Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification
Deep learning has significantly advanced image analysis across diverse domains but often depends on large, annotated datasets for success. Transfer learning addresses this challenge by utilizing pre-trained models to tackle new tasks with limited labeled data. However, discrepancies between source a...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11075778/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849254668093358080 |
|---|---|
| author | Ana Davila Jacinto Colan Yasuhisa Hasegawa |
| author_facet | Ana Davila Jacinto Colan Yasuhisa Hasegawa |
| author_sort | Ana Davila |
| collection | DOAJ |
| description | Deep learning has significantly advanced image analysis across diverse domains but often depends on large, annotated datasets for success. Transfer learning addresses this challenge by utilizing pre-trained models to tackle new tasks with limited labeled data. However, discrepancies between source and target domains can hinder effective transfer learning. We introduce BioTune, a novel adaptive fine-tuning technique utilizing evolutionary optimization. BioTune enhances transfer learning by optimally choosing which layers to freeze and adjusting learning rates for unfrozen layers. Through extensive evaluation on nine image classification datasets, spanning natural and specialized domains such as medical imaging, BioTune demonstrates superior accuracy and efficiency over state-of-the-art fine-tuning methods, including AutoRGN and LoRA, highlighting its adaptability to various data characteristics and distribution changes. Additionally, BioTune consistently achieves top performance across four different CNN architectures, underscoring its flexibility. Ablation studies provide valuable insights into the impact of BioTune’s key components on overall performance. |
| format | Article |
| id | doaj-art-a730b127e080463aab72074f3a5b7d64 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-a730b127e080463aab72074f3a5b7d642025-08-20T03:56:04ZengIEEEIEEE Access2169-35362025-01-011312923412924910.1109/ACCESS.2025.358752411075778Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image ClassificationAna Davila0https://orcid.org/0000-0002-2076-6842Jacinto Colan1https://orcid.org/0000-0002-8833-2215Yasuhisa Hasegawa2https://orcid.org/0000-0001-9917-098XInstitutes of Innovation for Future Society, Nagoya University, Chikusa-ku, Nagoya, JapanDepartment of Micro-Nano Mechanical Science and Engineering, Nagoya University, Chikusa-ku, Nagoya, JapanInstitutes of Innovation for Future Society, Nagoya University, Chikusa-ku, Nagoya, JapanDeep learning has significantly advanced image analysis across diverse domains but often depends on large, annotated datasets for success. Transfer learning addresses this challenge by utilizing pre-trained models to tackle new tasks with limited labeled data. However, discrepancies between source and target domains can hinder effective transfer learning. We introduce BioTune, a novel adaptive fine-tuning technique utilizing evolutionary optimization. BioTune enhances transfer learning by optimally choosing which layers to freeze and adjusting learning rates for unfrozen layers. Through extensive evaluation on nine image classification datasets, spanning natural and specialized domains such as medical imaging, BioTune demonstrates superior accuracy and efficiency over state-of-the-art fine-tuning methods, including AutoRGN and LoRA, highlighting its adaptability to various data characteristics and distribution changes. Additionally, BioTune consistently achieves top performance across four different CNN architectures, underscoring its flexibility. Ablation studies provide valuable insights into the impact of BioTune’s key components on overall performance.https://ieeexplore.ieee.org/document/11075778/Image classificationadaptive transfer learningfine-tuningevolutionary explorationbio-inspired optimizationmedical imaging |
| spellingShingle | Ana Davila Jacinto Colan Yasuhisa Hasegawa Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification IEEE Access Image classification adaptive transfer learning fine-tuning evolutionary exploration bio-inspired optimization medical imaging |
| title | Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification |
| title_full | Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification |
| title_fullStr | Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification |
| title_full_unstemmed | Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification |
| title_short | Bio-Inspired Fine-Tuning for Selective Transfer Learning in Image Classification |
| title_sort | bio inspired fine tuning for selective transfer learning in image classification |
| topic | Image classification adaptive transfer learning fine-tuning evolutionary exploration bio-inspired optimization medical imaging |
| url | https://ieeexplore.ieee.org/document/11075778/ |
| work_keys_str_mv | AT anadavila bioinspiredfinetuningforselectivetransferlearninginimageclassification AT jacintocolan bioinspiredfinetuningforselectivetransferlearninginimageclassification AT yasuhisahasegawa bioinspiredfinetuningforselectivetransferlearninginimageclassification |