Module projection 

Fast random projection methods for dimensionality reduction

This module provides algorithms for random projections, which are efficient techniques for reducing dimensionality of data while approximately preserving distances between points. These methods are particularly useful for large-scale machine learning applications where computational efficiency is crucial.

Available Methods

• Gaussian Random Projection: Projects data using a random matrix with entries drawn from a Gaussian distribution
• Sparse Random Projection: Uses a sparse random matrix for projection, which is computationally more efficient
• Very Sparse Random Projection: An extremely sparse random projection that only uses values {-1, 0, 1}
• Johnson-Lindenstrauss Transform: Implements the Johnson-Lindenstrauss lemma for dimension reduction with theoretical guarantees

Functions

gaussian_randommatrix

Generate a random projection matrix using Gaussian distribution

johnson_lindenstrauss_min_dim

Compute the minimum number of components needed for Johnson-Lindenstrauss transform

johnson_lindenstrauss_transform

Project data using Johnson-Lindenstrauss transform

project

Project data using a random projection matrix

sparse_randommatrix

Generate a sparse random projection matrix

very_sparse_randommatrix

Generate a very sparse random projection matrix