laughy

The one I use is slightly more advanced, can't copy it for some strange reason, but I use a service that checks for a valid domain first

A better alternative is to assume t-distributed errors

This suggests that the effective number of parameters is far lower than the nominal number. My head canon for neural networks as overparametrized models still holds.

Very cool! Found this from 9 years ago, deprecated now however. https://news.ycombinator.com/item?id=8304409

I have a custom written function in R that loads in .Rprofile to push messages e.g. when some script is done.

The latest research suggests Linear A and Linear B is the same script, but used for different languages analogous to how the Latin script is both used for say English and Polish, but with language specific adjustments.

I have applied it to the names in a population database. It learnt interesting, and expected structure. Visualized with UMAP it clustered by gender first, and then something that probably could be described as cultural origin of name.

There is also the tidytable package. But dtplyr works really well. Have used it in a couple of shiny apps that wrangle some heavy input files.

Put it here as well https://archive.ph/yuQyu