lsst.scarlet.lite.models.fit_psf.FittedPsfBlend Class Reference

Inheritance diagram for lsst.scarlet.lite.models.fit_psf.FittedPsfBlend:

Public Member Functions
tuple[int, float]	fit (self, int max_iter, float e_rel=1e-4, int min_iter=1, int resize=10)

Public Attributes
	bbox
	observation
	it

Protected Member Functions
Image	_grad_log_likelihood (self)

Detailed Description

A blend that attempts to fit the PSF along with the source models.

Definition at line 148 of file fit_psf.py.

Member Function Documentation

_grad_log_likelihood()

Image lsst.scarlet.lite.models.fit_psf.FittedPsfBlend._grad_log_likelihood ( self )

protected

Gradient of the likelihood wrt the unconvolved model

Reimplemented from lsst.scarlet.lite.blend.Blend.

Definition at line 151 of file fit_psf.py.

    def _grad_log_likelihood(self) -> Image:
        """Gradient of the likelihood wrt the unconvolved model"""
        model = self.get_model(convolve=True)
        # Update the loss
        self.loss.append(self.observation.log_likelihood(model))
        # Calculate the gradient wrt the model d(logL)/d(model)
        result = self.observation.weights * (model - self.observation.images)
        return result
 

fit()

tuple[int, float] lsst.scarlet.lite.models.fit_psf.FittedPsfBlend.fit	(		self,
		int	max_iter,
		float	e_rel = 1e-4,
		int	min_iter = 1,
		int	resize = 10 )

Fit all of the parameters

Parameters
----------
max_iter: int
    The maximum number of iterations
e_rel: float
    The relative error to use for determining convergence.
min_iter: int
    The minimum number of iterations.
resize: int
    Number of iterations before attempting to resize the
    resizable components. If `resize` is `None` then
    no resizing is ever attempted.

Reimplemented from lsst.scarlet.lite.blend.Blend.

Definition at line 160 of file fit_psf.py.

    ) -> tuple[int, float]:
        """Fit all of the parameters
 
        Parameters
        ----------
        max_iter: int
            The maximum number of iterations
        e_rel: float
            The relative error to use for determining convergence.
        min_iter: int
            The minimum number of iterations.
        resize: int
            Number of iterations before attempting to resize the
            resizable components. If `resize` is `None` then
            no resizing is ever attempted.
        """
        it = self.it
        while it < max_iter:
            # Calculate the gradient wrt the on-convolved model
            grad_log_likelihood = self._grad_log_likelihood()
            _grad_log_likelihood = self.observation.convolve(grad_log_likelihood, grad=True)
            # Check if resizing needs to be performed in this iteration
            if resize is not None and self.it > 0 and self.it % resize == 0:
                do_resize = True
            else:
                do_resize = False
            # Update each component given the current gradient
            for component in self.components:
                overlap = component.bbox & self.bbox
                component.update(it, _grad_log_likelihood[overlap].data)
                # Check to see if any components need to be resized
                if do_resize:
                    component.resize(self.bbox)
 
            # Update the PSF
            cast(FittedPsfObservation, self.observation).update(
                it, grad_log_likelihood.data, self.get_model()
            )
            # Stopping criteria
            it += 1
            if it > min_iter and np.abs(self.loss[-1] - self.loss[-2]) < e_rel * np.abs(self.loss[-1]):
                break
        self.it = it
        return it, self.loss[-1]

Member Data Documentation

bbox

lsst.scarlet.lite.models.fit_psf.FittedPsfBlend.bbox

Definition at line 198 of file fit_psf.py.

it

lsst.scarlet.lite.models.fit_psf.FittedPsfBlend.it

Definition at line 208 of file fit_psf.py.

observation

lsst.scarlet.lite.models.fit_psf.FittedPsfBlend.observation

Definition at line 201 of file fit_psf.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/models/fit_psf.py