scikit-clarans

scikit-clarans implements the CLARANS (Clustering Large Applications based on RANdomized Search) and FastCLARANS algorithms for Python using the scikit-learn estimator interface.

Key characteristics:

• Scikit-learn compatibility: estimators implement fit, predict, and transform as standard clusterers.

• Low memory footprint: computes distances dynamically with \(O(n)\) memory instead of allocating an \(O(n^2)\) pairwise matrix.

• C acceleration: core swap evaluation runs in Cython with a pure Python fallback.

• Flexible metrics: supports dense arrays via SciPy, sparse matrices via scikit-learn, and custom callables.

Note

This library is designed for coursework, algorithm study, and research prototyping. It scales well to tens of thousands of samples on a single machine, but it is not intended for distributed data systems.

Getting Started

• Installation
  • Requirements
  • Install with pip
  • Install from source
• User guide
  • Why use k-medoids?
  • Quick start
  • Choosing an estimator
  • Configuration and parameter tuning
  • Fitted attributes
  • Monitoring search progress
  • Search mechanics
• Examples
  • Overview of examples
  • Interactive execution

API Reference

• API Reference
  • Estimators
  • Helper Modules

Example Gallery

• Cluster Geometries
• Distance Metrics and Representations
• Seeding and Validation
• Inference and Pipelines
• Robustness and Anomaly Detection
• Benchmarks and Convergence

Project

• Project
  • Scope
  • Citation
  • Academic References
  • Resources
• Contributing
  • How to contribute
  • Pull request guidelines
• License

Indices and tables

• Index

• Module Index