Python implementation of an unsupervised learning algorithm: leveraged affinity propagation

Abstract

This work presents a new Python library for Leveraged Affinity Propagation (LAP), a clustering technique used in machine learning and data analysis. LAP is designed to reduce memory usage compared to existing Affinity Propagation (AP) implementations, including those in scikit-learn and the original R library. The library’s performance and memory usage are evaluated against these existing methods through various experiments on benchmark datasets, both synthetic and real-world. The results demonstrate that the proposed library achieves competitive performance while significantly reducing memory usage. The library is a valuable addition to the Python data science ecosystem and offers an efficient and effective tool for AP-based clustering analysis.