File Download

There are no files associated with this item.

  • Find it @ UNIST can give you direct access to the published full text of this article. (UNISTARs only)
Related Researcher

공태식

Gong, Taesik
Ubiquitous AI Lab
Read More

Views & Downloads

Detailed Information

Cited time in webofscience Cited time in scopus
Metadata Downloads

NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation

Author(s)
Gong, TaesikJeong, J.Kim, T.Kim, Y.Shin, J.Lee, S.-J.
Issued Date
2022-11-28
URI
https://scholarworks.unist.ac.kr/handle/201301/84397
Citation
Neural Information Processing Systems
Abstract
Test-time adaptation (TTA) is an emerging paradigm that addresses distributional shifts between training and testing phases without additional data acquisition or labeling cost; only unlabeled test data streams are used for continual model adaptation. Previous TTA schemes assume that the test samples are independent and identically distributed (i.i.d.), even though they are often temporally correlated (non-i.i.d.) in application scenarios, e.g., autonomous driving. We discover that most existing TTA methods fail dramatically under such scenarios. Motivated by this, we present a new test-time adaptation scheme that is robust against non-i.i.d. test data streams. Our novelty is mainly two-fold: (a) Instance-Aware Batch Normalization (IABN) that corrects normalization for out-of-distribution samples, and (b) Prediction-balanced Reservoir Sampling (PBRS) that simulates i.i.d. data stream from non-i.i.d. stream in a class-balanced manner. Our evaluation with various datasets, including real-world non-i.i.d. streams, demonstrates that the proposed robust TTA not only outperforms state-of-the-art TTA algorithms in the non-i.i.d. setting, but also achieves comparable performance to those algorithms under the i.i.d. assumption. Code is available at https://github.com/TaesikGong/NOTE. © 2022 Neural information processing systems foundation. All rights reserved.
Publisher
Neural information processing systems foundation
ISSN
1049-5258

qrcode

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.