Self-Masking Networks for Unsupervised Adaptation
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912024123932672 |
|---|---|
| author | Warmerdam, Alfonso Taboada Caron, Mathilde Asano, Yuki M. |
| author_facet | Warmerdam, Alfonso Taboada Caron, Mathilde Asano, Yuki M. |
| contents | With the advent of billion-parameter foundation models, efficient fine-tuning has become increasingly important for the adaptation of models to downstream tasks. However, especially in computer vision, it can be hard to achieve good performance when access to quality labeled data is lacking. In this work, we propose a method adapting pretrained generalist models in a self-supervised manner by learning binary masks. These self-supervised masking networks (SMNs) are up to 79x more efficient to store and significantly improve performance on label-efficient downstream tasks. We validate the usefulness of learning binary masks as a fine-tuning method on 8 datasets and 3 model architectures, and we demonstrate the effectiveness of SMNs in 3 label-efficient settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_07577 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Self-Masking Networks for Unsupervised Adaptation Warmerdam, Alfonso Taboada Caron, Mathilde Asano, Yuki M. Computer Vision and Pattern Recognition Machine Learning With the advent of billion-parameter foundation models, efficient fine-tuning has become increasingly important for the adaptation of models to downstream tasks. However, especially in computer vision, it can be hard to achieve good performance when access to quality labeled data is lacking. In this work, we propose a method adapting pretrained generalist models in a self-supervised manner by learning binary masks. These self-supervised masking networks (SMNs) are up to 79x more efficient to store and significantly improve performance on label-efficient downstream tasks. We validate the usefulness of learning binary masks as a fine-tuning method on 8 datasets and 3 model architectures, and we demonstrate the effectiveness of SMNs in 3 label-efficient settings. |
| title | Self-Masking Networks for Unsupervised Adaptation |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2409.07577 |