Dimensionality Reduction Clustering for Similarity Sample Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing similarity analysis methods face challenges with high computational complexity and reduced prediction accuracy due to increased dimensions of feature vectors, and they often result in low diversity of selected similarity samples.
Innovation Solution
The proposed solution involves performing multiple dimensionality reduction processes in parallel, followed by a clustering operation on the intermediate sets of similarity samples, to generate an output set with improved diversity and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If dimensionality reduction is applied to reduce computational complexity, then computational complexity is reduced, but prediction accuracy deteriorates due to information loss
Solution Approach 1:
The patent divides the dimensionality reduction process into multiple independent steps, each applying a different reduction technique to create multiple intermediate feature sets. This segmentation allows the system to balance computational complexity reduction with accuracy preservation by not relying on a single aggressive reduction method, thereby resolving the contradiction between reducing complexity and maintaining prediction accuracy.
Solution Approach 2:
The patent changes the parameters of dimensionality reduction by applying multiple different reduction techniques with varying degrees of aggression. Instead of using one fixed reduction method, the system varies the reduction parameters across multiple intermediate representations, enabling optimal balance between computational complexity and prediction accuracy for different application scenarios.
2Productivity
If traditional similarity analysis is used with high-dimensional feature vectors, then analysis can be performed, but computational complexity becomes intractable
Solution Approach 1:
The patent segments the high-dimensional similarity analysis into multiple lower-dimensional sub-problems by applying dimensionality reduction techniques. This divides the complex computational task into manageable intermediate steps, making the analysis tractable while preserving the ability to perform comprehensive similarity analysis across the entire feature space.
Solution Approach 2:
The patent transforms the problem from high-dimensional space to multiple lower-dimensional spaces through dimensionality reduction. By changing the dimensional representation of the data, the system makes computational operations feasible while maintaining the essential information needed for similarity analysis, thus resolving the intractability issue.
3Adaptability or versatility
If clustering is performed on base samples with large dimensions, then sample diversity can be improved, but computational complexity becomes intractable
Solution Approach 1:
The patent segments the clustering process by first performing dimensionality reduction to create multiple intermediate feature sets, then applying clustering to these reduced-dimensional representations. This segmentation makes the clustering computationally feasible while still achieving diverse sample selection, as each intermediate representation captures different aspects of the data structure.
Solution Approach 2:
The patent changes the dimensional space in which clustering is performed by reducing the dimensionality of base samples before clustering. This dimensionality transformation enables clustering operations that would be intractable in the original high-dimensional space, while maintaining the ability to select diverse samples through the multiple intermediate representations.
Data Source
AI summary
Systems, methods, and software can be used to determine similarity samples. In some aspects, a method includes: obtaining a first feature vector of a sample; processing the first feature vector through a plurality of dimensionality reduction processes, wherein each of the plurality of dimensionality reduction processes generates a respective second feature vector, each of the second feature vectors has a smaller dimension than a dimension of the first feature vector; for each of the second feature vectors, determining an intermediate set of similarity samples; generating a plurality of clusters based on the intermediate sets of similarity samples; and generating an output set of similarity samples based on the plurality of clusters.