News for hackers. Stuff that matters.
Re:Deep learning for twelve hour precipitation forecasts by dr_dshiv on Sunday February 12, 2023 @08:34AM (#34762692)
> About as good as you can get without collecting new data.

So let’s call holdout datasets “second-best practice”?

Re:Deep learning for twelve hour precipitation forecasts by zamfi on Sunday February 12, 2023 @12:47PM (#34764925)
LOL, love it!

Sadly too common that collecting new data is practically speaking impossible, because the actual collection process is underspecified.