m2lines / L96_demo

Lorenz 1996 two time-scale model for learning machine learning
https://m2lines.github.io/L96_demo
Other
36 stars 20 forks source link

Prepare for JOSE submission #135

Closed dhruvbalwada closed 1 year ago

dhruvbalwada commented 1 year ago

We will submit this repo to JOSE. The submission requirements are detailed here.

Tasks:

What should be in paper.md:

dhruvbalwada commented 1 year ago

Draft:
I wrote a first draft of the submission for JOSE: https://github.com/m2lines/L96_demo/blob/neural_net_for_L96/paper.md (I know this can be annoying to edit, so the text can be edited on this google doc) The main things that need to be covered in a JOSE article are in the list above (I got it from JOSE website). It should be short ~ 1000 words. I based this on the introduction that is part of our jupyter book, but made some minor modifications. Also based on example provided by JOSE.

Authors: I included everyone who replied to me about this (https://docs.google.com/spreadsheets/d/14DM7wuhKjexrYV9e9GT0fYKZfW0AvqkrDh0xPQnO3dM/edit#gid=0). However, a few people are left who have not replied. I am assuming they are too busy. Should I individually ping them?

LaureZanna commented 1 year ago

@dhruvbalwada : the draft is outstanding!! just made some minor edits.
For pinging authors: @johannag126 might be able to help you and reach out to them directly so you can move ahead with submission?

Thanks again @dhruvbalwada for your leadership!

dhruvbalwada commented 1 year ago

Submitted https://jose.theoj.org/papers/c644a0264f445698f212a051d8ace6e8