AI Model Training Frequency Control for Accuracy and Power Balance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Variations in the accuracy of machine learning models trained in AI devices due to changes in external environments or saturation lead to unreliable outputs and inefficient power consumption in smart city applications.
Innovation Solution
A model management device that communicates with multiple AI devices to manage training frequency based on model accuracy, adjusting the execution frequency of training processes to maintain accuracy within predetermined thresholds, thereby reducing variations and optimizing power usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are continuously trained in each AI device, then model accuracy may be improved, but power consumption increases and accuracy variations occur
Solution Approach 1:
The patent implements dynamic adjustment of training execution frequency based on real-time model accuracy monitoring. The management device changes the training frequency adaptively - increasing it when accuracy drops below thresholds and decreasing it when accuracy is sufficient - thereby optimizing the balance between maintaining model accuracy and reducing power consumption in AI devices
Solution Approach 2:
The system establishes a feedback loop where the management device continuously monitors model accuracy from multiple AI devices and uses this information to adjust training frequencies. This feedback mechanism enables the system to respond to accuracy variations and external environment changes, optimizing power consumption while maintaining required accuracy levels
2Reliability
If training frequency is increased to improve model accuracy, then reliability of output values improves, but power consumption increases
Solution Approach 1:
The patent applies dynamic frequency adjustment to match training intensity with actual reliability requirements. By monitoring model accuracy and adjusting training frequency accordingly, the system ensures reliable output values only when necessary, avoiding excessive power consumption during periods when model performance is already sufficient
3Stability of the object's composition
If continuous training is performed at high frequency, then model accuracy is maintained, but variations in accuracy across different AI devices occur
Solution Approach 1:
The patent implements a centralized management device that coordinates training across multiple AI devices, applying universal accuracy thresholds and frequency adjustment rules to all devices. This universal approach ensures consistent accuracy maintenance across different devices while adapting to individual device conditions, reducing accuracy variations among the plurality of AI devices
Data Source
AI summary
The model management device includes a communication unit capable of communicating with a plurality of AI devices, each of which continuously performs training of a machine learning model, and a processor configured to manage training of a machine learning model in each of the plurality of AI devices, and acquire accuracy of the machine learning model trained in each of the plurality of AI devices. The processor is configured to change an execution frequency of a process relating to training of the machine learning model having an accuracy within a predetermined range.