Practical Probabilistic Programming with Figaro

on YouTube ↗24m

" summary

A mature probabilistic programming library implemented as Figaro in Scala provides a Scala-embedded model representation and inference engines; it runs hydrological site-selection and a DARPA malware-lineage task where the malware lineage scored higher than prior methods, and a tracklet-merging routine that fits in a few lines — tracklet merging in a few lines. Structured factored inference decomposes models into reusable subproblems, the hybrid solver picks between variable elimination and belief propagation, is orders-of-magnitude more accurate, and variable elimination runs out of memory on larger cases.

AI-written summary · about

" chapters (26)

1. 0:00Intro
2. 0:35Overview
3. 0:56What Are We Trying To Do?
4. 1:21Probabilistic Reasoning Lets You Do All These Things
5. 2:04Probabilistic Reasoning: Predicting the Future
6. 2:36Observations
7. 3:33But Probabilistic Reasoning Is Hard!
8. 4:09Goal of Probabilistic Programming
9. 4:21How Probabilistic Programming Achieves This
10. 5:03Probabilistic Programming Compared to Deep Learning
11. 7:06Figaro goals
12. 8:04Figaro as a Scala Library
13. 10:07Disadvantages of Scala Embedding
14. 11:20Hydrological Terrain Modeling for Army Logistics
15. 12:19Malware Lineage (DARPA Cyber Genome)
16. 13:59Tracklet Merging (DARPA PPAML Challenge Problem)
17. 16:20Tracklet Merging in Figaro
18. 17:37Current State of the Art
19. 18:12Our Goal
20. 19:26Automated Inference Strategy
21. 19:57Structured Factored inference (SFI)
22. 20:36Compiled Graphical Model of Figaro Program
23. 20:55Combine and Reuse Solutions
24. 21:18Optimize Each Subproblem Individually
25. 22:50Conclusion
26. 23:23Acknowledgement

" related talks

• 38mProbabilistic programming and meta-programming in ClojureVikash K. Mansinghka · Clojure/conj 2018
• 31mProbabilistic Programming: The What, Why and HowMaria I. Gorinova · SPLASH 2020 REBASE
• 01h 03mAn Overview of Probabilistic ProgrammingVikash K. Mansinghka · Strange Loop 2015
• 29mMotivating Probabilistic ProgrammingRahul Chitturi · Scale By The Bay 2018
• 40mProbabilistic Programming for Augmented IntelligenceVikash K. Mansinghka · Uncertainty in Computation 2016
• 35mScalable structure learning and inference for domain-specific probabilistic progFeras A. Saad · LAFI 2022
• 26mSemi-symbolic Inference for Efficient Streaming Probabilistic ProgrammingCharles Yuan, Michael Carbin, Eric Atkinson, Guillaume Baudart, Louis Mandel · OOPSLA 2022
• 36mProbabilistic Programs Which Make (Common) SenseZenna Tavares · Strange Loop 2015