Probabilistic Python: An Introduction to Bayesian Modeling with PyMC
summary
Overview of what is actually done and what comes out: introduces PyMC v4 and the Aesara/JAX backends, fits models with the No-U-Turn Sampler (NUTS), and shows concrete examples — a change-point model on MLB pitch spin that estimates a spin rate drop of ~100 rpm with a posterior change point around early June, and a Six Nations Poisson attack/defense model that yields predictive win probabilities (England ≈ 82%). Covers diagnostics (divergences, R-hat, BFMI), posterior predictive checks, and practical performance notes.
AI-written summary · about
chapters (44)
- 0:00Welcome!
- 1:19Probabilistic programming
- 1:53Stochastic language ”primitives”
- 3:06Bayesian inference
- 3:27What is Bayes?
- 3:57Inverse probability
- 4:39Why Bayes
- 5:13The Bayes formula
- 6:51Prior distribution
- 8:12Likelihood function
- 8:29Normal distribution
- 8:53Binomial distribution
- 9:14Poisson distribution
- 9:32Infer values for latent variables
- 9:54Posterior distribution
- 10:56Bayes by hand
- 12:18Conjugacy
- 16:43Probabilistic programming in Python
- 17:24PyMC and its features
- 19:15Question: Among the different probabilistic programming libraries, is there a difference in what they have to offer?
- 20:39Question: How can one know which likelihood distribution to choose?
- 21:35Question: Is there a methodology used to specify the likelihood distribution?
- 22:30Example: Building models in PyMC
- 27:31Stochastic and deterministic variables
- 37:11Observed Random Variables
- 41:00Question: To what extent are the features of PyMC supported if compiled in different backends?
- 41:47Markov Chain Monte Carlo and Bayesian approximation
- 43:04Markov chains
- 44:19Reversible Markov chains
- 45:06Metropolis sampling
- 48:00Hamiltonian Monte Carlo
- 49:10Hamiltonian dynamics
- 50:49No U-turn Sampler (NUTS)
- 52:11Question: How do you know the number of leap frog steps to take?
- 52:55Example: Markov Chain Monte Carlo in PyMC
- 1:13:30Divergences and how to deal with them
- 1:15:08Bayesian Fraction of Missing Information
- 1:16:25Potential Scale Reduction
- 1:17:57Goodness of fit
- 1:22:40Intuitive Bayes course
- 1:23:09Question: Do bookmakers use PyMC or Bayesian methods?
- 1:23:53Question: How does it work if you have different samplers for different variables?
- 1:25:09Question: What route should one take in case of data with many discrete variables and many possible values?
- 1:25:39Question: Is there a natural way to use PyMC over a cluster of CPUs?
related talks
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
42mBayesian Non-parametric Models for Data Science using PyMC3
43mProbabilistic Programming in the Real World
01h 31mProbabilistic Programming and Bayesian Inference in Python
01h 27mBayesian Time Series
47mThe State of the Art for Probabilistic Programming
44mAn introduction to PyMC3
39mProbablistic Programming Data Science with PyMC3