Deakin University
Browse

File(s) under embargo

Tackling Model Mismatch with Mixup Regulated Test-Time Training

conference contribution
posted on 2024-01-05, 04:12 authored by B Zhang, R Shao, J Du, P Yuen, Wei LuoWei Luo
Test-time training (TTT) is an emerging approach for addressing the problem of domain shift. In its framework, a test-time training phase is inserted between the training phase and the test phase. During the test-time training phase, the representation layers are adapted using an auxiliary task. Then the updated model will be used in the test phase. Although the idea is very intuitive, TTT does not demonstrate competitive performance compared with some other domain adaption methods. In this paper, we present both theoretical and empirical analyses to explain the subpar performance of TTT. In particular, we point out that TTT causes a new kind of problem, which we term as Model Mismatch. To address this problem of Model Mismatch, we analyse a simple yet effective method inspired by the idea of mixup in robust training. Such effectiveness is shown in the experimental results.

History

Volume

00

Pagination

1-7

Location

Thessaloniki, Greece

Start date

2023-10-09

End date

2023-10-13

ISBN-13

9798350345032

Language

eng

Title of proceedings

2023 IEEE 10th International Conference on Data Science and Advanced Analytics, DSAA 2023 - Proceedings

Event

2023 IEEE 10th International Conference on Data Science and Advanced Analytics (DSAA)

Publisher

IEEE

Place of publication

Piscataway, N.J.

Usage metrics

    Research Publications

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC