Description

The goal of this topic is a comparative analysis of several probabilistic programming languages (PPLs) available in Python, namely Tensorflow Probability, Pyro, Numpyro and PyMC.

It should consider both,  qualitative (available algorithms, how active is the community, completeness of documentation, what are community recommendation, ...) as well as a quantitative aspects (empirical time and memory complexity, scalability, utilization of specialized hardware) for selected reference models.

Please note: The topic very ambitious, not all languages / reference models / algorithms need to be evaluated for a good result!

Requirements:

• Python experience
• Basic probability theory
• General understanding of deep learning

If you are interested in this topic or a variation of it, feel free to contact me!