divergenceEM

divergenceEM is a computational framework designed to implement the Expectation‑Maximization (EM) algorithm for estimating divergence parameters in complex data sets. It was first released in 2015 as an open‑source Python package and has since evolved with contributions from the bioinformatics and machine learning communities. The core idea behind divergenceEM is to provide a modular interface for users to define custom likelihood functions, initial parameter settings, and convergence criteria while automating the iterative EM procedure. The library supports both batch and online variations of EM, which is particularly useful for streaming data or large‑scale genomic sequencing applications.

The principal use cases for divergenceEM include estimation of model‑based evolutionary distances in phylogenetic analyses, inference of mixture components in population genetics, and calculation of distributional divergences in anomaly detection tasks. In the context of clustering, divergenceEM can be combined with Gaussian Mixture Models (GMMs) or Dirichlet process priors to automatically refine cluster boundaries. The package provides diagnostic tools such as log‑likelihood trajectories, parameter convergence plots, and Bayesian Information Criterion (BIC) scores to aid model selection.

Users learn divergenceEM through its comprehensive API documentation and tutorials, which cover installation via pip, sample scripts for standard datasets, and instructions for extending the framework with custom likelihoods. The project is maintained under a permissive MIT license and welcomes bug reports and feature requests through its GitHub repository. By abstracting the repetitive components of EM implementation, divergenceEM allows researchers to focus on problem‑specific modeling decisions without reinventing the underlying algorithmic machinery.