Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Communication-Efficient Federated Learning with Dual-Sided Sparse Aggregation for Edge Sensing Systems.

He Zhao1, Jingwei Li1

  • 1School of Telecommunications Engineering, Xidian University, Xi'an 710071, China.

Sensors (Basel, Switzerland)
|May 13, 2026
PubMed
Summary

This study introduces Dual-Sided Sparse Aggregation (DSSA) with FedProx for efficient federated learning (FL) on edge devices. The method significantly cuts communication and computation costs while maintaining high accuracy for privacy-sensitive data.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

SpecNet: A hyperspectral band-learning method for slice-level differentiation of fibroadenoma and phyllodes tumor.

Photodiagnosis and photodynamic therapy·2025
Same author

FNDC4 Drives Metastasis and Immune Evasion in Pancreatic Cancer.

Cancer research·2025
Same author

Identification of five sleep-biopsychosocial profiles with specific neural signatures linking sleep variability with health, cognition, and lifestyle factors.

PLoS biology·2025
Same author

The IGF2BP3-FASN axis drives lipid metabolic reprogramming to promote brain colonization in non-small cell lung cancer.

Cell death & disease·2025
Same author

Correction: Effect of the renin-angiotensin-aldosterone system inhibitors on time to nucleic acid negative conversion in hypertensive patients with SARS-CoV-2 omicron infection: a propensity score matching study.

Hypertension research : official journal of the Japanese Society of Hypertension·2025
Same author

Metal cations effect of Ni active sites in layered double hydroxides enables efficient anion exchange membrane water electrolysis.

Journal of colloid and interface science·2025

Area of Science:

  • Edge computing
  • Machine learning
  • Data privacy

Background:

  • Edge sensing systems generate sensitive data, posing transmission cost and privacy challenges.
  • Federated learning (FL) enables collaborative training without raw data sharing but faces issues like non-IID data and resource constraints.
  • Existing FL methods struggle with the unique demands of resource-limited edge environments.

Purpose of the Study:

  • To propose a novel mechanism, Dual-Sided Sparse Aggregation (DSSA) integrated with FedProx, for efficient federated learning in edge sensing.
  • To address challenges of non-IID data, limited resources, and computation in practical FL deployments at the edge.
  • To reduce communication and computation overhead while maintaining model accuracy.

Main Methods:

  • Developed a Dual-Sided Sparse Aggregation (DSSA) mechanism combined with FedProx.
Keywords:
FedProx regularizationcommunication efficiencyfederated learningmodel pruningnon-IID data

Related Experiment Videos

  • Implemented a fixed-structure sparse training strategy where the server prunes the global model and clients update complementary parameters with sparse local gradients.
  • Evaluated the framework on CIFAR-10 and SVHN datasets under varying non-IID conditions and pruning ratios.

    • Main Results:

    • Achieved significant reductions in communication cost (up to 73.0%) and computation cost (up to 34.9%).
    • Maintained competitive model accuracy despite resource constraints and data heterogeneity.
    • Demonstrated substantial resource savings compared to FedAvg, especially in mildly heterogeneous scenarios.

    Conclusions:

    • The proposed DSSA integrated with FedProx offers a favorable resource-performance trade-off for federated learning in edge sensing.
    • The method effectively mitigates communication and computation burdens, making FL more practical for resource-constrained edge devices.
    • This approach shows promise for efficient and accurate collaborative model training in distributed edge environments.