Closed bearecinos closed 2 years ago
Hi @jrmaddison
I think some extra comments are needed to explain the processing steps / NaN handling etc.
A lot of the comments on how we handle NAN are written here: https://github.com/bearecinos/smith_glacier/blob/stable_inference/scripts/prepro_stages/02_gridded_data/generate_subsampling_vel_files.py#L162
That is handle outside of the subsampling function in the main velocity generation files. More info in the wiki
New or modified scripts should be linted with flake8. I will try to add this to Pycharm but not sure if there is a plugging at the moment ... will try to set up that in the future
I will try to add this to Pycharm but not sure if there is a plugging at the moment ... will try to set up that in the future
There are probably too many flake8 errors to add this to the test suite at the moment -- but it'd still be good to reduce their number.
Added suggested changes ... see latest commits
@dngoldberg
I think this is ready to be merge as I have ran all the experiments and code seems to be doing what we wanted. Making J_reg close to zero and calling comp_J_inv() to test alpha and beta generated by the training set against the testing velocity set.
There are a lot of new files but they all tackle issue #22 ...
The most important part of the code is the subsampling function, which we modified last week: https://github.com/bearecinos/smith_glacier/blob/stable_inference/ficetools/velocity.py#L349-L393
And the run_stable_inference_inv.py routine that we built with @jrmaddison the other day.
No need to review it in detail, this is just a PR to inform you of the many files (mostly tomls) to be added. I ran the experiments and all seems to work... we can discuss the results at the meeting.
Cheers Bea