Training Coordination for Context-Aware CNF Retraining

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing Cognitive Autonomous Networks (CANs) face challenges in efficiently coordinating the training of multiple Cognitive Network Functions (CNFs) due to changes in context, leading to potential conflicts and inefficiencies in resource utilization, as existing systems lack the ability to manage Machine Learning Algorithm (MLA) training requirements and dependencies across multiple CNF instances.

Innovation Solution

A Training Coordination Function (TCF) is introduced to plan, execute, and optimize CNF training requests, utilizing CNF Instance Metadata (CNF-IMD) and policy rules to coordinate retraining needs, ensuring efficient use of network resources and minimizing conflicts by batch processing and model reuse.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If MLAs are re-trained when context changes to maintain model accuracy, then model accuracy is improved, but training resource consumption and time increase

Engineering Contradiction:
Improvemodel accuracyVSAvoidtraining time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The TCF proactively identifies context changes and triggers retraining before model accuracy degrades significantly. By monitoring context parameters and detecting changes early, the system prepares and executes retraining operations in advance, preventing performance degradation while optimizing the timing of resource-intensive training operations

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system dynamically adjusts retraining parameters such as training frequency, data sampling rates, and model architecture based on the detected context change magnitude and type. This allows the system to maintain model accuracy while adapting training resource allocation to the actual needs of each context change scenario

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If multiple CNFs are trained simultaneously to improve network intelligence, then network intelligence is improved, but resource conflicts and training coordination complexity increase

Engineering Contradiction:
Improvenetwork intelligenceVSAvoidtraining coordination complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The TCF serves as an intermediary layer between multiple CNFs and the underlying training infrastructure. It centralizes the coordination logic, managing training requests, allocating resources, and resolving conflicts between concurrent training operations. This mediator approach simplifies the overall system architecture by consolidating coordination complexity in a single management point rather than distributing it across multiple CNFs

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The TCF dynamically adjusts training schedules, resource allocation, and priority levels based on real-time system state, context changes, and training progress. This dynamic coordination allows multiple CNFs to be trained efficiently by adaptively responding to changing conditions, optimizing resource utilization while maintaining network intelligence across multiple functions

Inventive Principle:
Principle #15Dynamics

3Adaptability or versatility

If MLA training is performed frequently to adapt to context changes, then adaptability is improved, but resource consumption and network performance degradation increase

Engineering Contradiction:
Improveadaptability to context changesVSAvoidnetwork performance
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The system implements periodic retraining operations triggered by detected context changes rather than continuous training. The TCF monitors context parameters and initiates training only when significant changes are detected, creating a periodic training rhythm that balances adaptability with network performance. This approach avoids unnecessary training operations that would degrade network performance while still maintaining model accuracy through timely retraining

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The TCF employs partial retraining strategies where only specific model components or parameters are updated based on the nature and scope of context changes, rather than performing complete model retraining. This selective approach maintains adaptability to context changes while reducing the computational burden and network performance impact compared to full retraining operations

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4268395B1Context-aware training coordination in communication network systems
Publication Date: 2026.03.04 NOKIA TECHNOLOGIES OY
  • EP4268395B1 patent drawingFigure 1
  • EP4268395B1 patent drawingFigure 2
  • EP4268395B1 patent drawingFigure 3

AI summary

According to a method for implementing a training coordination function, at least one training request with which training of a network function instance of an autonomous network of a communication network system is requested is processed. Based on a result of the processing, it is decided whether to approve the at least one training request, or not. If the at least one training request is approved, the training is planned based on the at least one training request, an identification of the at least one training request is returned, and in accordance with the planning, an existing training state associated with the network function instance is stored or updated in a database which is configured to store existing training states associated with network function instances of the autonomous network.