Raphaël Lafargue Thesis defense

Address:

Campus de Brest - Salle Archipel

Raphaël Lafargue from the Mathematical and Electronic Engineering (MEE) Deprtment and from the Lab-STICC laboratory, will present his research about : 

"Data-Centric Approaches for Few-shot Learning"

 

Thesis defense notice

See more

This thesis presents three key contributions aimed at advancing Few-Shot Learning (FSL) through improved model robustness, accurate performance assessment, and task-specific adaptation. First, we explore methods for building robust and universal feature extractors by incorporating data augmentation and self-supervised loss during pre-training, achieving state-of-the-art performance in in-domain classification tasks. Next, we address the need for reliable evaluations of FSL methods by emphasizing confidence intervals and revealing that widely used evaluation approaches often overlook data randomness, leading to conclusions that may be dataset-specific. We propose evaluation techniques that account for this randomness, demonstrating that claims of superiority between methods can change under these considerations. Lastly, we introduce a data-centric approach that enhances cross-domain task adaptation by selectively forgetting portions of the pre-training dataset, reallocating feature space to improve generalization. Together, these contributions provide comprehensive insights for developing robust, adaptable FSL models

Organizer(s)

As part of the IMT Atlantique's thesis co-accreditation within the SPIN doctoral school

 

Keywords : Few-shot Learning, Deep Learning, Computer Vision, Confidence Intervals

Published on 20.11.2024
Next events
02.12.2025 > 02.16.2025
MEXIQUE
06.16.2025 > 06.19.2025