Method and system for scheduling transmission of update information
The method and system address network congestion in distributed machine learning by predicting update completion times and adjusting transmission schedules based on architectural parameters and communication modes, effectively managing bursty traffic.
WO2026132207A1PCT designated stage Publication Date: 2026-06-25KONINK KPN NV
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- KONINK KPN NV
- Filing Date
- 2025-12-18
- Publication Date
- 2026-06-25
AI Technical Summary
Technical Problem
Existing scheduling systems for distributed machine learning models fail to account for multiple concurrent trainings, leading to network congestion due to bursty transmissions, which is not effectively managed.
Method used
A method and system that predicts the timing of update information completion at worker nodes, determines the risk of temporal overlap, and generates scheduling information to modify transmission times, using architectural parameters and communication modes to avoid congestion.
Benefits of technology
Reduces the risk of network congestion by orchestrating transmissions efficiently, ensuring smooth communication across multiple worker nodes participating in distributed training.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure EP2025087957_25062026_PF_FP_ABST
Abstract
Some embodiments are directed to scheduling of transmissions of update information from a plurality of worker nodes participating in distributed training of one or more machine learning models A risk of temporal overlap between update information transmissions is determined by predicting, using a prediction model, and based on architectural parameters, a timing of when the update information is complete for each worker node, is obtained.
Need to check novelty before this filing date? Find Prior Art