Applied Energy, Journal Year: 2024, Volume and Issue: 378, P. 124641 - 124641
Published: Nov. 14, 2024
Language: Английский
Applied Energy, Journal Year: 2024, Volume and Issue: 378, P. 124641 - 124641
Published: Nov. 14, 2024
Language: Английский
IEEE Internet of Things Journal, Journal Year: 2024, Volume and Issue: 11(21), P. 34617 - 34638
Published: May 30, 2024
Federated learning (FL) has been gaining attention for its ability to share knowledge while maintaining user data, protecting privacy, increasing efficiency, and reducing communication overhead. Decentralized FL (DFL) is a decentralized network architecture that eliminates the need central server in contrast centralized (CFL). DFL enables direct between clients, resulting significant savings resources. In this paper, comprehensive survey profound perspective are provided DFL. First, review of methodology, challenges, variants CFL conducted, laying background Then, systematic detailed on introduced, including iteration order, protocols, topologies, paradigm proposals, temporal variability. Next, based definition DFL, several extended categorizations proposed with state-of-the-art (SOTA) technologies. Lastly, addition summarizing current challenges some possible solutions future research directions also discussed.
Language: Английский
Citations
44International Journal of Production Research, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 23
Published: Jan. 4, 2025
Information sharing in supply chains can be challenged by privacy concerns. Equating data and information, the existing literature primarily focuses on incentivisation behind information between firms. The field of AI may bring a new way looking at this problem asking following question: what if we do not share raw but learned from it instead? This raises next question, with whom when should chain members which address paper. We develop novel adaptive federated learning approach for generation usage collective knowledge without direct exchange test use case collectively predicting risk. propose privacy-preserving network formation clustering algorithm, enables to decide enter information-sharing network, how they form teams. Using an e-commerce platform, illustrate our outperforms suppliers' own prediction models. further show that suppliers teams achieves best performance converges faster compared two benchmarks. heterogeneity contribution firms those who benefit also important research questions role cooperation chains.
Language: Английский
Citations
8Journal of Sensor and Actuator Networks, Journal Year: 2025, Volume and Issue: 14(1), P. 9 - 9
Published: Jan. 22, 2025
Federated Learning (FL) has emerged as a pivotal approach for decentralized Machine (ML), addressing the unique demands of Internet Things (IoT) environments where data privacy, bandwidth constraints, and device heterogeneity are paramount. This survey provides comprehensive overview FL, focusing on its integration with IoT. We delve into motivations behind adopting FL IoT, underlying techniques that facilitate this integration, challenges posed by IoT environments, diverse range applications is making an impact. Finally, submission also outlines future research directions open issues, aiming to provide detailed roadmap advancing in settings.
Language: Английский
Citations
6IEEE Transactions on Pattern Analysis and Machine Intelligence, Journal Year: 2023, Volume and Issue: 46(2), P. 712 - 728
Published: Oct. 25, 2023
Federated learning is an important privacy-preserving multi-party paradigm, involving collaborative with others and local updating on private data. Model heterogeneity catastrophic forgetting are two crucial challenges, which greatly limit the applicability generalizability. This paper presents a novel FCCL+, federated correlation similarity non-target distillation, facilitating both intra-domain discriminability inter-domain generalization. For issue, we leverage irrelevant unlabeled public data for communication between heterogeneous participants. We construct cross-correlation matrix align instance distribution logits feature levels, effectively overcomes barrier improves generalizable ability. in stage, FCCL+ introduces Non Target Distillation, retains knowledge while avoiding optimization conflict fulling distilling privileged information through depicting posterior classes relation. Considering that there no standard benchmark evaluating existing under same setting, present comprehensive extensive representative methods four domain shift scenarios, supporting homogeneous settings. Empirical results demonstrate superiority of our method efficiency modules various scenarios. The code reproducing available at https://github.com/WenkeHuang/FCCL.
Language: Английский
Citations
29Sensors, Journal Year: 2024, Volume and Issue: 24(3), P. 968 - 968
Published: Feb. 1, 2024
Federated learning (FL) is a machine (ML) technique that enables collaborative model training without sharing raw data, making it ideal for Internet of Things (IoT) applications where data are distributed across devices and privacy concern. Wireless Sensor Networks (WSNs) play crucial role in IoT systems by collecting from the physical environment. This paper presents comprehensive survey integration FL, IoT, WSNs. It covers FL basics, strategies, types discusses WSNs various domains. The addresses challenges related to heterogeneity summarizes state-of-the-art research this area. also explores security considerations performance evaluation methodologies. outlines latest achievements potential directions emphasizes significance surveyed topics within context current technological advancements.
Language: Английский
Citations
17Information Fusion, Journal Year: 2024, Volume and Issue: 106, P. 102290 - 102290
Published: Feb. 10, 2024
Language: Английский
Citations
11Neurocomputing, Journal Year: 2024, Volume and Issue: 599, P. 128089 - 128089
Published: June 22, 2024
Language: Английский
Citations
11Big Data and Cognitive Computing, Journal Year: 2025, Volume and Issue: 9(2), P. 18 - 18
Published: Jan. 21, 2025
Synthetic Data Generation (SDG) is a promising solution for healthcare, offering the potential to generate synthetic patient data closely resembling real-world while preserving privacy. However, scarcity and heterogeneity, particularly in under-resourced regions, challenge effective implementation of SDG. This paper addresses these challenges using Federated Learning (FL) SDG, focusing on sharing patients across nodes. By leveraging collective knowledge diverse distributions, we hypothesize that can significantly enhance quality representativeness generated data, institutions with limited or biased datasets. approach aligns meta-learning concepts, like Domain Randomized Search. We compare two FL techniques, FedAvg Sharing (SDS), latter being our proposed contribution. Both approaches are evaluated variational autoencoders Bayesian Gaussian mixture models medical Our results demonstrate both methods improve SDS consistently outperforms FedAvg, producing higher-quality, more representative data. Non-IID scenarios reveal achieves improvements 13–27% reducing divergence compared isolated training, reductions exceeding 50% worst-performing These findings underscore reduce disparities between data-rich data-poor institutions, fostering equitable healthcare research innovation.
Language: Английский
Citations
2Electronics, Journal Year: 2025, Volume and Issue: 14(4), P. 745 - 745
Published: Feb. 14, 2025
Clustered federated learning has garnered significant attention as an effective strategy for enhancing model performance in non-independent and identically distributed (non-IID) data scenarios. This approach improves such environments by calculating the similarity between users clustering them into multiple groups. However, several challenges arise when implementing this method, particularly balancing flexibility, communication costs, performance. To address these issues, paper proposes a novel hierarchical framework that balances both network The performs principal component analysis (PCA) on device-side image datasets to assess of private across devices and, conjunction with measurements, dynamically adjusts strategies minimize latency while ensuring stable By weighting metrics, optimizes efficiency without significantly compromising validate proposed method’s effectiveness, we employed three publicly available compared it against four baseline methods. experimental results demonstrate SC-Fed (segmented clustering-federated learning) achieves maximum accuracy improvement 7.56% over methods, also reducing average waiting time 54.6%. These indicate algorithm enhances applicability clustered practical training
Language: Английский
Citations
2Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 152, P. 110724 - 110724
Published: April 11, 2025
Language: Английский
Citations
2