lsst.scarlet.lite.component.FactorizedComponent Class Reference

Inheritance diagram for lsst.scarlet.lite.component.FactorizedComponent:

Public Member Functions
	__init__ (self, tuple bands, Parameter|np.ndarray spectrum, Parameter|np.ndarray morph, Box bbox, tuple[int, int]|None peak=None, np.ndarray|None bg_rms=None, float|None bg_thresh=0.25, float floor=1e-20, Monotonicity|None monotonicity=None, int padding=5)
tuple[int, int]|None	peak (self)
tuple[int, int]|None	component_center (self)
np.ndarray	spectrum (self)
np.ndarray	morph (self)
tuple	shape (self)
Image	get_model (self)
	grad_spectrum (self, np.ndarray input_grad, np.ndarray spectrum, np.ndarray morph)
	grad_morph (self, np.ndarray input_grad, np.ndarray morph, np.ndarray spectrum)
np.ndarray	prox_spectrum (self, np.ndarray spectrum)
np.ndarray	prox_morph (self, np.ndarray morph)
bool	resize (self, Box model_box)
	update (self, int it, np.ndarray input_grad)
None	parameterize (self, Callable parameterization)
	__str__ (self)
	__repr__ (self)

Public Attributes
	bg_rms
	bg_thresh
	floor
	monotonicity
	padding
	morph

Protected Attributes
	_spectrum
	_morph
	_peak
	_bbox

Detailed Description

A component that can be factorized into spectrum and morphology
parameters.

Parameters
----------
bands:
    The bands of the spectral dimension, in order.
spectrum:
    The parameter to store and update the spectrum.
morph:
    The parameter to store and update the morphology.
peak:
    Location of the peak for the source.
bbox:
    The `Box` in the `model_bbox` that contains the source.
bg_rms:
    The RMS of the background used to threshold, grow,
    and shrink the component.
floor:
    Minimum value of the spectrum or center morphology pixel
    (depending on which is normalized).
monotonicity:
    The monotonicity operator to use for making the source monotonic.
    If this parameter is `None`, the source will not be made monotonic.

Definition at line 115 of file component.py.

Constructor & Destructor Documentation

__init__()

lsst.scarlet.lite.component.FactorizedComponent.__init__	(		self,
		tuple	bands,
		Parameter | np.ndarray	spectrum,
		Parameter | np.ndarray	morph,
		Box	bbox,
		tuple[int, int] | None	peak = None,
		np.ndarray | None	bg_rms = None,
		float | None	bg_thresh = 0.25,
		float	floor = 1e-20,
		Monotonicity | None	monotonicity = None,
		int	padding = 5 )

Reimplemented from lsst.scarlet.lite.component.Component.

Reimplemented in lsst.scarlet.lite.models.free_form.FreeFormComponent.

Definition at line 142 of file component.py.

    ):
        # Initialize all of the base attributes
        super().__init__(
            bands=bands,
            bbox=bbox,
        )
        self._spectrum = parameter(spectrum)
        self._morph = parameter(morph)
        self._peak = peak
        self.bg_rms = bg_rms
        self.bg_thresh = bg_thresh
 
        self.floor = floor
        self.monotonicity = monotonicity
        self.padding = padding
 

Member Function Documentation

__repr__()

lsst.scarlet.lite.component.FactorizedComponent.__repr__

(

self

)

Reimplemented in lsst.scarlet.lite.models.free_form.FreeFormComponent.

Definition at line 340 of file component.py.

    def __repr__(self):
        return self.__str__()
 
 

__str__()

lsst.scarlet.lite.component.FactorizedComponent.__str__

(

self

)

Reimplemented in lsst.scarlet.lite.models.free_form.FreeFormComponent.

Definition at line 333 of file component.py.

    def __str__(self):
        result = (
            f"FactorizedComponent<\n    bands={self.bands},\n    center={self.peak},\n    "
            f"spectrum={self.spectrum},\n    morph_shape={self.morph.shape}\n>"
        )
        return result
 

component_center()

tuple[int, int] | None lsst.scarlet.lite.component.FactorizedComponent.component_center

(

self

)

The center of the component in its bounding box

This is likely to be different than `Component.center`,
since `Component.center` is the center of the component in the
full model, whereas `component_center` is the center of the component
inside its bounding box.

Returns
-------
center:
    The center of the component in its bounding box

Definition at line 182 of file component.py.

    def component_center(self) -> tuple[int, int] | None:
        """The center of the component in its bounding box
 
        This is likely to be different than `Component.center`,
        since `Component.center` is the center of the component in the
        full model, whereas `component_center` is the center of the component
        inside its bounding box.
 
        Returns
        -------
        center:
            The center of the component in its bounding box
        """
        _center = self.peak
        if _center is None:
            return None
        center = (
            _center[0] - self.bbox.origin[-2],
            _center[1] - self.bbox.origin[-1],
        )
        return center
 

get_model()

Image lsst.scarlet.lite.component.FactorizedComponent.get_model

(

self

)

Build the model from the spectrum and morphology

Reimplemented from lsst.scarlet.lite.component.Component.

Definition at line 219 of file component.py.

    def get_model(self) -> Image:
        """Build the model from the spectrum and morphology"""
        # The spectrum and morph might be Parameters,
        # so cast them as arrays in the model.
        spectrum = self.spectrum
        morph = self.morph
        model = spectrum[:, None, None] * morph[None, :, :]
        return Image(model, bands=self.bands, yx0=cast(tuple[int, int], self.bbox.origin))
 

grad_morph()

lsst.scarlet.lite.component.FactorizedComponent.grad_morph	(		self,
		np.ndarray	input_grad,
		np.ndarray	morph,
		np.ndarray	spectrum )

Gradient of the morph wrt. the component model

Definition at line 232 of file component.py.

    def grad_morph(self, input_grad: np.ndarray, morph: np.ndarray, spectrum: np.ndarray):
        """Gradient of the morph wrt. the component model"""
        return np.einsum("i,i...", spectrum, input_grad)
 

grad_spectrum()

lsst.scarlet.lite.component.FactorizedComponent.grad_spectrum	(		self,
		np.ndarray	input_grad,
		np.ndarray	spectrum,
		np.ndarray	morph )

Gradient of the spectrum wrt. the component model

Definition at line 228 of file component.py.

    def grad_spectrum(self, input_grad: np.ndarray, spectrum: np.ndarray, morph: np.ndarray):
        """Gradient of the spectrum wrt. the component model"""
        return np.einsum("...jk,jk", input_grad, morph)
 

morph()

np.ndarray lsst.scarlet.lite.component.FactorizedComponent.morph

(

self

)

The array of morphology values

Definition at line 210 of file component.py.

    def morph(self) -> np.ndarray:
        """The array of morphology values"""
        return self._morph.x
 

parameterize()

None lsst.scarlet.lite.component.FactorizedComponent.parameterize	(		self,
		Callable	parameterization )

Convert the component parameter arrays into Parameter instances

Parameters
----------
parameterization: Callable
    A function to use to convert parameters of a given type into
    a `Parameter` in place. It should take a single argument that
    is the `Component` or `Source` that is to be parameterized.

Reimplemented from lsst.scarlet.lite.component.Component.

Definition at line 315 of file component.py.

    def parameterize(self, parameterization: Callable) -> None:
        """Convert the component parameter arrays into Parameter instances
 
        Parameters
        ----------
        parameterization: Callable
            A function to use to convert parameters of a given type into
            a `Parameter` in place. It should take a single argument that
            is the `Component` or `Source` that is to be parameterized.
        """
        # Update the spectrum and morph in place
        parameterization(self)
        # update the parameters
        self._spectrum.grad = self.grad_spectrum
        self._spectrum.prox = self.prox_spectrum
        self._morph.grad = self.grad_morph
        self._morph.prox = self.prox_morph
 

peak()

tuple[int, int] | None lsst.scarlet.lite.component.FactorizedComponent.peak

(

self

)

The peak of the component

Returns
-------
peak:
    The peak of the component

Definition at line 171 of file component.py.

    def peak(self) -> tuple[int, int] | None:
        """The peak of the component
 
        Returns
        -------
        peak:
            The peak of the component
        """
        return self._peak
 

prox_morph()

np.ndarray lsst.scarlet.lite.component.FactorizedComponent.prox_morph	(		self,
		np.ndarray	morph )

Apply a prox-like update to the morphology

Reimplemented in lsst.scarlet.lite.models.free_form.FreeFormComponent.

Definition at line 243 of file component.py.

    def prox_morph(self, morph: np.ndarray) -> np.ndarray:
        """Apply a prox-like update to the morphology"""
        # monotonicity
        if self.monotonicity is not None:
            morph = self.monotonicity(morph, cast(tuple[int, int], self.component_center))
 
        if self.bg_thresh is not None and self.bg_rms is not None:
            bg_thresh = self.bg_rms * self.bg_thresh
            # Enforce background thresholding
            model = self.spectrum[:, None, None] * morph[None, :, :]
            morph[np.all(model < bg_thresh[:, None, None], axis=0)] = 0
        else:
            # enforce positivity
            morph[morph < 0] = 0
 
        # prevent divergent morphology
        shape = morph.shape
        if self.peak is None:
            peak = (shape[0] // 2, shape[1] // 2)
        else:
            peak = (
                self.peak[0] - self.bbox.origin[-2],
                self.peak[1] - self.bbox.origin[-1],
            )
        morph[peak] = np.max([morph[peak], self.floor])
 
        # Ensure that the morphology is finite
        morph[~np.isfinite(morph)] = 0
 
        # Normalize the morphology
        max_value = np.max(morph)
        if max_value > 0:
            morph[:] = morph / max_value
        return morph
 

prox_spectrum()

np.ndarray lsst.scarlet.lite.component.FactorizedComponent.prox_spectrum	(		self,
		np.ndarray	spectrum )

Apply a prox-like update to the spectrum

Reimplemented in lsst.scarlet.lite.models.free_form.FreeFormComponent.

Definition at line 236 of file component.py.

    def prox_spectrum(self, spectrum: np.ndarray) -> np.ndarray:
        """Apply a prox-like update to the spectrum"""
        # prevent divergent spectrum
        spectrum[spectrum < self.floor] = self.floor
        spectrum[~np.isfinite(spectrum)] = self.floor
        return spectrum
 

resize()

bool lsst.scarlet.lite.component.FactorizedComponent.resize	(		self,
		Box	model_box )

Test whether or not the component needs to be resized

Reimplemented from lsst.scarlet.lite.component.Component.

Reimplemented in lsst.scarlet.lite.models.free_form.FreeFormComponent.

Definition at line 278 of file component.py.

    def resize(self, model_box: Box) -> bool:
        """Test whether or not the component needs to be resized"""
        # No need to resize if there is no size threshold.
        # To allow box sizing but no thresholding use `bg_thresh=0`.
        if self.bg_thresh is None or self.bg_rms is None:
            return False
 
        model = self.spectrum[:, None, None] * self.morph[None, :, :]
        bg_thresh = self.bg_rms * self.bg_thresh
        significant = np.any(model >= bg_thresh[:, None, None], axis=0)
        if np.sum(significant) == 0:
            # There are no significant pixels,
            # so make a small box around the center
            center = self.peak
            if center is None:
                center = (0, 0)
            new_box = Box((1, 1), center).grow(self.padding) & model_box
        else:
            new_box = (
                Box.from_data(significant, threshold=0).grow(self.padding) + self.bbox.origin
            ) & model_box
        if new_box == self.bbox:
            return False
 
        old_box = self.bbox
        self._bbox = new_box
        self._morph.resize(old_box, new_box)
        return True
 

shape()

tuple lsst.scarlet.lite.component.FactorizedComponent.shape

(

self

)

Shape of the resulting model image

Definition at line 215 of file component.py.

    def shape(self) -> tuple:
        """Shape of the resulting model image"""
        return self.spectrum.shape + self.morph.shape
 

spectrum()

np.ndarray lsst.scarlet.lite.component.FactorizedComponent.spectrum

(

self

)

The array of spectrum values

Definition at line 205 of file component.py.

    def spectrum(self) -> np.ndarray:
        """The array of spectrum values"""
        return self._spectrum.x
 

update()

lsst.scarlet.lite.component.FactorizedComponent.update	(		self,
		int	it,
		np.ndarray	input_grad )

Update the spectrum and morphology parameters

Reimplemented from lsst.scarlet.lite.component.Component.

Definition at line 307 of file component.py.

    def update(self, it: int, input_grad: np.ndarray):
        """Update the spectrum and morphology parameters"""
        # Store the input spectrum so that the morphology can
        # have a consistent update
        spectrum = self.spectrum.copy()
        self._spectrum.update(it, input_grad, self.morph)
        self._morph.update(it, input_grad, spectrum)
 

Member Data Documentation

_bbox

lsst.scarlet.lite.component.FactorizedComponent._bbox

protected

Definition at line 303 of file component.py.

_morph

lsst.scarlet.lite.component.FactorizedComponent._morph

protected

Definition at line 161 of file component.py.

_peak

lsst.scarlet.lite.component.FactorizedComponent._peak

protected

Definition at line 162 of file component.py.

_spectrum

lsst.scarlet.lite.component.FactorizedComponent._spectrum

protected

Definition at line 160 of file component.py.

bg_rms

lsst.scarlet.lite.component.FactorizedComponent.bg_rms

Definition at line 163 of file component.py.

bg_thresh

lsst.scarlet.lite.component.FactorizedComponent.bg_thresh

Definition at line 164 of file component.py.

floor

lsst.scarlet.lite.component.FactorizedComponent.floor

Definition at line 166 of file component.py.

monotonicity

lsst.scarlet.lite.component.FactorizedComponent.monotonicity

Definition at line 167 of file component.py.

morph

lsst.scarlet.lite.component.FactorizedComponent.morph

Definition at line 312 of file component.py.

padding

lsst.scarlet.lite.component.FactorizedComponent.padding

Definition at line 168 of file component.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/gae0086650b+585e252eca/python/lsst/scarlet/lite/component.py