Welcome to ArchetypAX’s documentation!
archetypax is a GPU-accelerated implementation of Archetypal Analysis using JAX.
Overview
Archetypal Analysis is a statistical method for dimensionality reduction that represents data points as convex combinations of extreme points (archetypes). Unlike PCA, which finds orthogonal directions of maximum variance, archetypal analysis finds extreme points that lie on the convex hull of the data, making it particularly useful for interpretable feature extraction and data exploration.
archetypax provides:
High-performance implementation using JAX for GPU acceleration
Scikit-learn compatible API for seamless integration
Advanced optimization techniques for improved convergence
Comprehensive visualization tools for result interpretation
Extensive evaluation metrics for model assessment
Multiple analysis variants, including Sparse Archetypal Analysis and Biarchetypal Analysis
Innovative ArchetypeTracker for monitoring archetype evolution during optimization
Trajectory analysis tools for visualizing archetype movements and convergence patterns
Boundary proximity tracking with historical metrics for evolutionary pattern analysis