Incremental Centroid Clustering for Edge Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning algorithms for anomaly detection, such as One Class SVM and Isolation Forest, require significant storage and processing resources that are not available on edge devices, and cloud-based solutions incur high communication costs and latency, making them impractical for distributed systems.
Innovation Solution
The implementation of incremental centroid clustering, which stores only cluster information and not the entire dataset, allowing for efficient anomaly detection on edge devices with limited resources by incrementally adding data points to existing clusters or creating new ones based on distance thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning algorithms (One Class SVM, Isolation Forest) are used for anomaly detection, then detection accuracy is improved, but processing and memory requirements increase significantly
Solution Approach 1:
The patent extracts only the essential cluster centroid information from the full dataset, storing merely the centroid coordinates and cluster assignments rather than the complete data points. This extraction principle reduces memory requirements from storing all data points to storing only compact cluster representations, while maintaining anomaly detection capability through distance-based classification.
Solution Approach 2:
The patent creates simplified copies of the data in the form of cluster centroids - condensed representations that capture the essential structure of the data distribution. These centroid copies serve as proxies for the full dataset, enabling anomaly detection without requiring storage of the original large-scale data points.
2Adaptability or versatility
If traditional machine learning algorithms are deployed on edge devices, then local anomaly detection capability is improved, but device resource constraints are violated
Solution Approach 1:
The patent employs computationally inexpensive operations for incremental cluster updates, using simple distance calculations and basic arithmetic operations rather than complex machine learning computations. These lightweight operations can be executed on resource-constrained edge devices without requiring significant processing power, making the system suitable for deployment in embedded and IoT environments.
3Measurement precision
If cloud-based anomaly detection is used, then detection capability is improved, but communication costs and latency increase
Solution Approach 1:
The patent performs preliminary clustering and centroid computation during an offline training phase, preparing the anomaly detection model in advance. This preliminary action enables the edge device to perform only lightweight incremental updates and distance-based anomaly classification during online operation, eliminating the need for continuous cloud communication and reducing latency to minimal local processing delays.
Data Source
AI summary
A method at a computing device for anomaly detection, the method including storing, at the computing device, a full batch model for a plurality of data points, wherein each of the plurality of data points is associated with one of a plurality of clusters, and wherein the storing comprises a table for storing information about the plurality of clusters without storing the plurality of data points; receiving a new data point from a hardware sensor; determining that the new data point falls outside the full batch model, thereby detecting an anomaly; and performing an action based on the anomaly.