One-Class Similarity Machine for Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
One-class support vector machines (SVMs) require a full labeled training set and cannot handle changes in the problem domain, making them inefficient for anomaly detection, especially when only one class of data is available.
Innovation Solution
A system using one-class similarity machines computes similarity scores based on unlabeled training instances and determines boundary thresholds to classify testing instances as anomalies, enhancing data mining and outlier detection by learning from a single class of data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one-class support vector machines (SVMs) are used for anomaly detection, then classification accuracy can be achieved, but the system requires a full labeled training set and cannot handle changes in the problem domain, resulting in inefficiency
Solution Approach 1:
The patent transitions from using labeled training data to unlabeled training data, fundamentally changing the data parameter. The system learns from unlabeled data to establish normal patterns, then detects anomalies by identifying deviations from these patterns, eliminating the need for expensive labeled data while maintaining detection accuracy
Solution Approach 2:
The patent implements a dynamic anomaly detection system that can adapt to changes in the problem domain. The system continuously learns from new unlabeled data and updates its understanding of normal patterns, allowing it to handle domain changes without requiring re-labeling of training data
2Adaptability or versatility
If one-class support vector machines (SVMs) are used for anomaly detection, then classification capability is provided, but the requirement for full labeled training sets increases system complexity and reduces ease of operation
Solution Approach 1:
The system performs self-learning by automatically establishing normal patterns from unlabeled data without requiring manual labeling. The anomaly detection mechanism automatically identifies deviations from learned patterns, eliminating the need for operators to prepare and manage labeled training datasets
3Reliability
If one-class support vector machines (SVMs) are used for anomaly detection, then classification is achieved, but the inability to handle changes in the problem domain reduces adaptability
Solution Approach 1:
The patent creates a dynamic system that continuously adapts to domain changes by learning from new unlabeled data. The system updates its internal representation of normal patterns over time, allowing it to maintain reliable anomaly detection even when the problem domain evolves or changes
Data Source
AI summary
One embodiment provides a system for facilitating anomaly detection. During operation, the system determines, by a computing device, a set of training instances, wherein a training instance represents a single class of data within a predefined range. The system computes a similarity score for each testing instance in a set of testing instances, wherein the similarity score is based on a similarity function which takes as input a respective testing instance and the set of training instances. The system determines a boundary threshold based on an ordering of the similarity score for each testing instance. The system classifies a first testing instance as an anomaly responsive to determining that the first testing instance falls outside the boundary threshold, thereby enhancing data mining and outlier detection in the single class of data using unlabeled training instances.