An Empirical Study of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Naganuma, Hiroki, Hataya, Ryuichiro, Yoshida, Kotaro, Mitliagkas, Ioannis
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913808756244480
author Naganuma, Hiroki
Hataya, Ryuichiro
Yoshida, Kotaro
Mitliagkas, Ioannis
author_facet Naganuma, Hiroki
Hataya, Ryuichiro
Yoshida, Kotaro
Mitliagkas, Ioannis
contents In the field of computer vision, fine-tuning pre-trained models has become a prevalent strategy for out-of-distribution (OOD) generalization tasks. Different from most prior work that has focused on advancing learning algorithms, we systematically examined how pre-trained model size, pre-training dataset size, and training strategies impact generalization and confidence calibration on downstream tasks. We evaluated 100 models across diverse pre-trained model sizes, five pre-training datasets, and five data augmentations through extensive experiments on four distribution shift datasets totaling over 120,000 GPU hours. Our results demonstrate the significant impact of pre-trained model selection, with optimal choices substantially improving OOD accuracy over algorithm improvement alone. Additionally, we find that larger models and bigger pre-training datasets not only enhance OOD performance but also improve calibration, helping to mitigate overconfidence, contrary to some prior studies that found modern deep networks to calibrate worse than classical shallow models. Our work underscores the overlooked importance of pre-trained model selection for out-of-distribution generalization and calibration.
format Preprint
id arxiv_https___arxiv_org_abs_2307_08187
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle An Empirical Study of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration
Naganuma, Hiroki
Hataya, Ryuichiro
Yoshida, Kotaro
Mitliagkas, Ioannis
Machine Learning
Artificial Intelligence
In the field of computer vision, fine-tuning pre-trained models has become a prevalent strategy for out-of-distribution (OOD) generalization tasks. Different from most prior work that has focused on advancing learning algorithms, we systematically examined how pre-trained model size, pre-training dataset size, and training strategies impact generalization and confidence calibration on downstream tasks. We evaluated 100 models across diverse pre-trained model sizes, five pre-training datasets, and five data augmentations through extensive experiments on four distribution shift datasets totaling over 120,000 GPU hours. Our results demonstrate the significant impact of pre-trained model selection, with optimal choices substantially improving OOD accuracy over algorithm improvement alone. Additionally, we find that larger models and bigger pre-training datasets not only enhance OOD performance but also improve calibration, helping to mitigate overconfidence, contrary to some prior studies that found modern deep networks to calibrate worse than classical shallow models. Our work underscores the overlooked importance of pre-trained model selection for out-of-distribution generalization and calibration.
title An Empirical Study of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2307.08187