Nothing Special   »   [go: up one dir, main page]

Class based Influence Functions for Error Detection

Thang Nguyen-Duc, Hoang Thanh-Tung, Quan Hung Tran, Dang Huu-Tien, Hieu Nguyen, Anh T. V. Dau, Nghi Bui


Abstract
Influence functions (IFs) are a powerful tool for detecting anomalous examples in large scale datasets. However, they are unstable when applied to deep networks. In this paper, we provide an explanation for the instability of IFs and develop a solution to this problem. We show that IFs are unreliable when the two data points belong to two different classes. Our solution leverages class information to improve the stability of IFs.Extensive experiments show that our modification significantly improves the performance and stability of IFs while incurring no additional computational cost.
Anthology ID:
2023.acl-short.104
Volume:
Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1204–1218
Language:
URL:
https://aclanthology.org/2023.acl-short.104
DOI:
10.18653/v1/2023.acl-short.104
Bibkey:
Cite (ACL):
Thang Nguyen-Duc, Hoang Thanh-Tung, Quan Hung Tran, Dang Huu-Tien, Hieu Nguyen, Anh T. V. Dau, and Nghi Bui. 2023. Class based Influence Functions for Error Detection. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 1204–1218, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
Class based Influence Functions for Error Detection (Nguyen-Duc et al., ACL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.acl-short.104.pdf
Video:
 https://aclanthology.org/2023.acl-short.104.mp4