This repository is for the implementation and application of statistical methods. Specifically, it contains code and resources for the application of the Expectation-Maximization (EM) algorithm and Variational Inference (VI).
The Expectation-Maximization algorithm is a statistical tool for maximum likelihood estimation in latent variable models. In this project, the EM algorithm is applied to:
- Estimate model parameters
- Infer hidden variables from observable data points
- Identify clusters in the data
Variational Inference is an inference technique especially used in Bayesian statistics that approximates probability densities through optimization. In this project, VI is used to:
- Efficiently approximate posterior distributions
- Infer hidden variables (by approximating the posterior distribution of latent (hidden) variables)
- Identify clusters in the data