shrinkage.ProximalGradient

Shrinkage for Python · python · 0.2.0

ProximalGradient(
    *,
    stopping: StoppingCriterion | None = None,
    max_iterations: int = 10000,
    initial_step: float = 1.0,
    max_backtracks: int = 100,
    screening: bool = True
)

Backtracking PG on the original Gaussian or binomial objective.

max_iterations limits accepted proximal updates. initial_step is finite and positive, and max_backtracks limits step halvings per iteration, with zero permitting only the first trial. stopping=None uses absolute full-space mapping tolerance 1e-6. Gap and KKT criteria are rejected. Screening is enabled by default; heuristic working sets are available through HybridCoordinateDescent.

stopping: StoppingCriterion | None = field(default=None, kw_only=True)

max_iterations: int = field(default=10_000, kw_only=True)

initial_step: float = field(default=1.0, kw_only=True)

max_backtracks: int = field(default=100, kw_only=True)

screening: bool = field(default=True, kw_only=True)