RBF Network Excursion Classification Using Hyper-Cubes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Radial Basis Function (RBF) networks in artificial neural systems face increasing errors with higher dimensions, leading to false negatives and false positives in distinguishing normal and abnormal system operations, particularly in sensor data analysis from semi-conductor processing equipment.
Innovation Solution
The solution involves creating a system that uses RBF networks in conjunction with hyper-cube and hyper-sphere analysis, where nodes are expanded by increasing their radii, and samples are classified based on their residence within these geometric structures, allowing for more accurate classification with confidence estimation and minimizing errors by adding additional nodes along relevant axes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If RBF networks are used to analyze sensor data from semi-conductor processing equipment, then the system can differentiate between normal and abnormal operations, but errors increase with increasing numbers of dimensions (sensors)
Solution Approach 1:
The patent transforms the high-dimensional sensor data into a lower-dimensional feature space by extracting key features and patterns from the raw sensor readings. This dimensionality reduction allows the RBF network to maintain classification accuracy while reducing the impact of the curse of dimensionality, thereby resolving the contradiction between measurement precision and device complexity.
Solution Approach 2:
The patent extracts relevant features and patterns from the high-dimensional sensor data, separating the essential information from the redundant or noisy dimensions. By taking out only the critical features needed for classification, the system maintains accuracy while reducing the effective dimensionality that causes error increase in RBF networks.
2Ease of operation
If RBF networks differentiate only between normal and abnormal values, then the analysis is simple, but false negatives increase and reduce reliability
Solution Approach 1:
The patent segments the classification task into multiple stages: initial normal/abnormal differentiation by the RBF network, followed by secondary verification steps and confidence assessment. This segmentation allows the system to maintain operational simplicity while reducing false negatives through multi-layered validation, thereby improving reliability without sacrificing ease of operation.
Solution Approach 2:
The patent implements feedback mechanisms where classification results are continuously evaluated and refined. False negatives are detected through confidence scoring and verification processes, with the system learning from these errors to improve future classifications. This feedback loop maintains simplicity while significantly reducing false negative rates.
Data Source
AI summary
A method and system for analysis of data, including creating a first node, determining a first hyper-cube for the first node, determining whether a sample resides within the first hyper-cube. If the sample does not reside within the first hyper-cube, the method includes determining whether the sample resides within a first hyper-sphere, wherein the first hyper-sphere has a radius equal to a diagonal of the first hyper-cube.