Saved in:
Bibliographic Details
Main Authors: Shaked, Tom, Goldman, Yuval, Shayer, Oran
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2409.07582
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916390455214080
author Shaked, Tom
Goldman, Yuval
Shayer, Oran
author_facet Shaked, Tom
Goldman, Yuval
Shayer, Oran
contents Foundational models, trained on vast and diverse datasets, have demonstrated remarkable capabilities in generalizing across different domains and distributions for various zero-shot tasks. Our work addresses the challenge of retaining these powerful generalization capabilities when adapting foundational models to specific downstream tasks through fine-tuning. To this end, we introduce a novel approach we call "similarity loss", which can be incorporated into the fine-tuning process of any task. By minimizing the distortion of fine-tuned embeddings from the pre-trained embeddings, our method strikes a balance between task-specific adaptation and preserving broad generalization abilities. We evaluate our approach on two diverse tasks: image classification on satellite imagery and face recognition, focusing on open-class and domain shift scenarios to assess out-of-distribution (OOD) performance. We demonstrate that this approach significantly improves OOD performance while maintaining strong in-distribution (ID) performance.
format Preprint
id arxiv_https___arxiv_org_abs_2409_07582
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Minimizing Embedding Distortion for Robust Out-of-Distribution Performance
Shaked, Tom
Goldman, Yuval
Shayer, Oran
Computer Vision and Pattern Recognition
Foundational models, trained on vast and diverse datasets, have demonstrated remarkable capabilities in generalizing across different domains and distributions for various zero-shot tasks. Our work addresses the challenge of retaining these powerful generalization capabilities when adapting foundational models to specific downstream tasks through fine-tuning. To this end, we introduce a novel approach we call "similarity loss", which can be incorporated into the fine-tuning process of any task. By minimizing the distortion of fine-tuned embeddings from the pre-trained embeddings, our method strikes a balance between task-specific adaptation and preserving broad generalization abilities. We evaluate our approach on two diverse tasks: image classification on satellite imagery and face recognition, focusing on open-class and domain shift scenarios to assess out-of-distribution (OOD) performance. We demonstrate that this approach significantly improves OOD performance while maintaining strong in-distribution (ID) performance.
title Minimizing Embedding Distortion for Robust Out-of-Distribution Performance
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2409.07582