Anomaly Detection Space Selection for Unlabelled Target Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Analyzing complex target systems like mobile communication networks and industrial processes poses a challenge in determining whether anomalies are best characterized in Euclidean or non-Euclidean space, especially when starting with unlabelled datasets, as existing methods lack automation in selecting appropriate anomaly detection methods.
Innovation Solution
A computer-implemented method that computes alignment scores using both linear and non-linear kernels to decide between Euclidean and non-Euclidean space measures for anomaly detection, selecting appropriate methods based on these scores, and applying centred kernel target alignment to maximize alignment scores relative to an unknown label-vector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If anomaly detection is performed using Euclidean space measures, then the detection process is simple and computationally efficient, but the detection accuracy deteriorates when anomalies are better characterized in non-Euclidean space
Solution Approach 1:
The patent changes the parameter of space representation by computing alignment scores for both linear (Euclidean) and non-linear (non-Euclidean) kernels, then selecting the appropriate space based on which score is higher. This dynamic parameter selection resolves the contradiction by adapting the detection approach to the specific characteristics of the dataset.
Solution Approach 2:
The patent introduces a dynamic selection mechanism that determines whether to use Euclidean or non-Euclidean space measures based on computed alignment scores. Rather than statically choosing one approach, the system dynamically adapts its detection methodology based on the data characteristics, thereby resolving the efficiency-accuracy trade-off.
2Measurement precision
If anomaly detection methods are manually selected based on domain expertise, then the detection accuracy improves, but the complexity of the system increases and automation is reduced
Solution Approach 1:
The patent enables the anomaly detection system to automatically select its own appropriate detection methods by computing alignment scores and determining whether Euclidean or non-Euclidean space measures are more suitable. This self-service capability eliminates the need for manual expert intervention while maintaining high detection accuracy.
Solution Approach 2:
The patent implements a feedback mechanism where alignment scores are computed and used to guide the selection of anomaly detection methods. This closed-loop approach allows the system to automatically adjust its methodology based on data characteristics, reducing complexity while preserving accuracy.
3Adaptability or versatility
If anomaly detection is performed on unlabelled datasets, then the applicability to real-world scenarios improves, but the difficulty of determining appropriate detection methods increases
Solution Approach 1:
The patent performs preliminary computation of alignment scores for both linear and non-linear kernels before selecting the anomaly detection method. This preliminary action on unlabelled data enables the system to determine the appropriate detection approach without requiring labelled examples, thereby maintaining versatility while reducing difficulty.
Data Source
AI summary
A computer implemented method for analyzing a target system for the purpose of controlling the target system. The method is performed by obtaining (301) a dataset comprising observations related to the target system; computing (302) alignment score for the dataset using a linear kernel to obtain a linear alignment score; computing (302) alignment score for the dataset using a non-linear kernel to obtain a non-linear alignment score; comparing (303) the linear alignment score and the non-linear alignment score; and if linear alignment score>non-linear alignment score, selecting (304) anomaly detection that uses Euclidean space measures, and else selecting anomaly detection that uses non-Euclidean space measures.