FEAT - Support for Poisson Group Lasso

[jonpedros]

Hi skglm team,

I'm trying to fit a Poisson Group Lasso model by combining the Poisson datafit, WeightedGroupL2 penalty, and the GroupBCD solver.

This fails with

ValueError: 'Poisson' is not block-separable. Missing 'grp_ptr' attribute., because the standard Poisson datafit isn't compatible with the group solver.

Is there a current workaround for this? If not, please consider this a feature request to add a PoissonGroup datafit, similar to the existing QuadraticGroup and LogisticGroup. This would be a great addition for modeling count data with grouped features.

import numpy as np
from skglm import GeneralizedLinearEstimator
from skglm.datafits import Poisson
from skglm.penalties import WeightedGroupL2
from skglm.solvers import GroupBCD

# Sample data and group structure
n_samples, n_features = 20, 10
X = np.random.randn(n_samples, n_features)
y = np.random.poisson(np.abs(X[:, 0] + X[:, 5]))

grp_ptr = np.array([0, 3, 5, 8, 10], dtype=np.int32)
grp_indices = np.arange(n_features, dtype=np.int32)

# Estimator setup
estimator = GeneralizedLinearEstimator(
    datafit=Poisson(),
    penalty=WeightedGroupL2(alpha=0.1, grp_ptr=grp_ptr, grp_indices=grp_indices,
                            weights=np.ones(len(grp_ptr) - 1)),
    solver=GroupBCD()
)

estimator.fit(X, y)