Parallelized Feature Selection for Variance Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data mining techniques face challenges in efficiently selecting explanatory or predictive features in large-scale data sets, particularly in parallelized computing environments, where redundant features increase data dimensionality and worsen model performance.
Innovation Solution
A parallelized computing system that accesses a set of features characterized by variance measures, determines components of the covariance matrix, and selects multiple features using a greedy search method executed with parallelized computation, minimizing redundant feature selection and achieving high-quality solutions in polynomial time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional feature selection methods are used on large-scale data sets, then feature selection can be performed, but the computational complexity and time consumption increase significantly
Solution Approach 1:
The patent divides the large set of features into multiple subsets or groups, allowing parallel processing of feature selection across different groups. This segmentation reduces the computational burden on single processors and enables distributed computing approaches, thereby decreasing overall computational time while maintaining selection accuracy through coordinated evaluation of all feature subsets.
Solution Approach 2:
The patent introduces a new dimension of parallelization by distributing feature selection computations across multiple processing nodes or cores simultaneously. By transforming the single-threaded sequential process into a multi-dimensional parallel architecture, the system achieves linear or near-linear speedup with respect to the number of processors, effectively reducing computational time without sacrificing accuracy.
2Loss of information
If all features are retained in the data set, then no information is lost, but data dimensionality increases and model performance deteriorates
Solution Approach 1:
The patent extracts and removes redundant or less informative features from the complete feature set through systematic evaluation metrics. By identifying and eliminating features that contribute minimally to predictive power or introduce noise, the method reduces data dimensionality and improves model performance while retaining the essential information contained in the most relevant features.
Solution Approach 2:
The patent employs parameter-based feature selection by evaluating features according to statistical parameters such as variance, information gain, or correlation coefficients. By changing the perspective from raw feature values to derived parameters that quantify feature quality, the system can objectively identify and retain only those features that significantly contribute to model performance, thereby reducing dimensionality without information loss.
3Loss of information
If redundant features are selected, then more comprehensive coverage is achieved, but data dimensionality increases and computational efficiency decreases
Solution Approach 1:
The patent replaces brute-force exhaustive feature evaluation with efficient computational algorithms that use mathematical approximations, heuristics, or iterative optimization methods. By substituting computationally intensive mechanical enumeration with smarter algorithmic approaches, the system achieves comprehensive feature coverage through selective evaluation rather than exhaustive analysis, thereby maintaining productivity and computational efficiency.
Data Source
AI summary
This disclosure describes a method, system and computer-program product for parallelized feature selection. The method, system and computer-program product may be used to access a first set of features, wherein the first set of features includes multiple features, wherein the features are characterized by a variance measure, and wherein accessing the first set of features includes using a computing system to access the features, determine components of a covariance matrix, the components of the covariance matrix indicating a covariance with respect to pairs of features in the first set, and select multiple features from the first set, wherein selecting is based on the determined components of the covariance matrix and an amount of the variance measure attributable to the selected multiple features, and wherein selecting the multiple features includes executing a greedy search performed using parallelized computation.