Clustered Network Node Configuration for Federated Learning Scalability

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing Federated Learning systems face scalability issues and inefficiencies due to hardware and software disparities among network nodes, leading to inappropriate model updates and potential data noise, especially in large communications networks.

Innovation Solution

Implement a clustering mechanism within the network nodes, grouping nodes with similar properties into clusters, and using cluster computing devices to manage Federated Learning, allowing for localized model updates and efficient configuration parameter sharing through cloning, thereby optimizing energy performance and service quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If Federated Learning is implemented across all network nodes, then model accuracy and service quality assurance are improved, but system complexity and scalability issues arise due to hundreds of thousands of nodes

Engineering Contradiction:
Improveservice quality assuranceVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the large-scale Federated Learning system into hierarchical clusters. Network nodes are organized into clusters with designated cluster heads, dividing the overall system into manageable sub-systems. This segmentation reduces the complexity of coordinating updates across all nodes while maintaining the benefits of distributed learning for service quality assurance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces cluster heads as intermediary entities between individual network nodes and the central server. These cluster heads aggregate and preprocess learning data from their member nodes before transmitting to the central server, reducing communication overhead and system complexity while preserving the federated learning approach for reliability.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If network nodes with different hardware and software configurations are included, then network coverage and adaptability are improved, but model update accuracy deteriorates due to inappropriate comparisons between dissimilar nodes

Engineering Contradiction:
Improvenetwork adaptabilityVSAvoidmodel update accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by forming clusters of network nodes with similar hardware and software configurations. Within each cluster, nodes share comparable characteristics, ensuring that model updates and performance comparisons are accurate and meaningful. This allows the system to maintain network adaptability across diverse configurations while ensuring precision in model updates within homogeneous clusters.

Inventive Principle:
Principle #3Local quality

3Reliability

If real-time service assurance is implemented across the entire network, then problem detection and localization are improved, but latency increases due to the scale of the network

Engineering Contradiction:
Improveservice assuranceVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the network into clusters with local cluster heads that perform real-time monitoring and analysis within their respective clusters. This segmentation enables localized problem detection and rapid response without requiring communication across the entire network, reducing latency while maintaining comprehensive service assurance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements feedback mechanisms at the cluster level where cluster heads continuously monitor node performance and provide localized feedback to both member nodes and the central server. This feedback loop enables real-time service assurance with reduced latency, as problems are detected and addressed within local clusters before propagating to the central system.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4079028B1Configuring network nodes in communication network
Publication Date: 2026.02.18 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP4079028B1 patent drawingFigure 1
  • EP4079028B1 patent drawingFigure 2
  • EP4079028B1 patent drawingFigure 3

AI summary

A cluster computing device (80, 82) is defined, associated with a plurality of network nodes of the communications network, and it generates a performance model relating configuration parameters of a network node to at least one energy performance parameter of the network node. The cluster computing device transfers (192) the generated performance model to the plurality of network nodes, and then receives (194) information from network nodes, indicating configuration parameters and at least one energy performance parameter of the respective network node. When the cluster computing device (80, 82) identifies (196) a first network node as having a poor performance, it also identifies (198) a second network node as having a good performance, and it causes (200) the first network node to operate with cloned configuration parameters of the second network node; and