Optimal prediction for sparse linear models? Lower bounds for coordinate-separable M-estimators

Open in new window