Monitoring System Disturbance Indication for ML Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Monitoring systems face challenges in accurately tracking the operation of data processing systems due to disturbances caused by assistance requests, which can affect the training and inference operations of machine learning models, leading to less accurate results.
Innovation Solution
A monitoring system that includes a processor providing a disturbance indication to a machine learning model when assistance requests are made, allowing the model to adjust for disturbances in output data, thereby improving accuracy during both training and inference phases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the monitoring system provides assistance requests to the monitored system, then the monitoring capability is improved, but the output data becomes disturbed and ML model accuracy deteriorates
Solution Approach 1:
The monitoring system applies preliminary anti-action by providing a disturbance indication signal to the ML model before the assistance request fully impacts the output data. This advance warning allows the ML model to compensate for the expected distortion, maintaining accuracy while enabling the monitoring system to request necessary assistance from the monitored system.
2Ease of operation
If assistance requests are provided to the monitored system, then the monitoring system can obtain necessary information, but the output data distortion increases and affects training and inference operations
Solution Approach 1:
The system implements feedback by providing a disturbance indication signal from the monitoring system to the ML model. This feedback loop communicates when assistance requests will cause distortion, allowing the ML model to adjust its processing accordingly. This enables the monitoring system to access necessary information while the ML model compensates for the resulting data distortion.
Data Source
AI summary
A data processing system includes a monitoring system, the monitoring system includes a processor and a data analysis block. The processor executes a monitoring application for monitoring an operation of a monitored system coupled to the monitoring system. When assistance is needed from the monitored system, the processor has an output coupled to the monitored system for providing an assistance request. When the assistance request is sent to the monitored system, the processor also sends a disturbance indication to the data analysis block. The disturbance indication indicates that the output data from the monitored system may be disturbed by the assistance request. The data analysis block can then take an action to reduce the effect the disturbance may have on the analysis results. A method for monitoring the monitored system is also provided.