Outlier Detection via Subspace Distance Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current outlier detection methods, such as sparse representation based methods, are not suited for real-time applications due to their high complexity.
Innovation Solution
A system and method for outlier detection that calculates distances from an input data point to multiple subspaces, selects a minimum distance, normalizes it using remaining distances, and compares it to a threshold value to determine if it's an outlier, with low algorithmic complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sparse representation based methods (SCI, GPD) are used for outlier detection, then detection capability is improved, but algorithmic complexity increases significantly
Solution Approach 1:
The patent segments the feature space into multiple subspaces, each spanned by a subset of features. Instead of using complex global sparse representation methods, the algorithm calculates distances to these segmented subspaces independently, significantly reducing computational complexity while maintaining detection effectiveness
Solution Approach 2:
The patent uses partial action by considering only the minimum distance from the data point to the subspaces rather than computing full sparse representations. This partial approach (using only essential distance information) achieves effective outlier detection without the excessive computational burden of complete sparse representation methods
2Measurement precision
If complex outlier detection methods are applied, then detection accuracy improves, but real-time application capability deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-computing the subspaces during an offline training phase using historical data. During real-time inference, the algorithm only needs to calculate distances to these pre-computed subspaces, which are stored in memory, enabling fast real-time detection without the computational burden of complex methods
Solution Approach 2:
The patent replaces complex mechanical computation systems (full sparse representation algorithms) with a simpler geometric approach (distance calculation to subspaces). This substitution maintains detection accuracy while dramatically improving processing speed for real-time applications
Data Source
AI summary
A method of outlier detection includes steps as follows. Distances from an input data point to a plurality of subspaces respectively are calculated. A minimum distance is selected from the distances to leave one or more remaining distances. The one or more remaining distances are utilized to normalize the minimum distance to obtain the normalized distance value. Whether the normalized distance value is greater than a threshold value is detected, so as to output a detection result.