Personalized Federated Learning with Model Interpolation among Client Clusters and its Application in Smart Home

Yang, Zhikai; Liu, Yaping; Zhang, Shuo; Zhou, Keshen

doi:10.21203/rs.3.rs-2173011/v1

“…To contextualize and benchmark our findings, we conducted a comparative analysis in Table 5. We contrasted our results with those of prior studies that utilized analogous datasets, drawing upon the APPLE [16], FedFOMO [23], and FedSup [9] frameworks to train Non-DP FL models. This comparative exercise enabled us to gauge the relative effectiveness and privacy implications of our DP FL model in relation to its Non-DP FL counterparts utilized in previous research endeavors.…”

Section: B Experimental Resultsmentioning

confidence: 94%

“…The Pathological Non-IID setting is intended to assess how well FL algorithms perform under extremely non-IID conditions, while the Practical Non-IID setting aims to replicate a more realistic scenario seen in medical applications where institutes located in different regions may have datasets with varying sizes and imbalanced distributions due to different demographic distributions of patients and categories of data. (Table 1) compares their results to [23]'s results under practical Non-IID settings. [24], techniques for preserving privacy were reviewed in conjunction with machine learning, and subsequently applied in the classification of chest X-rays and segmentation of CT scans.…”

Section: Related Workmentioning

confidence: 99%

A Differentially Private Federated Learning Application in Privacy-Preserving Medical Imaging

FARES,

SERTBAŞ

2024

Preprint

0

View full text Add to dashboard Cite

This research addresses the escalating concerns surrounding privacy, particularly in the context of safeguarding sensitive medical data within the increasingly demanding healthcare landscape. We undertake an experimental exploration of differentially private federated learning systems, employing three benchmark datasets—PathMNIST, BloodMNIST, and OrganAMNIST—for medical image classification. This study pioneers the application of federated learning with differential privacy in healthcare, closely simulating real-world data distribution across twelve hospitals. Additionally, we introduce a novel deep-learning architecture tailored for differentially private models. Our findings demonstrate the superior performance of federated learning models compared to traditional approaches, with accuracy levels approaching those of non-private settings. By leveraging resilient deep learning models, we aim to enhance privacy, efficiency, and effectiveness in healthcare solutions, benefiting patients, healthcare practitioners, and the overall healthcare system through privacy-protected healthcare.

show abstract

“…Authors in [23] proposed a new approach to personalized federated learning. The authors introduce a method for efficiently calculating optimal weighted model combinations for each client, based on determining how much a client can benefit from a model trained on other clients' data.…”

Section: Related Workmentioning

confidence: 99%

A Differentially Private Federated Learning Application in Privacy-Preserving Medical Imaging

FARES,

SERTBAŞ

2024

Preprint

0

View full text Add to dashboard Cite

This research addresses the escalating concerns surrounding privacy, particularly in the context of safeguarding sensitive medical data within the increasingly demanding healthcare landscape. We undertake an experimental exploration of differentially private federated learning systems, employing three benchmark datasets—PathMNIST, BloodMNIST, and OrganAMNIST—for medical image classification. This study pioneers the application of federated learning with differential privacy in healthcare, closely simulating real-world data distribution across twelve hospitals. Additionally, we introduce a novel deep-learning architecture tailored for differentially private models. Our findings demonstrate the superior performance of federated learning models compared to traditional approaches, with accuracy levels approaching those of non-private settings. By leveraging resilient deep learning models, we aim to enhance privacy, efficiency, and effectiveness in healthcare solutions, benefiting patients, healthcare practitioners, and the overall healthcare system through privacy-protected healthcare.

show abstract

“…The others are dedicated to improve the mixture of global and local model. There are many ways to help the local models to mix the global model including L2GD [17], AL2GD [18] and many other methods [7, 19–22]. And certainly, there are also lots of papers studying the personalised federated learning using clustering methods [8, 23].…”

Section: Related Workmentioning

confidence: 99%

A novel parameter decoupling approach of personalised federated learning for image analysis

Su

¹

,

Pang

²

,

Wang

³

2023

IET Computer Vision

View full text Add to dashboard Cite

Given the importance of privacy protection in databases and other institutions, federated learning (FL) is used to benefit training machine learning models based on these decentralised and private data so as to address the growing vision tasks. However, for federated learning, statistical heterogeneity continues to be a major problem. Recently, plenty of personalised federated learning methods have been explored to solve the problem of statistical heterogeneity. The usage of trained base layers and the effect of feature extraction in personalised layers, however, are hardly considered in those methods that employ the learning personalised models approach. To address the problem of the statistical heterogeneity in image analysis, PCCFED, a personalised federated learning method utilizing the strategy of parameter decoupling is proposed. It should be emphasised that the authors’ personalised federated learning method decouples the personalised (P) layers into a connecting (C) layer and classifier (C) layer in order to enhance the effectiveness of feature learning for personalised layers. Further, an approach is proposed to fully use the base layers to adapt a personalised model based on the newly admitted institution's dataset through meta‐transfer. The performance of the proposed PCCFED on three datasets is evaluated under the practical non‐independent and identically distributed (non‐IID) setting. Extensive experiments demonstrate that compared with baseline methods, the proposed framework achieves the best performance in federated learning and fine‐tuning. Through FL, the investigation reveals a method to reduce statistical heterogeneity while protecting the institutions' privacy.

show abstract

Personalized Federated Learning with Model Interpolation among Client Clusters and its Application in Smart Home

Cited by 4 publications

References 24 publications

A Differentially Private Federated Learning Application in Privacy-Preserving Medical Imaging

A Differentially Private Federated Learning Application in Privacy-Preserving Medical Imaging

A Differentially Private Federated Learning Application in Privacy-Preserving Medical Imaging

A novel parameter decoupling approach of personalised federated learning for image analysis

Contact Info

Product

Resources

About