Concurrent ML Function Coordination via Mediated Parameter Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network automation systems face challenges in coordinating concurrent machine learning (ML) functions due to limited or no direct communication between these functions, leading to potential conflicts and suboptimal network performance.
Innovation Solution
A distributed coordination mechanism allows ML functions to learn and optimize their actions based on the history of parameter updates and strategies of other functions, without direct communication, by using a mediation function to facilitate parameter value updates and notifications, ensuring minimal impact on concurrent functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If ML functions operate independently without direct communication, then device complexity is reduced, but coordination between functions deteriorates leading to conflicts
Solution Approach 1:
The patent introduces a mediation function that acts as an intermediary between ML functions. This mediator collects parameter update information from one ML function and communicates it to other ML functions, enabling indirect coordination without requiring direct communication channels between all ML function pairs, thus maintaining low complexity while improving coordination reliability
Solution Approach 2:
The patent implements a feedback mechanism where ML functions receive notifications about parameter updates made by other functions. This feedback allows each ML function to adjust its behavior based on the actions of others, achieving coordination through iterative learning and adaptation without complex direct communication infrastructure
2Reliability
If ML functions directly communicate to coordinate actions, then coordination reliability improves, but device complexity increases
Solution Approach 1:
The mediation function serves as a centralized intermediary that handles all communication between ML functions. Instead of requiring direct peer-to-peer communication channels between all ML function pairs (which would create O(n²) complexity), the mediator centralizes communication through a single node, reducing infrastructure complexity while maintaining coordination reliability
3Productivity
If ML functions optimize parameters independently, then productivity of individual functions increases, but network performance deteriorates due to conflicts
Solution Approach 1:
The patent enables ML functions to maintain high optimization productivity by allowing independent operation, while simultaneously improving network performance through feedback mechanisms. Each ML function continues to optimize parameters independently at high speed, but receives feedback about other functions' actions and adjusts its optimization strategy accordingly, preventing conflicts and improving overall network performance
Data Source
AI summary
Method comprising:receiving an update information on an updated value of a parameter of a network function and an identifier of a managing function responsible for the updating of the value of the parameter;defining a favorable range of values of the parameter based on the update information and a history of previous values of the parameter, wherein the history comprises, for each of the previous values of the parameter, the identifier of a respective managing function responsible for updating the value of the parameter to the respective previous value;calculating a new value of the parameter by optimizing a utility function and taking the favorable range as a constraint for the new value;updating the value of the parameter of the network function to the new value.