A federated anti-forgetting representation method based on hybrid model architecture and gradient truncation

Hui Wang; Jie Sun; Tianyu Wo; Xudong Liu; Suzhen Pei

doi:10.1007/s11704-024-40557-w

A federated anti-forgetting representation method based on hybrid model architecture and gradient truncation

Hui Wang^*
, Jie Sun
, Tianyu Wo^*
, Xudong Liu
, Suzhen Pei

^*Corresponding author for this work

Beihang University
Zhongguancun Laboratory
Southwest University

Research output: Contribution to journal › Letter › peer-review

1 Scopus citations

Abstract

This paper formulates the anti-forgetting representation problem under UFCL and proposes FedAFR. The experiments show that FedAFR effectively improves the model’s anti-forgetting performance.

Original language	English
Article number	196339
Journal	Frontiers of Computer Science
Volume	19
Issue number	6
DOIs	https://doi.org/10.1007/s11704-024-40557-w
State	Published - Jun 2025

Access to Document

10.1007/s11704-024-40557-w

Cite this

@article{c78aa755fb3148d598cf4f7aeb212c7f,

title = "A federated anti-forgetting representation method based on hybrid model architecture and gradient truncation",

abstract = "This paper formulates the anti-forgetting representation problem under UFCL and proposes FedAFR. The experiments show that FedAFR effectively improves the model{\textquoteright}s anti-forgetting performance.",

author = "Hui Wang and Jie Sun and Tianyu Wo and Xudong Liu and Suzhen Pei",

note = "Publisher Copyright: {\textcopyright} Higher Education Press 2025.",

year = "2025",

month = jun,

doi = "10.1007/s11704-024-40557-w",

language = "英语",

volume = "19",

journal = "Frontiers of Computer Science",

issn = "2095-2228",

publisher = "Higher Education Press Limited Company",

number = "6",

}

TY - JOUR

T1 - A federated anti-forgetting representation method based on hybrid model architecture and gradient truncation

AU - Wang, Hui

AU - Sun, Jie

AU - Wo, Tianyu

AU - Liu, Xudong

AU - Pei, Suzhen

PY - 2025/6

Y1 - 2025/6

N2 - This paper formulates the anti-forgetting representation problem under UFCL and proposes FedAFR. The experiments show that FedAFR effectively improves the model’s anti-forgetting performance.

AB - This paper formulates the anti-forgetting representation problem under UFCL and proposes FedAFR. The experiments show that FedAFR effectively improves the model’s anti-forgetting performance.

UR - https://www.scopus.com/pages/publications/85211773686

U2 - 10.1007/s11704-024-40557-w

DO - 10.1007/s11704-024-40557-w

M3 - 快报

AN - SCOPUS:85211773686

SN - 2095-2228

VL - 19

JO - Frontiers of Computer Science

JF - Frontiers of Computer Science

IS - 6

M1 - 196339

ER -

A federated anti-forgetting representation method based on hybrid model architecture and gradient truncation

Abstract

Access to Document

Other files and links

Fingerprint

Cite this