Ankur Ankan

Radboud University  PhD Candidate

Project: Enhancing and Promoting Data Simulation Capabilities of pgmpy 

The goal: pgmpy is a Python package that offers flexible simulation methods for both standard and time-series data from graphical models such as Bayesian Networks, Directed Acyclic Graphs, and Dynamic Bayesian Networks. Currently, one of the main limitations of this simulation feature is its restriction to only simulating categorical data. This project aims to extend the simulation capabilities of pgmpy to include continuous and mixed data types. Additionally, I will develop two tutorials focusing on predictive modeling and causal inference use cases for these simulations. Through these efforts, this project aims to significantly enhance the simulation capabilities of pgmpy and increase awareness within the community about the applications of simulations and the features offered by pgmpy.  

The why: Simulation methods are widely used in statistics and machine learning for education and to evaluate methods and algorithms. This project would give researchers a flexible Python tool to perform simulation studies.