This repository has been archived by the owner on Nov 29, 2023. It is now read-only.
Critique of the MetNet paper #28
Labels
Comments
jacobbieker
added a commit
that referenced
this issue
Jun 30, 2021
The correlation between days of the week might mean that there is leakage between the training and val/test sets. This is now split up so that 2020 is used for training, and 2021 is used for val/testing. This does mean that there is only data for the first half of 2021 in the test/val set for now, but that does cover winter/spring/part of summer so far, with more as time goes on.
|
Oh interesting! I hadn't thought of the correlation between days of the week, didn't really think about that. Changed the training splits to address that. And hopefully adding in more data sources will help the predictions of the model too. So far, I still haven't added radar data yet, as its in a odd format, or NWP data, both which should help as well. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Not sure if this is the right place to post this (!) (maybe we should use github discussions?!?) but here's an interesting critique of MetNet by @NiklasRiewald: https://niklasriewald.com/2021/06/29/paper-review-metnet-a-neural-weather-model-for-precipitation-forecasting/
The text was updated successfully, but these errors were encountered: