Thomas Wiecki: The State of the Art for Probabilistic Programming | PyData Global 2022

Probabilistic Programming as a field is moving at breakneck speed, with innovations being driven on all levels: language, algorithms, compilers, computation, hardware. In this expert briefing I will give a brief overview of where the field is today and where it is headed. One big trend is what I call The Great Decoupling: rather than monolithic PPL systems, we are seeing how various layers of abstraction are introduced and separated. This allows more interoperability, as well as innovation to occur at every level of the stack. Finally, I will talk about a convergence of Bayesian modeling and Causal Inference to a new paradigm called Bayesian Causal Inference.
Event Description
Probabilistic Programming as a field is moving at breakneck speed, with innovations being driven on all levels: language, algorithms, compilers, computation, hardware. In this expert briefing I will give a brief overview of where the field is today and where it is headed. One big trend is what I call The Great Decoupling: rather than monolithic PPL systems, we are seeing how various layers of abstraction are introduced and separated. This allows more interoperability, as well as innovation to occur at every level of the stack. Finally, I will talk about a convergence of Bayesian modeling and Causal Inference to a new paradigm called Bayesian Causal Inference.
Speaker bio:
Dr. Thomas Wiecki is an author of PyMC, the leading platform for statistical data science. To help businesses solve some of their trickiest data science problems, he assembled a world class team of Bayesian modelers founded PyMC Labs -- the Bayesian consultancy. He did his PhD at Brown University studying cognitive neuroscience.
- GitHub: github.com/twi...
- Twitter: / twiecki
- Website: twiecki.io/
PyMC Labs
- PyMC Labs: www.pymc-labs.io
#bayesian #statistics #python
===
www.pydata.org
PyData is an educational program of NumFOCUS, a 501(c)3 non-profit organization in the United States. PyData provides a forum for the international community of users and developers of data analysis tools to share ideas and learn from each other. The global PyData network promotes discussion of best practices, new approaches, and emerging technologies for data management, processing, analytics, and visualization. PyData communities approach data science using many languages, including (but not limited to) Python, Julia, and R.
PyData conferences aim to be accessible and community-driven, with novice to advanced level presentations. PyData tutorials and talks bring attendees the latest project features along with cutting-edge use cases.
00:00 Introduction by the moderator
00:38 Thomas introduces himself
02;34 PyMC Labs
03:20 Probabilistic programming GitHub star history
05:17 PyTensor Announcement
07:18 The Great Uncoupling of packages
12:44 BlackJAX: Inference Algorithms in JAX
15:40 Nutpie sampler: NUTS written in Rust
15:51 PyScript and pyodide
19:01 Demonstration of how PyScript and pyodide work
20:26 Auto-marginalization
22:07 Bayesian Causal Inference
23:18 CausalPy
24:15 Bayes in business trends
26:30 PyMCon Web series: pymcon.com/
27:06 Intuitive Bayes course
29:04 Q & A Are you working with banks or insurers on capital modeling?
30:33 Q & A Is there an overlap between Aesara and JAX?
32:50 Q & A How general is the auto-marginalization, can it work for arbitrary distributions?
34:06 Q & A Does PyMC have any functionality for time series models yet?
35:36 Q & A Do you encourage and work with external bodies to contribute to the PyMC universe?
38:13 Q & A It feels like the library is moving fast, is the documentation and learning resources or examples keeping the same fast pace?
41:00 Q & A If one wanted to contribute to PyMC, is it beginner friendly?
44:37 Q & A Will the video of the talk be shared?
46:44 How to get in touch with PyMC Labs
47:09 Thank you!

[@kirill2842]

Thank you, Thomas!

[@rafaelortega1376]

Very good talk!

[@dinoscheidt]

<a href="#" class="seekto" data-time="227">3:47</a> just note that the x-axis labels are screwed up - or the Romans started smashing that gh star button 😂

[@brendanhansknecht4650]

It's correct, just a bad font where 2010 looks like 200