Probabilistic Programming in the Real World
summary
Shows concrete models and results: uses PyMC3 and MCMC variants (Metropolis, HMC/NUTS) to build a truncated-normal joke-rating model from three ratings, a Bernoulli model for classifier accuracy, a discrete-mixture model for document types, and a Student‑t comparison for error rates. Produces posterior samples, credibile intervals, and summary probabilities—about 80% confident the joke is funnier at PyData and roughly 75% posterior probability the new process yields fewer errors—enough to drive data‑collection and business decisions.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 2:28Joke funniness modeling
- 9:51Bayesian modeling theory
- 17:35Markov chain Monte Carlo and Metropolis algorithm
- 19:39Real-world problems solved
- 21:45Model validation
- 24:34Expert knowledge in models
- 28:05Uncertainty in business
- 32:41Probabilistic programming
- 36:53Q&A
related talks
01h 27mProbabilistic Python: An Introduction to Bayesian Modeling with PyMC
47mThe State of the Art for Probabilistic Programming
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
42mSolving Real-World Business Problems with Bayesian Modeling
01h 31mProbabilistic Programming and Bayesian Inference in Python
42mBayesian Non-parametric Models for Data Science using PyMC3
01h 03mAn Overview of Probabilistic Programming
39mProbablistic Programming Data Science with PyMC3