Nearest Neighbor Dataset Reduction via Iterative Sample Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computational expense and storage requirements of large databases in nearest neighbor techniques, such as k-nearest neighbor (KNN), become significant when classifying objects, especially in languages with thousands of characters, leading to inefficient classification processes.
Innovation Solution
A reduced dataset is generated by initializing an empty set and iteratively adding points based on wrongly classified samples, using techniques like averaging or random selection to generate representative points, which are then used for nearest neighbor searches, reducing the dataset size without sacrificing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large database with millions of samples is used for nearest neighbor classification, then classification accuracy is improved, but computational expense and storage requirements increase significantly
Solution Approach 1:
The patent extracts only the essential samples needed for accurate classification by identifying and removing redundant samples. It uses a filtering mechanism that evaluates each sample's contribution to classification accuracy and retains only those samples that provide unique classification value, thereby reducing database size while maintaining accuracy.
Solution Approach 2:
The patent changes the parameter of database size by transforming the complete sample database into a reduced subset. It applies transformation rules that modify the sample selection criteria, keeping only samples that meet specific diversity and representativeness thresholds, thus reducing storage and computational requirements while preserving classification performance.
2Reliability
If a large database with millions of samples is used for nearest neighbor classification, then classification accuracy is improved, but storage requirements increase significantly
Solution Approach 1:
The patent extracts only the essential samples needed for accurate classification by identifying and removing redundant samples. It uses a filtering mechanism that evaluates each sample's contribution to classification accuracy and retains only those samples that provide unique classification value, thereby reducing database size while maintaining accuracy.
Solution Approach 2:
The patent changes the parameter of database size by transforming the complete sample database into a reduced subset. It applies transformation rules that modify the sample selection criteria, keeping only samples that meet specific diversity and representativeness thresholds, thus reducing storage and computational requirements while preserving classification performance.
3Adaptability or versatility
If the number of characters in the alphabet is increased to support languages with thousands of characters, then versatility of the recognition system is improved, but both storage and computational requirements increase
Solution Approach 1:
The patent segments the large character set into multiple smaller groups or clusters based on visual similarity or feature characteristics. By organizing characters into segments, the system can perform nearest neighbor searches within smaller subsets rather than the entire database, significantly reducing search time while still supporting a comprehensive character set.
Solution Approach 2:
The patent changes the parameter of database size by transforming the complete sample database into a reduced subset. It applies transformation rules that modify the sample selection criteria, keeping only samples that meet specific diversity and representativeness thresholds, thus reducing storage and computational requirements while preserving classification performance.
Data Source
AI summary
A set S is initialized. Initially, S is empty; but, as the disclosed process is performed, items are added to it. It may contain one or more samples (e.g., items) from each class. One or more labeled samples for one or more classes may be obtained. A series of operations may be performed, iteratively, until a stopping criterion is reach to obtain the reduced set. For each class of the one or more classes, a point may be generated based on at least one sample in the class having a nearest neighbor in a set S with a different class label than the sample. The point may be added to the set S. The process may be repeated unless a stopping criterion is reached. A nearest neighbor for a submitted point in the set S may be identified and a candidate nearest neighbor may be output for the submitted point.