search

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

Equation discovery notebook #104

Closed LaureZanna closed 1 year ago

LaureZanna commented 1 year ago

Just to follow up on our discussion last week, we decided to start with equation discovery as the next big section in the Jupyter book. I will list below a possible structure for us to iterate on .

LaureZanna commented 1 year ago

I think the following might work: 1) starting with the notebook comparison of method. We might need an intro section to explain the motivation in general.

2) Then, we can do an example with L96 with SinDy, with this notebook: sindy_L96_2scale.ipynb . This is a nice and simple one.

3) And then close with learning sub grid from equation discovery vs. NN [symbolic_vs_nn_multiscale_L96.ipynb]) https://github.com/m2lines/team-meetings/blob/main/22-01-12-deep_vs_other_ML/symbolic_vs_nn_multiscale_L96.ipynb)

I don't think we need to cover L63 or L96 with 1 timescale. It might be too much. Of course, this is just a suggestion. Let me know what you think. cc: @dhruvbalwada @IamShubhamGupto @shan18