Enregistré dans:
Détails bibliographiques
Auteurs principaux: Tao, Chen, Shen, Li, Mondal, Soumik
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2410.01709
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866929524429553664
author Tao, Chen
Shen, Li
Mondal, Soumik
author_facet Tao, Chen
Shen, Li
Mondal, Soumik
contents Test-time domain adaptation is a challenging task that aims to adapt a pre-trained model to limited, unlabeled target data during inference. Current methods that rely on self-supervision and entropy minimization underperform when the self-supervised learning (SSL) task does not align well with the primary objective. Additionally, minimizing entropy can lead to suboptimal solutions when there is limited diversity within minibatches. This paper introduces a meta-learning minimax framework for test-time training on batch normalization (BN) layers, ensuring that the SSL task aligns with the primary task while addressing minibatch overfitting. We adopt a mixed-BN approach that interpolates current test batch statistics with the statistics from source domains and propose a stochastic domain synthesizing method to improve model generalization and robustness to domain shifts. Extensive experiments demonstrate that our method surpasses state-of-the-art techniques across various domain adaptation and generalization benchmarks, significantly enhancing the pre-trained model's robustness on unseen domains.
format Preprint
id arxiv_https___arxiv_org_abs_2410_01709
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Meta-TTT: A Meta-learning Minimax Framework For Test-Time Training
Tao, Chen
Shen, Li
Mondal, Soumik
Machine Learning
Test-time domain adaptation is a challenging task that aims to adapt a pre-trained model to limited, unlabeled target data during inference. Current methods that rely on self-supervision and entropy minimization underperform when the self-supervised learning (SSL) task does not align well with the primary objective. Additionally, minimizing entropy can lead to suboptimal solutions when there is limited diversity within minibatches. This paper introduces a meta-learning minimax framework for test-time training on batch normalization (BN) layers, ensuring that the SSL task aligns with the primary task while addressing minibatch overfitting. We adopt a mixed-BN approach that interpolates current test batch statistics with the statistics from source domains and propose a stochastic domain synthesizing method to improve model generalization and robustness to domain shifts. Extensive experiments demonstrate that our method surpasses state-of-the-art techniques across various domain adaptation and generalization benchmarks, significantly enhancing the pre-trained model's robustness on unseen domains.
title Meta-TTT: A Meta-learning Minimax Framework For Test-Time Training
topic Machine Learning
url https://arxiv.org/abs/2410.01709