fdtomasi / regain Goto Github PK

Clean notebooks folder. There are many notebooks and pickles which are not needed anymore, or that can be put in specific folders, along with experimental notebooks.

Expected Behavior

New data set should be created

Actual Behavior

An error occurred

Steps to Reproduce the Problem

1. pip install regain
2. pip install all dependencies
3. run Quickstart
4. error in gaussian.py line 393: "assert (is_pos_def(theta_observed))"

Hello!

I have just found that it is not possible to generate synthetic data using your functionalities in regain/datasets/base.py

The reason is a recent sklearn update of datasets module, you might want to change line 36 in the file above as following from sklearn.utils import Bunch

I'm interested in your graphical lasso with latent variables, would be cool if you can check this issue.
Currently, I have the following error when try to reproduce your example:

`--------------------------------------------------------------------------
ModuleNotFoundError Traceback (most recent call last)
in
----> 1 from regain.datasets import datasets

~/opt/anaconda3/envs/gglasso/lib/python3.8/site-packages/regain/datasets/init.py in
28 # OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
---> 30 from .base import make_dataset

~/opt/anaconda3/envs/gglasso/lib/python3.8/site-packages/regain/datasets/base.py in
34
35 import numpy as np
---> 36 from sklearn.datasets.base import Bunch
37
38 from .gaussian import (

ModuleNotFoundError: No module named 'sklearn.datasets.base'`

pip includes deprecated version of the repo as the script to upload to pip does not work with python 2 (currently supported from the package).
As Python 2 is deprecated, we'll remove Py2 support, which should also fix this error.

Describe the bug
Attempting to import regain.linear_model raises error

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "...\lib\site-packages\regain\linear_model\__init__.py", line 31, in <module>
    from .group_lasso_overlap_ import GroupLassoOverlap, GroupLassoOverlapClassifier
  File "...\lib\site-packages\regain\linear_model\group_lasso_overlap_.py", line 39, in <module>
    from sklearn.linear_model.base import LinearClassifierMixin, LinearModel, _pre_fit
ModuleNotFoundError: No module named 'sklearn.linear_model.base'

To Reproduce
Install regain and scikit-learn >=1.0.2 (or possibly earlier version)
Attempt to import regain.linear_model

Expected behavior
No error

Screenshots
N/A

Desktop (please complete the following information):

• OS: Windows 10
• Browser: N/A
• Version: regain v0.3.3; scikit-learn v1.0.2

As sklearn v0.20, we need to import GraphicalLasso instead of GraphLasso to avoid warnings.

At the same time, it does make sense to rename also our classes.

Expected Behavior

Create Ising dataset.

Actual Behavior

Function does not terminate, hence we cannot add it in the tests.

Steps to Reproduce the Problem

from regain import datasets
datasets.make_dataset(distribution='ising')

Attempting to import GraphicalLasso (or any other module) from regain raises error:

from GL import GraphicalLasso
Results in the Module 'collections' has no attribute 'Mapping' error as in the screenshot below:

Screenshots

Getting this error when executing in AWS SageMaker notebook (Amazon Linux 2) with Python 3.10.10 | packaged by conda-forge

I have noticed the parameter over_relax in classes LatentGraphLasso, LatentTimeGraphLasso, and TimeGraphLasso cannot be changed. It will always be 1.0 (as specified in ancestor class GraphLasso). Is this a bug ?

When I try to use the TimeGraphicalLasso function on a dataset where each time point only has one sample, the function returns only diagonal matrices.

According to the authors who came up with Time-Varying Graphical Lasso (https://dl.acm.org/doi/pdf/10.1145/3097983.3098037), a Time-Varying Graphical Lasso should be "able to estimate a network at a time where there is only one observation."

Is your implementation of this function not designed to handle the extreme case where there is only one observation for each time-point? If it isn't, then it would be nice if you were to adjust this function for this extreme case. If not, are there any recommendations you can give for how to get your implementation of the time-varying Graphical Lasso to work on a dataset where there is only one observation per time point?

Thank you!