A Fair Contribution Measurement Method for Federated Learning DOI Creative Commons
Peng Guo, Yanqing Yang, Wei Guo

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(15), P. 4967 - 4967

Published: July 31, 2024

Federated learning is an effective approach for preserving data privacy and security, enabling machine to occur in a distributed environment promoting its development. However, urgent problem that needs be addressed how encourage active client participation federated learning. The Shapley value, classical concept cooperative game theory, has been utilized valuation services. Nevertheless, existing numerical evaluation schemes based on the value are impractical, as they necessitate additional model training, leading increased communication overhead. Moreover, participants' may exhibit Non-IID characteristics, posing significant challenge evaluating participant contributions. have greatly affected accuracy of global model, weakened marginal effect participants, led underestimated contribution measurement results participants. Current work often overlooks impact heterogeneity aggregation. This paper presents fair scheme addresses need computations. By introducing novel aggregation weight, it enhances measurement. Experiments MNIST Fashion dataset show proposed method can accurately compute contributions Compared baseline algorithms, significantly improved, with similar time cost.

Language: Английский

Sensing emotional valence and arousal dynamics through automated facial action unit analysis DOI Creative Commons
Junyao Zhang, Wataru Sato,

Naoya Kawamura

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Aug. 22, 2024

Information about the concordance between dynamic emotional experiences and objective signals is practically useful. Previous studies have shown that valence dynamics can be estimated by recording electrical activity from muscles in brows cheeks. However, whether facial actions based on video data analyzed without electrodes used for sensing emotion remains unknown. We investigated this issue of participants' faces obtaining arousal ratings while they observed films. Action units (AUs) 04 (i.e., brow lowering) 12 lip-corner pulling), detected through an automated analysis data, were negatively positively correlated with subjective valence, respectively. Several other AUs also or ratings. Random forest regression modeling, interpreted using SHapley Additive exPlanation tool, revealed non-linear associations arousal. These results suggest expression to estimate states, which could applied various fields including mental health diagnosis, security monitoring, education.

Language: Английский

Citations

2
A Fair Contribution Measurement Method for Federated Learning DOI Creative Commons
Peng Guo, Yanqing Yang, Wei Guo

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(15), P. 4967 - 4967

Published: July 31, 2024

Federated learning is an effective approach for preserving data privacy and security, enabling machine to occur in a distributed environment promoting its development. However, urgent problem that needs be addressed how encourage active client participation federated learning. The Shapley value, classical concept cooperative game theory, has been utilized valuation services. Nevertheless, existing numerical evaluation schemes based on the value are impractical, as they necessitate additional model training, leading increased communication overhead. Moreover, participants' may exhibit Non-IID characteristics, posing significant challenge evaluating participant contributions. have greatly affected accuracy of global model, weakened marginal effect participants, led underestimated contribution measurement results participants. Current work often overlooks impact heterogeneity aggregation. This paper presents fair scheme addresses need computations. By introducing novel aggregation weight, it enhances measurement. Experiments MNIST Fashion dataset show proposed method can accurately compute contributions Compared baseline algorithms, significantly improved, with similar time cost.

Language: Английский

Citations

1