I really wish I could run church in a notebook as I need a bit more space to experiment in order to solve the problems in probmod.
Maybe just a repl would be enough.
I guess the bottom line is that it made sense to do probmod in a more “normal” programming language. Figuring out the lisp like code and the probabilistic problems at the same time is quite a mouthful.
Is it possible that it's just custom? It looks very familiar, but https://probmods.org/assets/css/default.css seems to be the bulk of the non-bootstrap styling and it doesn't look like it was generated.
Took a peek. Information is accessible and interesting. Code snippets work and visualizations are clear. Can anyone who has read this (or similar) speak to deeper points? Did it change anything for you? Is it outdated? etc.
It's not outdated. Many ideas there, like mixture models or hierarchical models are pretty timeless mathematical concepts.
A nice follow up is [1], which delves more into more technical inference aspects.
There are also follow up texts by the same authors [2,3].
Another approach to further learning is to implement your own library. My favorite implementation is [4], which is both advanced and easy to understand.
8 comments
[ 5.6 ms ] story [ 29.3 ms ] thread[0]: http://v1.probmods.org/
Maybe just a repl would be enough.
I guess the bottom line is that it made sense to do probmod in a more “normal” programming language. Figuring out the lisp like code and the probabilistic problems at the same time is quite a mouthful.
A nice follow up is [1], which delves more into more technical inference aspects.
There are also follow up texts by the same authors [2,3].
Another approach to further learning is to implement your own library. My favorite implementation is [4], which is both advanced and easy to understand.
[1] https://arxiv.org/abs/1809.10756
[2] http://dippl.org/
[3] https://agentmodels.org/
[4] https://gen.dev/