zx @zx@mastodon.xyz

Would anybody have any interest in me tooting out cool #emnlp2018 papers, #iclr2019 preprints, #arxiv preprints?

I'm thinking to replace my lab notebooks with a personal wiki to better tie all my projects and idle thoughts together.

Has anyone else done this? How have you liked the result?

I have been wrecked by a need for explanations to be testable. That theory building without making predictions or having a way to tell if you are wrong is just wasting time.

I think about ICML deadline I need to bite the bullet and really engage with the causality literature https://gradientinstitute.org/blog/6/

Basically unless I stick to prediction all the high-impact projects on my plate need causality.

I feel really guilty making a shitty joint model, but I only need the conditional distribution to be easy to sample from.

@mikiobraun I'm glad to see you back on this. I really want all of us to migrate off of birdsite.

Anybody else feel like TemplateHaskell makes your code way too brittle?

Don't you hate it when words mix Greek and Latin roots?!

https://en.wikipedia.org/wiki/Hybrid_word

One interesting way of finding places to use probabilistic programming is look for fields that have developed their own libraries effectively wrapping MCMC

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5624502/

I have successfully migrated off of docker onto podman! I don't know why this makes me so happy but it does!

So this week I learned that the number of nodes in an AST follows a lognormal distribution. This result holds across programming languages.

Hey why are so many instances closed for new users? Any one you like folks that I can get an invite for?