search

so-wise / weddell_gyre_clusters

Unsupervised classification of Weddell Gyre profiles
MIT License
2 stars 1 forks source link

Implement data subsetting #3

Closed DaniJonesOcean closed 3 years ago

DaniJonesOcean commented 3 years ago

At present, in the example notebook 1.0, I train using the entire dataset, which isn't necessarily the best practice. I should implement a better test-train-split that uses roughly uniform coverage where available.

DaniJonesOcean commented 3 years ago

Can we use a command like profiles = profiles.sel(iDEPTH=slice(zmin,zmax)) with the sel function, to slice using longitude and latitude? It would be an iterative function, I guess. It would iterate over lat/lon and try to grab a sample in every box (e.g. 2°, if possible).

DaniJonesOcean commented 3 years ago

I think this will be tricky in such an inhomogeneous, somewhat sparsely populated dataset. Closing for now, may reopen later.