We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.DC

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Distributed, Parallel, and Cluster Computing

Title: Data-Quality Based Scheduling for Federated Edge Learning

Abstract: FEderated Edge Learning (FEEL) has emerged as a leading technique for privacy-preserving distributed training in wireless edge networks, where edge devices collaboratively train machine learning (ML) models with the orchestration of a server. However, due to frequent communication, FEEL needs to be adapted to the limited communication bandwidth. Furthermore, the statistical heterogeneity of local datasets' distributions, and the uncertainty about the data quality pose important challenges to the training's convergence. Therefore, a meticulous selection of the participating devices and an analogous bandwidth allocation are necessary. In this paper, we propose a data-quality based scheduling (DQS) algorithm for FEEL. DQS prioritizes reliable devices with rich and diverse datasets. In this paper, we define the different components of the learning algorithm and the data-quality evaluation. Then, we formulate the device selection and the bandwidth allocation problem. Finally, we present our DQS algorithm for FEEL, and we evaluate it in different data poisoning scenarios.
Comments: 2021 IEEE 46th Conference on Local Computer Networks (LCN)
Subjects: Distributed, Parallel, and Cluster Computing (cs.DC)
DOI: 10.1109/LCN52139.2021.9524974
Cite as: arXiv:2201.11247 [cs.DC]
  (or arXiv:2201.11247v1 [cs.DC] for this version)

Submission history

From: Afaf Taik [view email]
[v1] Thu, 27 Jan 2022 00:40:44 GMT (2774kb,D)

Link back to: arXiv, form interface, contact.