ActPerFL: Active Personalized Federated Learning

Huili Chen; Jie Ding; Eric Tramel; Shuang Wu; Anit Kumar Sahu; Salman Avestimehr; Tao Zhang

ActPerFL: Active Personalized Federated Learning

Huili Chen
, Jie Ding
, Eric Tramel
, Shuang Wu
, Anit Kumar Sahu
, Salman Avestimehr
, Tao Zhang

Statistics (Twin Cities)

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

8 Scopus citations

Abstract

In the context of personalized federated learning (FL), the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned. Inspired by Bayesian hierarchical models, we develop ActPerFL, a self-aware personalized FL method where each client can automatically balance the training of its local personal model and the global model that implicitly contributes to other clients' training. Such a balance is derived from the inter-client and intra-client uncertainty quantification. Consequently, ActPerFL can adapt to the underlying clients' heterogeneity with uncertainty-driven local training and model aggregation. With experimental studies on Sent140 and Amazon Alexa audio data, we show that ActPerFL can achieve superior personalization performance compared with the existing counterparts.

Original language	English (US)
Title of host publication	FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop
Editors	Bill Yuchen Lin, Chaoyang He, Chulin Xie, Fatemehsadat Mireshghallah, Ninareh Mehrabi, Tian Li, Mahdi Soltanolkotabi, Xiang Ren
Publisher	Association for Computational Linguistics (ACL)
Pages	1-5
Number of pages	5
ISBN (Electronic)	9781955917377
State	Published - 2022
Event	1st Workshop on Federated Learning for Natural Language Processing, FL4NLP 2022 - Dublin, Ireland Duration: May 27 2022 → …

Publication series

Name	FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop

Conference

Conference	1st Workshop on Federated Learning for Natural Language Processing, FL4NLP 2022
Country/Territory	Ireland
City	Dublin
Period	5/27/22 → …

Bibliographical note

Publisher Copyright:
© 2022 Association for Computational Linguistics.

Find It

Find It at U of M Libraries

Cite this

Chen, H., Ding, J., Tramel, E., Wu, S., Sahu, A. K., Avestimehr, S., & Zhang, T. (2022). ActPerFL: Active Personalized Federated Learning. In B. Y. Lin, C. He, C. Xie, F. Mireshghallah, N. Mehrabi, T. Li, M. Soltanolkotabi, & X. Ren (Eds.), FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop (pp. 1-5). (FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop). Association for Computational Linguistics (ACL).

ActPerFL: Active Personalized Federated Learning. / Chen, Huili; Ding, Jie; Tramel, Eric et al.
FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop. ed. / Bill Yuchen Lin; Chaoyang He; Chulin Xie; Fatemehsadat Mireshghallah; Ninareh Mehrabi; Tian Li; Mahdi Soltanolkotabi; Xiang Ren. Association for Computational Linguistics (ACL), 2022. p. 1-5 (FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Chen, H, Ding, J, Tramel, E, Wu, S, Sahu, AK, Avestimehr, S & Zhang, T 2022, ActPerFL: Active Personalized Federated Learning. in BY Lin, C He, C Xie, F Mireshghallah, N Mehrabi, T Li, M Soltanolkotabi & X Ren (eds), FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop. FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop, Association for Computational Linguistics (ACL), pp. 1-5, 1st Workshop on Federated Learning for Natural Language Processing, FL4NLP 2022, Dublin, Ireland, 5/27/22.

Chen H, Ding J, Tramel E, Wu S, Sahu AK, Avestimehr S et al. ActPerFL: Active Personalized Federated Learning. In Lin BY, He C, Xie C, Mireshghallah F, Mehrabi N, Li T, Soltanolkotabi M, Ren X, editors, FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop. Association for Computational Linguistics (ACL). 2022. p. 1-5. (FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop).

Chen, Huili ; Ding, Jie ; Tramel, Eric et al. / ActPerFL : Active Personalized Federated Learning. FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop. editor / Bill Yuchen Lin ; Chaoyang He ; Chulin Xie ; Fatemehsadat Mireshghallah ; Ninareh Mehrabi ; Tian Li ; Mahdi Soltanolkotabi ; Xiang Ren. Association for Computational Linguistics (ACL), 2022. pp. 1-5 (FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop).

@inproceedings{9a3c77a2044d4891985bfad5b9372bde,

title = "ActPerFL: Active Personalized Federated Learning",

abstract = "In the context of personalized federated learning (FL), the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned. Inspired by Bayesian hierarchical models, we develop ActPerFL, a self-aware personalized FL method where each client can automatically balance the training of its local personal model and the global model that implicitly contributes to other clients' training. Such a balance is derived from the inter-client and intra-client uncertainty quantification. Consequently, ActPerFL can adapt to the underlying clients' heterogeneity with uncertainty-driven local training and model aggregation. With experimental studies on Sent140 and Amazon Alexa audio data, we show that ActPerFL can achieve superior personalization performance compared with the existing counterparts.",

author = "Huili Chen and Jie Ding and Eric Tramel and Shuang Wu and Sahu, \{Anit Kumar\} and Salman Avestimehr and Tao Zhang",

note = "Publisher Copyright: {\textcopyright} 2022 Association for Computational Linguistics.; 1st Workshop on Federated Learning for Natural Language Processing, FL4NLP 2022 ; Conference date: 27-05-2022",

year = "2022",

language = "English (US)",

series = "FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop",

publisher = "Association for Computational Linguistics (ACL)",

pages = "1--5",

editor = "Lin, \{Bill Yuchen\} and Chaoyang He and Chulin Xie and Fatemehsadat Mireshghallah and Ninareh Mehrabi and Tian Li and Mahdi Soltanolkotabi and Xiang Ren",

booktitle = "FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop",

}

TY - GEN

T1 - ActPerFL

T2 - 1st Workshop on Federated Learning for Natural Language Processing, FL4NLP 2022

AU - Chen, Huili

AU - Ding, Jie

AU - Tramel, Eric

AU - Wu, Shuang

AU - Sahu, Anit Kumar

AU - Avestimehr, Salman

AU - Zhang, Tao

PY - 2022

Y1 - 2022

N2 - In the context of personalized federated learning (FL), the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned. Inspired by Bayesian hierarchical models, we develop ActPerFL, a self-aware personalized FL method where each client can automatically balance the training of its local personal model and the global model that implicitly contributes to other clients' training. Such a balance is derived from the inter-client and intra-client uncertainty quantification. Consequently, ActPerFL can adapt to the underlying clients' heterogeneity with uncertainty-driven local training and model aggregation. With experimental studies on Sent140 and Amazon Alexa audio data, we show that ActPerFL can achieve superior personalization performance compared with the existing counterparts.

AB - In the context of personalized federated learning (FL), the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned. Inspired by Bayesian hierarchical models, we develop ActPerFL, a self-aware personalized FL method where each client can automatically balance the training of its local personal model and the global model that implicitly contributes to other clients' training. Such a balance is derived from the inter-client and intra-client uncertainty quantification. Consequently, ActPerFL can adapt to the underlying clients' heterogeneity with uncertainty-driven local training and model aggregation. With experimental studies on Sent140 and Amazon Alexa audio data, we show that ActPerFL can achieve superior personalization performance compared with the existing counterparts.

UR - https://www.scopus.com/pages/publications/85133020845

UR - https://www.scopus.com/pages/publications/85133020845#tab=citedBy

M3 - Conference contribution

AN - SCOPUS:85133020845

T3 - FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop

SP - 1

EP - 5

BT - FL4NLP 2022 - 1st Workshop on Federated Learning for Natural Language Processing, Proceedings of the Workshop

A2 - Lin, Bill Yuchen

A2 - He, Chaoyang

A2 - Xie, Chulin

A2 - Mireshghallah, Fatemehsadat

A2 - Mehrabi, Ninareh

A2 - Li, Tian

A2 - Soltanolkotabi, Mahdi

A2 - Ren, Xiang

PB - Association for Computational Linguistics (ACL)

Y2 - 27 May 2022

ER -

ActPerFL: Active Personalized Federated Learning

Abstract

Publication series

Conference

Bibliographical note

Find It

Other files and links

Fingerprint

Cite this