search

lalvim / PartialLeastSquaresRegressor.jl

Implementation of a Partial Least Squares Regressor
MIT License
40 stars 8 forks source link

Refactoring to support MLJ #10

Closed filipebraida closed 4 years ago

filipebraida commented 4 years ago

I created a PR.

This version still has some warnings, but it works.

@tlienart

tlienart commented 4 years ago

I've just had a brief look but it's probably best if we iterate in several shots, eventually when we've cleared out some stuff I'll also get Anthony to have a look.

This is exciting!

ablaom commented 4 years ago

There is an open issue here: https://github.com/alan-turing-institute/MLJModels.jl/issues/269 @vollmersj has assigned himself .

ablaom commented 4 years ago

Update. You can now inverse_transform the standardisation of tables. And the one model Standardizer handles both tables and vector data.

You will need to update MLJModels

filipebraida commented 4 years ago

Thanks @ablaom ! We (@lalvim ) will update with this new update.

filipebraida commented 4 years ago

@tlienart , we refactored the code with the new MLJ version. It's working now with standardization on the target.

We can close a version and change the package name. What do you think?

azev77 commented 4 years ago

@ablaom @tlienart Is MLJ now ready for integration w PLS?

ablaom commented 4 years ago

Issue to add PLSRegressor to MLJ registry: https://github.com/alan-turing-institute/MLJModels.jl/issues/312

ablaom commented 4 years ago

My vote would be for PartialLeastSquaresRegressor. I had to look up PLSRegressor.

But then "MLJ" is pretty awful 😉 and I haven't had the courage to change it.

azev77 commented 4 years ago

FWIW, I love MLJ. To me, more minimalist = better.