Learning Bayesian Statistics #3.1 What is Probabilistic Programming & Why use it, with Colin Carroll
Nov 5, 2019
Colin Carroll, a machine learning researcher and key contributor to PyMC3 and ArviZ, discusses the intricacies of probabilistic programming. He explains its value in the realm of Bayesian statistics and provides insights on selecting between various libraries like Stan and Pyro based on project requirements. Colin shares his journey from pure mathematics to data science and emphasizes the importance of quantifying uncertainty for better decision-making, particularly in high-stakes scenarios like flight insurance.
AI Snips
Chapters
Transcript
Episode notes
Colin's Python and Bayesian Journey
- Colin Carroll's initial data science job allowed him to learn Python over several years.
- His introduction to Bayesian methods and PyMC3 happened later during a long commute.
First Open-Source Contributions
- Colin's first open-source contributions involved improving PyMC3's test suite.
- He was motivated by the friendly community and the quick acceptance of his pull requests.
No-U-Turn Sampler Mistake
- Colin rewrote the No-U-Turn Sampler in PyMC3, but it was incorrect and later reverted.
- He learned the importance of careful review and testing in open-source projects.