Open jcvall opened 6 years ago
I think adding this would be super nice. My students were actually asking me about other instance-based learning method but I didn't have a nice implementation of LOESS handy. I think roughly it should use a scikit-learn-like API, so that other methods on top of it (like cross-validation and grid search) can be used conveniently.
Maybe https://github.com/rasbt/mlxtend/blob/master/mlxtend/regressor/linear_regression.py can be used as a template?!
I guess for a first implementation, also let's not worry about KD-trees for data storage and retrieval
Later, this can maybe be refactored into a LocalRegression
class as a parent for both LOESS and LOWESS.
What do you think?
Hi @rasbt , Can I take this up?
Great! Olamyy, I say give it a go. Unless there are any objections? I use Loess in R but would like to use it more in Python.
Sent with GitHawk
Either of you implementing it would be totally fine with me :)
Suggestion to add loess regression to MLXtend? Thoughts?
Sent with GitHawk