Efficient Unified Online Feature Selection via GMM Parameter Matrix and Binary Relevance-Redundancy Assessment

Read full text Summarise with VIRAT
Authors: Junyang Wu, Linjing You, jiabao lu
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This capsule provides the official Python implementation of **UOFS (Efficient Unified Online Feature Selection via GMM Parameter Matrix and Binary Relevance-Redundancy Assessment)**. Feature selection is crucial for processing high-dimensional visual data and continuous image streams in modern computer vision and image processing tasks. However, many existing methods are limited to handling specific types of streams. **UOFS** brings a unified framework designed to efficiently handle varying dynamic data and feature streams by leveraging Gaussian Mixture Model (GMM) parameter matrices and a novel binary relevance-redundancy assessment strategy. **Key Features & Capabilities:** * **Versatile Stream Support:** The codebase inherently simulates and evaluates four different feature selection scenarios common in dynamic visual environments: * Offline Feature Selection (`offline`) * Data Stream Feature Selection (`data`) * Feature Stream Feature Selection (`feature`) * Trapezoidal Data Stream Feature Selection (`trap`) * **Comprehensive Pipeline:** The experiment runner handles everything from data scaling (`StandardScaler`) and UOFS execution to training a Support Vector Machine (Linear SVM) to measure the effectiveness of the selected features on image classification tasks. * **Reproducibility:** Includes multiple benchmark datasets (e.g., USPS, DrivFace, ORL—widely used in face recognition and digit classification) directly within the `data/` directory. Click **"Reproducible Run"** to execute the pipeline. The console will display the selection process, the exact number of features retained, and the final classification performance (Accuracy & F1 Score).

Fields of Study & Subjects

Face and Expression Recognition Data Stream Mining Techniques Image Retrieval and Classification Techniques

Publication details

Journal / source
Code Ocean
Publication date
1 January 2026
Language
EN