Dropout Mixture Low-Rank Adaptation for Visual Parameters-Efficient Fine-Tuning

  • Zhengyi Fang
  • , Yue Wang
  • , Ran Yi*
  • , Lizhuang Ma
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Parameter-efficient fine-tuning methods adjust a small subset of parameters in large models, achieving performance comparable to or even surpassing that of models fine-tuned with the full parameter set, and significantly reducing the time and computational costs associated with the fine-tuning process. Despite the developments of parameter-efficient fine-tuning methods for large models, we observe significant performance disparities across different vision tasks. We attribute this pronounced performance variability to the insufficient robustness of current parameter-efficient fine-tuning methods. In this paper, we propose a robust reparameterization framework for parameter-efficient fine-tuning. This framework has a dynamic training structure and introduces no additional computational overhead during the inference stage. Specifically, we propose Dropout-Mixture Low-Rank Adaptation (DMLoRA), yrrwhich incorporates multiple up and down branches, to provide the model with a more robust gradient descent path. As training proceeds, DMLoRA gradually drops out branches to achieve a balance between accuracy and regularization. Additionally, we employ a 2-Stage Learning Scalar (LS) strategy to optimize the scale factor for each layer’s DMLoRA module. Experimental results demonstrate that our method achieves state-of-the-art performance on the benchmark VTAB-1k and FGVC datasets for parameter-efficient fine-tuning.

Original languageEnglish
Title of host publicationComputer Vision – ECCV 2024 - 18th European Conference, Proceedings
EditorsAleš Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, Gül Varol
PublisherSpringer Science and Business Media Deutschland GmbH
Pages369-386
Number of pages18
ISBN (Print)9783031726668
DOIs
StatePublished - 2025
Externally publishedYes
Event18th European Conference on Computer Vision, ECCV 2024 - Milan, Italy
Duration: 29 Sep 20244 Oct 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume15065 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference18th European Conference on Computer Vision, ECCV 2024
Country/TerritoryItaly
CityMilan
Period29/09/244/10/24

Keywords

  • 2-Stage Learning Scalar
  • Dropout-Mixture Low-Rank Adaptation
  • Gradual Branch Dropout
  • Paramter-Efficient Fine-Tuning

Fingerprint

Dive into the research topics of 'Dropout Mixture Low-Rank Adaptation for Visual Parameters-Efficient Fine-Tuning'. Together they form a unique fingerprint.

Cite this