O'Reilly logo

Probabilistic Programming from Scratch 1: A/B Testing with Uncertainty

Probabilistic programming systems make it easier to learn from data. Fast Forward Lab's Mike Lee Williams demos a novel no-math-required way to make one.

You have two web site designs. Which one best converts customers? Choose wrong and you're out a million bucks. See how Mike Lee William's Approximate Bayesian Computation (ABC) inference machine transforms a trove of typically imperfect and incomplete data into an A/B choice you can live with.

What you will learn:

  • Build a probabilistic programming system from scratch using ABC and basic Python
  • Understand ABC: A no-math-required, easy-to-code algorithm that performs Bayesian inference
  • Explore three highly attractive advantages to the Bayesian approach