Add documentation for LassoNetAutoEncoder

ElrondL commented 1 year ago

In the example mnist_ae.py the module LassoNetAutoEncoder is imported from lassonet, but this is not in the documentation?

ElrondL commented 1 year ago

Does LassoNet have an autoencoder implementation? If not, is it possible to create one by connecting two LassoNetRegressors?

louisabraham commented 1 year ago

@ilemhadri do you know how the autoencoders were implemented in the paper?

ilemhadri commented 1 year ago

My best guess is that it was something along the lines of

from sklearn.datasets import fetch_california_housing
from sklearn.preprocessing import StandardScaler

from lassonet import LassoNetRegressor

X, _ = fetch_california_housing(return_X_y=True)
X = StandardScaler().fit_transform(X)

model = LassoNetRegressor(verbose=2)
path = model.path(X, X)

My reference is https://github.com/lasso-net/lassonet/issues/25

ElrondL commented 1 year ago

That works for me too. Is there any idea similar to a ‘latent space’ in LassoNetRegressor when used for unsupervised learning?

Sent from Mailhttps://go.microsoft.com/fwlink/?LinkId=550986 for Windows

From: @.> Sent: 04 November 2023 07:04 To: @.> Cc: @.>; @.> Subject: Re: [lasso-net/lassonet] Add documentation for LassoNetAutoEncoder (Issue #47)

As of Oct 2022, it seems this used to work

from sklearn.datasets import fetch_california_housing

from sklearn.preprocessing import StandardScaler

from lassonet import LassoNetRegressor

X, _ = fetch_california_housing(return_X_y=True)

X = StandardScaler().fit_transform(X)

model = LassoNetRegressor(verbose=2)