OES-Fed: a federated learning framework in vehicular network based on noise data filtering

View article
PeerJ Computer Science

Main article text

 

Introduction

Literature review

Federated learning and isolated data islands

Solutions of the noise data

  • Current perspective: data outlier

  • Historical perspective: data iteration

Scene description

Our contribution

  • From the current perspective of noise data, this article provides an outlier detection method based on K-means. By screening excellent subsets and taking them as the initial clustering center, we solved the problem of significant changes in vehicle data sets in IOV, and preliminarily reduced the noise data.

  • From the historical perspective of noise data, we consider the past training performance of the vehicle. Due to the influence of noise data, this article proposed the fusion of the Kalman filter and exponential smoothing to achieve the effect of noise reduction.

  • After current and historical noise reduction, the actual accuracy of vehicles will be significantly affected. This article solved the problem by reintroducing the results of the iterative data filtering into the K-means algorithm and updating the filtering criteria.

  • The proposed OES-Fed framework is trained on three datasets, and the global ML models are evaluated using accuracy, loss and area under the curve (AUC) metrics respectively. As a result, the OES-Fed framework proposed in this article achieved higher accuracy, lower loss, and better AUC.

System model and problem description

System model

  1. The central server is the core processing node of IoV, with high-speed computing capability and good scalability, which can run reliably for a long time and can undertake any level of computing work in the network. Meanwhile, the server does not receive the original data collected by the vehicle, but only the local ML model parameters transmitted by the access point (AP), which are the parameters of the ML model formed by the vehicle through training the local dataset of the vehicle client. This article assumes that the server has infinite computing power.

  2. The AP is the base station or roadside unit, which is equipped with communication and computational processing. In addition, the AP adjusts the training resources of the vehicle according to the instructions of the server.

  3. The vehicle will generate a large amount of driving data and related pictures at the vehicle equipment during the driving process, and the driving data will be processed and saved locally, forming the local dataset: DataLocaln,i={xn,iTi,yn,iTi},xn,i is the input sample vector for vehicle n to participate in training, and yn,i is the label of the input sample vector. When the server has started some task, the vehicle will train the ML model for the base and upload the local ML model parameters to the server via the AP.

Problem description

Our proposed framework: oes-fed

Phase A: current perspective of nosie data

Motivation: current perspective of nosie data filtering

Selecting the initial clustering centre: the excellent subset

Outlier algorithm based on excellent subset

Current perspective on noise data: the improved K-means scheme based on the outlier algorithm

Phase B: historical perspective of noise data

Motivation: historical perspective of noise data filtering

Kalman filter processing

Historical perspective of noise data: the improved exponential smoothing scheme based on Kalman filter

Phase C: the overall scheme of data iteration and resource alignment

Motivation: the overall scheme

Data iteration: (Kalman filter + Exponential smoothing) + (Excellent subset + Outlier + Improved K-means)

Client training adjustment

Results

Current and historical perspective’s noise data filtering results

OES-FED framework’s accuracy, loss and AUC results

Discussion

Conclusion

Supplemental Information

Coding vehicle classification dataset.

DOI: 10.7717/peerj-cs.1101/supp-3

Part of experimental results.

DOI: 10.7717/peerj-cs.1101/supp-4

MNIST.

Training set images and labels and test set images and labels.

DOI: 10.7717/peerj-cs.1101/supp-5

Additional Information and Declarations

Competing Interests

The authors declare that they have no competing interests.

Author Contributions

Yuan Lei conceived and designed the experiments, performed the experiments, performed the computation work, authored or reviewed drafts of the article, and approved the final draft.

Shir Li Wang conceived and designed the experiments, performed the experiments, performed the computation work, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Caiyu Su performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Theam Foo Ng performed the experiments, analyzed the data, performed the computation work, authored or reviewed drafts of the article, and approved the final draft.

Data Availability

The following information was supplied regarding data availability:

The raw data is available in the Supplemental Files.

CIFAR-10 python is available at the University of Toronto:

http://www.cs.toronto.edu/~kriz/cifar.html.

The “vehicle classification dataset” (913 MB) is available at AIStudio and accessible using GitHub:

https://aistudio.baidu.com/aistudio/datasetdetail/95878?lang=en.

Funding

This work was supported by the Higher Education Department of the Ministry of Education of P. R. China, Industry and University Cooperation Collaborative Education Project, under Grant 202002118061 and by the young and middle-aged teachers’ basic ability improvement of Guangxi colleges in 2022 under Grant 2022KY1296. There was no additional external funding received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

19 Citations 1,297 Views 116 Downloads

Your institution may have Open Access funds available for qualifying authors. See if you qualify

Publish for free

Comment on Articles or Preprints and we'll waive your author fee
Learn more

Five new journals in Chemistry

Free to publish • Peer-reviewed • From PeerJ
Find out more