Penalties

Penalty objects are the low-level interface behind selection_penalty= and spline_penalty=. Prefer the model-level arguments; reach for these classes when you need a specific group structure or an adaptive weighting.

superglm.GroupElasticNet	Group elastic net: lambda1 * [alpha * group_L2 + (1-alpha)/2 * L2^2].
superglm.GroupLasso	Group lasso penalty: lambda1 * sum_g(w_g * ||beta_g||_2).
superglm.SparseGroupLasso	Sparse group lasso: lambda1 * [(1-alpha) * group_L2 + alpha * L1].
superglm.Ridge	Ridge penalty: lambda1 * ||beta||_2^2 / 2.
superglm.Adaptive	Adaptive weighting flavor.