Skip to main content

Probabilistic Programming: What, Why, How, and When

Beau Cronin (Embedding.js)
Data Science
Ballroom AB
Average rating: ****.
(4.10, 10 ratings)

Probabilistic programming languages are systems in which stochastic modeling primitives are “first class” citizens. At a minimum, these environments provide a clean separation of probabilistic modeling, which captures our assumptions about the world and the hypothesis space, and inference, i.e. learning which hypotheses provide the best explanation of the observed data. Though still a subject of intense academic research, these systems have recently emerged in various open source projects and are now available for widespread evaluation and use.

This talk will provide a high-level introduction to the state of the probabilistic programming field, and give a feel for what it’s like to use these environments.

Questions that will be answered include:

  • What are the advantages of probabilistic programming over other approaches?
  • How should one choose from the available probabilistic programming systems? What are the relative strengths and weaknesses of the various options?
  • What do probabilistic programs “look like”? How are these programs different from more traditional solutions?
  • Are these systems ready for scaled production use? What further developments are most needed to accelerate their adoption in industry?
Photo of Beau Cronin

Beau Cronin

Lead Developer, Embedding.js

Beau co-founded two startups based on probabilistic inference, the second of which was acquired by Salesforce in 2012. He knows works there as a product manager. He received his PhD in computational neuroscience from MIT in 2008.