an effective regularizer for DG tasks. Surprisingly, a naive adapter implementation for large models achieve superior performance on common datasets. However, in situations of large distribution shifts, additional factors such as optimal amount of regularization due to the strength of distribution shifts should be considered for a sophisticated adapter implementation. As a result, we propose a mixture-of-expert based adapter fine-tuning method, dubbed as mixture-of-adapters (MoA). We employ multiple adapters that have varying capacities, and by using learnable routers, we allocate each token to a proper adapter. By using both PEFT and MoA methods we effectively alleviate the performance deterioration caused by distribution shifts and achieve state-of-the-art performance on diverse DG benchmarks.
Primary Area: representation learning for computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5302
Loading