lsst.scarlet.lite.parameters.FistaParameter Class Reference

Inheritance diagram for lsst.scarlet.lite.parameters.FistaParameter:

Public Member Functions
	__init__ (self, np.ndarray x, float step, Callable|None grad=None, Callable|None prox=None, float t0=1, np.ndarray|None z0=None)
	update (self, int it, np.ndarray input_grad, *args)

Public Attributes
	t

Detailed Description

A `Parameter` that updates itself using the Beck-Teboulle 2009
FISTA proximal gradient method.

See https://www.ceremade.dauphine.fr/~carlier/FISTA

Definition at line 183 of file parameters.py.

Constructor & Destructor Documentation

__init__()

lsst.scarlet.lite.parameters.FistaParameter.__init__	(		self,
		np.ndarray	x,
		float	step,
		Callable | None	grad = None,
		Callable | None	prox = None,
		float	t0 = 1,
		np.ndarray | None	z0 = None )

Reimplemented from lsst.scarlet.lite.parameters.Parameter.

Definition at line 190 of file parameters.py.

    ):
        if z0 is None:
            z0 = x
 
        super().__init__(
            x,
            {"z": z0},
            step,
            grad,
            prox,
        )
        self.t = t0
 

Member Function Documentation

update()

lsst.scarlet.lite.parameters.FistaParameter.update	(		self,
		int	it,
		np.ndarray	input_grad,
		*	args )

Update the parameter and meta-parameters using the PGM

See `Parameter` for the full description.

Reimplemented from lsst.scarlet.lite.parameters.Parameter.

Definition at line 211 of file parameters.py.

    def update(self, it: int, input_grad: np.ndarray, *args):
        """Update the parameter and meta-parameters using the PGM
 
        See `Parameter` for the full description.
        """
        step = self.step / np.sum(args[0] * args[0])
        _x = self.x
        _z = self.helpers["z"]
 
        y = _z - step * cast(Callable, self.grad)(input_grad, _x, *args)
        if self.prox is not None:
            x = self.prox(y)
        else:
            x = y
        t = 0.5 * (1 + np.sqrt(1 + 4 * self.t**2))
        omega = 1 + (self.t - 1) / t
        self.helpers["z"] = _x + omega * (x - _x)
        _x[:] = x
        self.t = t
 
 
# The following code block contains different update methods for
# various implementations of ADAM.
# We currently use the `amsgrad_phi_psi` update by default,
# but it can easily be interchanged by passing a different
# variant name to the `AdaproxParameter`.
 
 
# noinspection PyUnusedLocal

Member Data Documentation

t

lsst.scarlet.lite.parameters.FistaParameter.t

Definition at line 209 of file parameters.py.

---

The documentation for this class was generated from the following file:

• /j/snowflake/release/lsstsw/stack/lsst-scipipe-8.0.0/Linux64/scarlet_lite/gee10cc3b42+a6fc98d2e7/python/lsst/scarlet/lite/parameters.py