This a penalized regression formulation of sparse linear regression using the adaptive shrinkage (Mr.ASH) prior. A VEB formulation of Mr.ASH is available here: mr.ash.alpha. Theory for Mr.AshPen: Link to Overleaf
For development, download this repository and install using pip:
git clone https://github.com/banskt/mr-ash-pen.git # or use the SSH link
cd mr-ash-pen
pip install -e .
Link to demonstration on simple examples of linear data and trend-filtering data.
Functions are not documented yet. Here is only a quick start.
from mrashpen.inference.penalized_regression import PenalizedRegression as PLR
plr = PLR(method = 'L-BFGS-B', optimize_w = True, optimize_s = True, is_prior_scaled = True, debug = False)
plr.fit()
| Returns | Description |
|---|---|
plr.coef |
optimized regression coefficients |
plr.prior |
optimized Mr.ASH prior mixture coefficients |
plr.obj_path |
Value of the objective function for all iterations |
plr.theta |
optimized parameter theta from the objective function |
plr.fitobj |
OptimizeResult object from scipy.optimize |
| --- | --- |
Run the unittest from the /path/to/download/mr-ash-pen directory.
python -m unittest