I do machine learning and natural language processing. Tells jokes and anecdotes.
Joined June 2009
- Tweets5.4K
- Following413
- Followers1.8K
- Likes5.6K
What priors should I use for my parameters?
How do I fit a Bayesian model when I have so much data?
How do I diagnose and debug what's wrong with my Bayesian model?
1/ 🧵
You'll learn how to:
* Use variational inference to massively speed up inference on large datasets,
* Build causal models that quantify true impact, not just correlation,
* Identify when your model is under-specified or over-determined, and how to rewrite it
5/
Below is a Bayesian causal impact model built in PyMC — recovering an intervention effect from observational time series, with uncertainty on every estimate.
This is our only in-person workshop in London this year. Register now to secure your seat: dub.sh/x28nV7N
6/6
∬ ⌬ + ☢ retweeted
🚀 Thrilled to welcome our new Bayesian Wizard, @zaxtax, as Data Scientist @pymc_labs🧙♂️ With over a decade of expertise in Bayesian modeling & probabilistic programming, he's set to push boundaries on fascinating tech challenges across industries.🔥 Let's do this! #Bayesian #PyMC
If you use #pymc, do you prefer an explicit or implicit style for passing in coords?
github.com/pymc-devs/pymc-ex…
∬ ⌬ + ☢ retweeted
We are excited to announce the Call for Proposals for the PyMCon Web Series! 🚀
We are looking for folks like you to share your knowledge, best practices, or tricks with the wider PyMC community 📣
Send your proposal: pymcon.com/cfp
∬ ⌬ + ☢ retweeted
#MachineLearning has value & staying power but it's not optimal for every data-wrangling task. Our latest post explores #StatisticalModeling & #ProbabilisticProgramming and how to use domain expertise to model the processes underlying data
quansight.com/post/robust-pr… #Python #PyData
∬ ⌬ + ☢ retweeted
1/5
📣We are thrilled to announce that PyMC 4.0 is out!
Try it out with:
>conda install -c conda-forge pymc
#PyMCv4 Release announcement:
pymc.io/blog/v4_announcement…
∬ ⌬ + ☢ retweeted
Incredibly excited that #PyMC 4.0 is finally released. While this release is a major rewrite (it is now based on @AesaraDevs instead of theano), with many new features (JAX sampling, GPU support), it still has the same modeling API that you love.
1/5
📣We are thrilled to announce that PyMC 4.0 is out!
Try it out with:
>conda install -c conda-forge pymc
#PyMCv4 Release announcement:
pymc.io/blog/v4_announcement…
∬ ⌬ + ☢ retweeted
Did you know that PyMC variables can have symbolic shapes and be resized anytime? - 1/3🧵
∬ ⌬ + ☢ retweeted
A new table for Computer Science / AI academics willing to host Ukranian researchers. Many top @ETH Zurich groups are entering. Please enter your lab here, we will then append it to the global list in #ScienceForUkraine
#CSforUkraine #AIforUkraine
docs.google.com/spreadsheets…
∬ ⌬ + ☢ retweeted
We had a lot of fun writing this survey! Thank you, @rupakmajumdar, for inviting us to write this.
You can find a free excerpt here: nowpublishers.com/article/Do…
Below, a 🧵on what neurosymbolic programming (NSP) is and why we think it's important. (1/n)
Hey, ML/PL enthusiasts! Looking for some "light" reading for the holiday break?
FnT just published our survey on "Neurosymbolic Programming", written jointly with @swarat, Kevin Ellis, @rishabhs, Armando Solar-Lezama, and @yisongyue.
nowpublishers.com/article/De…
∬ ⌬ + ☢ retweeted
"The Many Faces of Information Geometry": Gentle
short introduction to information geometry. Open access PDF by @amermathsoc "Notices of the AMS". IG aims at unravelling geometric structures of families of probability distributions
ams.org/journals/notices/202… #InformationGeometry
∬ ⌬ + ☢ retweeted
Congrats to the Bean Machine team for launching a new PPL and thank you for mentioning @arviz_devs. It's excited as a Bayesian practitioner to see new tools, and as an OSS maintainer to see that folks use our tools!
research.facebook.com/blog/2…
