Cognitive Function Inference Using Data Quality and Confidence Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cognitive autonomous networks (CAN) face performance degradation due to erroneous and inconsistent data inputs, which affect the accuracy of cognitive functions (CFs) and the optimal configuration calculations, leading to suboptimal network performance.
Innovation Solution
Implementing a system that communicates Value Quality Scores (VQS) and Confidence Scores (CCS) to indicate the quality and confidence of data inputs, and using these scores to filter and weight the input data during training and inference, ensuring robustness against erroneous data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If cognitive functions use erroneous network data for learning and inference, then the system operates autonomously without manual intervention, but the accuracy of optimal configuration calculations deteriorates
Solution Approach 1:
The patent applies preliminary action by training cognitive functions with imperfect data before deployment. The training process prepares the CFs to handle erroneous inputs by exposing them to such data during the learning phase, enabling them to robustly infer optimal configurations even when runtime data contains errors.
Solution Approach 2:
The patent implements feedback mechanisms where the system evaluates the quality of data inputs and adjusts the learning process accordingly. By monitoring data quality metrics and providing feedback to the training algorithm, the system continuously improves its ability to handle erroneous data and refine its configuration calculations.
2Speed
If cognitive functions process all incoming data without verification, then processing speed is maintained, but reliability of network configuration decisions deteriorates
Solution Approach 1:
The patent applies local quality by implementing data quality assessment at specific critical points in the data processing pipeline rather than uniformly across all data. The system evaluates quality metrics for particular data inputs and applies selective verification only where necessary, maintaining processing speed while improving reliability of configuration decisions.
3Reliability
If the system implements comprehensive data verification mechanisms, then data quality improves, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary data quality assessment layer between raw data inputs and the cognitive functions. This intermediary component evaluates data quality metrics and provides filtered, quality-assessed data to the CFs without requiring complex verification mechanisms throughout the entire system, thus improving data quality while limiting complexity increase.
4Measurement precision
If cognitive functions are trained only with perfect data, then initial accuracy is high, but adaptability to real-world erroneous data deteriorates
Solution Approach 1:
The patent applies dynamics by making the training data composition flexible and adaptable. The system dynamically adjusts the mix of perfect and imperfect training data, and can adapt the training process based on the actual quality of runtime data. This enables the CFs to maintain high initial accuracy while developing adaptability to handle erroneous real-world data through continuous learning.
Data Source
AI summary
It is provided a method comprising: receiving one or more received data sets, wherein each of the received data sets comprises data each representing a value of a respective status parameter of a system, and at least one of the data sets comprises a respective value quality score representing a quality of the value of the respective status parameter: calculating, by a cognitive function, an optimal configuration range set and a confidence score of the optimal configuration range set based on the one or more received data sets; providing the calculated optimal configuration range set and the confidence score to a controller.