New unsupervised feature selection method LIDFS outperforms existing techniques

Researchers have developed LIDFS, a novel unsupervised feature selection method that leverages Local Intrinsic Dimensionality (LID) to identify the most discriminative features in high-dimensional datasets. This approach treats features as points and estimates their LID based on their nearest neighbors, assigning higher importance to features with lower LID. LIDFS does not require class labels and has a time complexity of O(d^2n). Evaluations on 23 datasets showed LIDFS outperforming existing methods in supervised tasks and performing competitively in clustering evaluations, even when compared to methods that do use class labels. AI

IMPACT This method could improve the efficiency and effectiveness of machine learning algorithms by reducing dimensionality without requiring labeled data.

RANK_REASON The cluster contains a research paper detailing a new method for unsupervised feature selection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• DagsHub
• Hugging Face
• LIDFS
• Litmaps
• Muhammad Rajabinasab
• scite Smart Citations

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →