Key-Pattern-Extraction for Training Set Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current classification techniques for pattern recognition, such as k-Nearest Neighbor and Radial Basis Function networks, face significant computational challenges with increasing training pattern sizes, leading to prohibitively long running times, inconsistency in decision boundaries, and suboptimal training set sizes, limiting their applicability in large databases and online diagnostics.
Innovation Solution
The Key-Pattern-Extraction technique reduces the training set size by identifying 'key patterns' that substantially affect the decision boundary, using R-function-based representations to compute Euclidean distances and iteratively remove non-key patterns, preserving the decision boundary and reducing computation time to O(dn^3) operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the training set size is reduced to decrease computational time, then productivity is improved, but measurement precision deteriorates due to bias and inconsistency in decision boundaries
Solution Approach 1:
The patent extracts only the essential training patterns (key patterns) that are necessary to define the decision boundary. By identifying and removing redundant patterns through iterative application of deletion rules, the method extracts a minimal subset that preserves decision boundary consistency while reducing computational burden.
Solution Approach 2:
The patent applies deletion rules in preliminary iterations to identify and remove redundant patterns before final classification. By pre-processing the training set to eliminate unnecessary patterns, the method prepares an optimized training subset that maintains accuracy while reducing computational time for the actual classification task.
2Productivity
If iterative deletion rules are applied to reduce training set size, then productivity is improved, but loss of information increases due to removal of potentially important patterns
Solution Approach 1:
The patent employs feedback mechanisms where each deletion rule application is evaluated based on its impact on classification performance. The iterative process continuously monitors whether pattern removal degrades decision boundary consistency, and adjusts the deletion process accordingly to prevent loss of critical information while removing redundant patterns.
Solution Approach 2:
The patent applies deletion rules iteratively but stops before excessive removal occurs. By controlling the number of iterations and monitoring decision boundary consistency, the method removes sufficient redundant patterns to improve efficiency while maintaining enough training patterns to preserve all necessary information for accurate classification.
3Measurement precision
If complex techniques like Voronoi-editing are used to preserve decision boundary consistency, then measurement precision is improved, but device complexity increases due to high computational requirements
Solution Approach 1:
The patent segments the complex problem of decision boundary preservation into simpler deletion rules that can be applied iteratively. Instead of computing complex Voronoi diagrams, the method divides the task into multiple simple pattern removal steps, each evaluating a specific deletion criterion, thereby reducing computational complexity while maintaining boundary consistency.
Solution Approach 2:
The patent uses simple, computationally inexpensive deletion rules that can be applied quickly and discarded after each iteration. Rather than maintaining complex data structures like Voronoi diagrams throughout the process, the method uses lightweight, disposable evaluation criteria that are computed, applied, and discarded in each iteration, reducing overall computational complexity.
Data Source
AI summary
A system that reduces the size of a design data set. During this design data set reduction operation, the system computes a decision boundary which separates a first group of data patterns in a training data set from a second group of data patterns in the training data set. For each data pattern in the training data set, the system determines if removing the data pattern from the training data set substantially affects the resulting decision boundary. If so, the system marks the data pattern as a key pattern. The system then removes all data patterns that are not marked as key patterns to produce a reduced training data set which represents the decision boundary.