Federated Learning Mode Selection for Limited-Capability User Equipment

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Federated learning is often ineffective due to insufficient computing or communication capabilities of user equipment, leading to inefficient model training.

Innovation Solution

An electronic apparatus determines whether user equipment should participate in federated learning independently or dependently based on self-state information, enabling efficient training by splitting models between user equipment and assistant equipment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If user equipment trains the complete model independently in federated learning, then model training can be performed without cooperation, but the computational burden on user equipment becomes too heavy when computing capability is insufficient

Engineering Contradiction:
Improvemodel training effectivenessVSAvoidcomputational burden on user equipment
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent divides the complete model into multiple sub-models and assigns different sub-models to different user equipment for parallel training. This segmentation reduces the computational burden on each individual device while maintaining overall training effectiveness through coordinated optimization of all sub-models.

Inventive Principle:
Principle #1Segmentation

2Reliability

If user equipment with insufficient computing capability participates in federated learning independently, then data privacy is maintained, but the federated learning cannot be performed effectively

Engineering Contradiction:
Improvefederated learning effectivenessVSAvoidadaptability to user equipment with limited capabilities
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

By segmenting the model into sub-models, the patent enables user equipment with limited capabilities to participate effectively in federated learning. Each device trains only its assigned sub-model, making the system adaptable to diverse hardware capabilities while preserving data privacy through distributed training.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent assigns different sub-models to different user equipment based on their specific computing capabilities and data characteristics. This local quality approach ensures that each device operates within its capability constraints while contributing to the overall federated learning effectiveness.

Inventive Principle:
Principle #3Local quality

3Use of energy by moving object

If the model is split into sub-models for cooperative training, then computational burden is reduced on individual devices, but the system complexity increases

Engineering Contradiction:
Improvecomputational efficiency on user equipmentVSAvoidsystem complexity for model splitting and coordination
Core Design Contradiction:
Use of energy by moving objectVSDevice complexity

Solution Approach 1:

The patent introduces a server as an intermediary that manages the model splitting, sub-model assignment, and training coordination. This intermediary handles the complex tasks of dividing the model and coordinating the training process, reducing the complexity burden on individual user equipment while enabling efficient distributed training.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4675505A1Electronic device and method for wireless communication, and computer-readable storage medium
Publication Date: 2026.01.07 SONY GROUP CORP
  • EP4675505A1 patent drawingFigure 1~2
  • EP4675505A1 patent drawingFigure 3
  • EP4675505A1 patent drawingFigure 4

AI summary

The present application relates to an electronic device and a method for wireless communication, and a computer-readable storage medium. The electronic device for wireless communication comprises a processing circuit, wherein the processing circuit is configured to determine, according to state information reported by a user equipment within a service range of the electronic device, whether the user equipment is to cooperate with other user equipments to train a segmented model obtained by segmenting a model to be trained, so as to participate in federated learning in a non-independent way, or to independently train the model to be trained, so as to independently participate in federated learning. (FIG. 1)