Utilities Overview

This page provides an organized reference of all utility modules in scikit-feature.

Module	File	Description
Construct W	construct_W	Construct weight matrix for similarity-based methods
Data Discretization	data_discretization	Discretize continuous features into categorical bins
Entropy Estimators	entropy_estimators	Estimate entropy and mutual information from data
Mutual Information	mutual_information	Compute mutual information between variables
Sparse Learning Utils	sparse_learning	Utilities for sparse learning algorithms
Unsupervised Evaluation	unsupervised_evaluation	Evaluation metrics for unsupervised feature selection
Util	util	General utility functions

Common patterns

Many utilities are shared across algorithm categories:

• Entropy estimators and mutual information computations are used by all information-theoretic methods
• Weight matrix construction is essential for similarity-based algorithms (LapScore, SPEC, etc.)
• Sparse learning utils support the l2,1-norm based algorithms
• Data discretization prepares continuous data for the information-theoretic methods

See also

• Algorithms overview — every algorithm that relies on these utilities