Edge Data Gathering Using Reinforcement Learning Divergence Thresholds

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies face challenges in efficiently managing machine learning models in edge networks, particularly in updating divergence threshold values to maximize model quality while minimizing computational overhead.

Innovation Solution

The system employs a reinforcement learning-based divergence model to dynamically adjust the divergence threshold value based on model metrics, resource availability, and clique configurations, using a graph neural network to identify distribution cliques and select representative edge nodes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the divergence threshold value is updated frequently to maximize model quality, then the model quality is improved, but the computational overhead and data transmission requirements increase

Engineering Contradiction:

Improvemodel qualityVSAvoidcomputational overhead

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The divergence threshold value is made dynamic through reinforcement learning, allowing it to adapt automatically based on current system conditions and model performance. The RL agent continuously learns optimal threshold values that balance model quality improvement with computational resource consumption, resolving the contradiction between frequent updates for better accuracy and reduced computational overhead

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of divergence threshold value from a static configuration to a dynamically learned value. By using reinforcement learning to optimize this parameter, the system achieves adaptive control that maximizes model quality while minimizing the computational resources required for data gathering and model updates

Inventive Principle:

Principle #35Parameter changes

2Reliability

If more edge nodes are selected for data gathering to improve model accuracy, then the model quality is improved, but the data transmission cost and computational overhead increase

Engineering Contradiction:

Improvemodel accuracyVSAvoiddata transmission cost

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The reinforcement learning agent receives feedback from the central ML model's performance metrics and uses this information to adjust the divergence threshold. This feedback mechanism enables the system to select the optimal number and identity of edge nodes for data gathering, achieving sufficient model accuracy without excessive data transmission by learning from past performance

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs self-optimization through reinforcement learning, where the divergence threshold is automatically adjusted based on model performance without manual intervention. The RL agent autonomously determines the optimal data gathering strategy, selecting edge nodes that provide the most valuable information for model accuracy while minimizing data transmission requirements

Inventive Principle:

Principle #25Self-service

3Device complexity

If a static divergence threshold is used to simplify the system, then the device complexity is reduced, but the adaptability to changing network conditions and model requirements deteriorates

Engineering Contradiction:

Improvesystem simplicityVSAvoidadaptability to changing conditions

Core Design Contradiction:

Device complexityVSAdaptability or versatility

Solution Approach 1:

The system replaces the mechanical approach of manually setting and adjusting divergence thresholds with an automated reinforcement learning mechanism. The RL agent substitutes human intervention with an autonomous learning system that adapts to changing network conditions and model requirements, achieving high adaptability while keeping the operational interface simple

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The reinforcement learning agent acts as an intermediary between the static system components and the dynamic requirements. It mediates between the simplicity of a fixed threshold approach and the need for adaptability, translating complex changing conditions into optimized threshold values through learned policies

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20250045592A1Edge data gathering using reinforcement learning and distribution cliques

Publication Date: 2025.02.06 EMC IP HLDG CO LLC

AI summary

Techniques are disclosed for edge node data gathering. One example method includes receiving probability distributions from edge nodes; using the probability distributions to identify a set of distribution cliques of the edge nodes; selecting one or more representative edge nodes from each clique; receiving feature data from the edge nodes, the feature data comprising resource information that includes a resource availability and a utilization status of the edge node at a first time, t−1; training a ML-based model using a portion of the feature data; associating the feature data with the corresponding clique for the edge node at the first time; using the probability distributions, cliques, and feature data to obtain episode data for each clique for the first time; and training a ML-based divergence model using a portion of the episode data to update a divergence threshold value for the clique for a second time, t.