Back to the Basics on Predicting Transfer Performance

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chaves, Levy, Valle, Eduardo, Bissoto, Alceu, Avila, Sandra
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909213570105344
author Chaves, Levy
Valle, Eduardo
Bissoto, Alceu
Avila, Sandra
author_facet Chaves, Levy
Valle, Eduardo
Bissoto, Alceu
Avila, Sandra
contents In the evolving landscape of deep learning, selecting the best pre-trained models from a growing number of choices is a challenge. Transferability scorers propose alleviating this scenario, but their recent proliferation, ironically, poses the challenge of their own assessment. In this work, we propose both robust benchmark guidelines for transferability scorers, and a well-founded technique to combine multiple scorers, which we show consistently improves their results. We extensively evaluate 13 scorers from literature across 11 datasets, comprising generalist, fine-grained, and medical imaging datasets. We show that few scorers match the predictive performance of the simple raw metric of models on ImageNet, and that all predictors suffer on medical datasets. Our results highlight the potential of combining different information sources for reliably predicting transferability across varied domains.
format Preprint
id arxiv_https___arxiv_org_abs_2405_20420
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Back to the Basics on Predicting Transfer Performance
Chaves, Levy
Valle, Eduardo
Bissoto, Alceu
Avila, Sandra
Machine Learning
Computer Vision and Pattern Recognition
In the evolving landscape of deep learning, selecting the best pre-trained models from a growing number of choices is a challenge. Transferability scorers propose alleviating this scenario, but their recent proliferation, ironically, poses the challenge of their own assessment. In this work, we propose both robust benchmark guidelines for transferability scorers, and a well-founded technique to combine multiple scorers, which we show consistently improves their results. We extensively evaluate 13 scorers from literature across 11 datasets, comprising generalist, fine-grained, and medical imaging datasets. We show that few scorers match the predictive performance of the simple raw metric of models on ImageNet, and that all predictors suffer on medical datasets. Our results highlight the potential of combining different information sources for reliably predicting transferability across varied domains.
title Back to the Basics on Predicting Transfer Performance
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.20420