All of my other projects are on hold because I discovered that #Stan, despite using its own model specification language, is really superb for Markov chain Monte Carlo (#MCMC) analysis of statistical models. I am being much more productive with it than #PyMC3.
I'm revisiting #Ebisu after folks patiently explaining a core modeling defect ten different ways till I finally understood.
As much as I hated my stint in grad school, it's the reason I know how to wield these super-powered tools.
Ebisu quiz scheduling library
Instead of pushing forward with research on a new probabilistic model for #Ebisu, I released a handful of new features folks had long requested with just the old Beta distribution model.
If anyone needs a Python (or JS, or JVM, or Dart, though those haven't gotten the new features) library for flashcard review scheduling, it might be useful:
Hope to get back to Stan soon and finding a better underlying model to handle memory growth.
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!