I got a Probabilistic Programming starter pack going. Hit me up if you're involved with #probprog R&D and want in!
go.bsky.app/JfvubEf
RE: https://bsky.brid.gy/convert/ap/at://did:plc:6ls7x4kw3wsz2opik4wobgkm/app.bsky.graph.starterpack/3lbcssvf7yz2x
#probprog
0 posts0 participants0 posts today
Tired of waiting forever for MCMC chains to converge? We experimented with using Pathfinder VI to initialize HMC and get early model diagnostics. https://mlcolab.org/public-events/faster-bayesian-inference-with-pathfinder #bayesian #probprog #probml
Faster Bayesian inference with PathfinderFaster Bayesian inference with PathfinderLet me tell you about the most frustrating part of Bayesian modeling. Often the first models you build either make bad assumptions or contain bugs. Both can cause the already expensive step of drawing posterior samples using Markov Chain Monte Carlo (MCMC) to become unbearably slow, but those samples are often the best way to check if our model makes sense. So we draw samples, encode some more better assumptions, draw more samples, fix some bugs, draw more samples, go check our error model against the laboratory equipment, rinse and repeat. And gradually we move toward higher quality, more useful models, which often can be sampled much faster. This is known as the folk theorem of statistical computing.
Updated some new thoughts regarding the TerpreT problem and my naive solution
https://luxxxlucy.github.io/projects/2021_terpret/index.html
The original TerpreT paper(https://arxiv.org/abs/1608.04428)
discussed solving program induction by gradient based optimization(after making the program differentiable by relaxation ).
I'm Chad. Hello from #Seattle! 
Since this is @fosstodon ... My #FOSS work is in #julialang, mostly around #Bayesian modeling and probabilistic programming (#probprog)
This started with Soss.jl, a probabilistic programming language (#PPL). I eventually realized I needed primitives with better composability, and started work on MeasureTheory:
https://github.com/cscherrer/MeasureTheory.jl
That's coming along well - next it's back to PPL, now with Tilde.jl, similar to Soss but a bit more flexible
GitHubGitHub - JuliaMath/MeasureTheory.jl: "Distributions" that might not add to one."Distributions" that might not add to one. Contribute to JuliaMath/MeasureTheory.jl development by creating an account on GitHub.
Continued thread
Introduction
Hi all, my #introduction:
I'm a prof at #UCLA CS, living in #LosAngeles, and researching #ArtificialIntelligence.
I enjoy bridging #machinelearning with probabilistic and logical #reasoning.
That makes me work on probabilistic programming (#probprog), tractable probabilistic models (e.g., #probcircuit), and #neurosymbolic #AI.
Looking forward to some more authentic discourse about AI on this platform.