Task-Inference Association Unit for ML Model Data Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning model operation management systems lack the capability to efficiently associate tasks with inference data, leading to manual and costly processes for identifying problematic data, especially when issues arise with model accuracy or inference results.
Innovation Solution
A system that includes a task-inference association unit, which calculates associations between system data streams and recorded data streams based on configuration information representing dependencies between the system and the machine learning model, enabling rapid extraction of problematic data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual data association is performed to identify problematic data, then association between tasks and inference data can be established, but immense cost and time are incurred
Solution Approach 1:
The system performs preliminary action by pre-establishing association rules between tasks and inference data streams before problems occur. Configuration information defining these associations is prepared in advance, enabling automatic matching when issues arise, thus eliminating the need for manual data association and significantly reducing identification time.
2Reliability
If all related data is collected and checked to identify problematic data, then comprehensive analysis can be performed, but immense cost is incurred
Solution Approach 1:
The system extracts only the necessary association rules and configuration information needed for problem identification, rather than processing all related data. By taking out only the essential association metadata, the system achieves comprehensive problem analysis while minimizing data processing costs and resource consumption.
3Measurement precision
If 1:1 data association is performed on enormous data, then precise task-inference mapping can be achieved, but manual effort and cost increase
Solution Approach 1:
The system implements self-service by automatically performing task-inference data association using pre-configured association rules. The association mechanism serves itself by autonomously matching inference data to tasks based on configuration information, eliminating manual effort while maintaining precise mapping through systematic rule-based association.
Data Source
AI summary
A system for performing operation management of a machine learning model includes a task-inference ID association unit configured to, on the basis of a system data stream for outputting content of a task performed by a system that calls a machine learning model and a recorded data stream for inputting/outputting data when inference of the machine learning model is processed, calculate, with regard to association between the system data stream and the recorded data stream, the association between the system data stream and the recorded data stream from configuration information representing dependency between the system and the machine learning model.