autogalaxy.FitEllipse#

class FitEllipse[source]#

Bases: FitDataset

A fit to a DatasetInterp dataset, using a model image to represent the observed data and noise-map.

Parameters:

  • dataset (Imaging) – The dataset containing the signal and noise-map that is fitted.

  • ellipse (Ellipse) – The ellipse profile used to fit the dataset.

  • multipole_list (Optional[List[EllipseMultipole]]) – Optional list of multipole perturbations applied to the ellipse points.

  • use_jax (bool) – If True, internal arrays are computed with JAX and downstream reductions return raw JAX arrays. The algorithm is identical to the NumPy path — both use a single perimeter-sampling pass with NaN-marking for masked positions.

Methods

points_from_major_axis_from

Return the (N, 2) (y, x) coordinates on the ellipse perimeter where interpolation occurs.

Attributes

chi_squared

The sum of the chi-squared-map, which quantifies how well the model data represents the data and noise-map.

chi_squared_map

Returns the chi-squared-map of the fit, which is the normalized residual-map squared.

data

The data of the dataset being fitted.

data_interp

Returns the data values of the dataset that the ellipse fits, which are computed by overlaying the ellipse over the 2D data and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse.

figure_of_merit

The figure of merit of the fit, which is passed by the Analysis class to the non-linear search to determine the best-fit solution.

grids

The grids of (y,x) coordinates associated with the dataset, adjusted by any grid_offset and grid_rotation_angle specified in the dataset_model.

interp

A property that is only computed once per instance and then replaces itself with an ordinary attribute.

inversion

Overwrite this method so it returns the inversion used to fit the dataset.

log_evidence

Returns the log Bayesian evidence of the inversion's fit to a dataset, which extends the log likelihood by including penalty terms that quantify the complexity of the inversion's reconstruction:

log_likelihood

The log likelihood of the fit, which quantifies how well the model data represents the data and noise-map.

log_likelihood_with_regularization

Returns the log likelihood of an inversion's fit to the dataset, including a regularization term which comes from an inversion:

mask

The 2D mask of the dataset being fitted, where False entries are unmasked and included in the fit and True entries are masked and excluded.

model_data

Returns the model-data, which is the data values where the ellipse overlaps the data minus the mean value of these data values.

noise_map

The noise-map of the dataset being fitted, representing the RMS noise in each pixel.

noise_map_interp

Returns the noise-map values of the dataset that the ellipse fits, which are computed by overlaying the ellipse over the 2D noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse.

noise_normalization

The noise normalization term of the log likelihood, which is the sum of the log noise-map values squared.

normalized_residual_map

Returns the normalized residual-map of the fit, which is the residual-map divided by the noise-map.

reduced_chi_squared

The reduced chi-squared of the model's fit to the dataset, defined as:

residual_flux_fraction_map

Returns the residual flux fraction map, which shows the fraction of signal in each pixel that is not fitted by the model, therefore where:

residual_map

Returns the residual-map of the fit, which is the data minus the model data and therefore the same as the model data.

signal_to_noise_map

The signal-to-noise_map of the dataset and noise-map which are fitted.

signal_to_noise_map_interp

Returns the signal-to-noise-map of the dataset that the ellipse fits, which is computed by overlaying the ellipse over the 2D data and noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse.
interp#

A property that is only computed once per instance and then replaces itself with an ordinary attribute. Deleting the attribute resets the property.

Source: https://github.com/bottlepy/bottle/commit/fa7733e075da0d790d809aa3d2f53071897e6f76

points_from_major_axis_from()[source]#

Return the (N, 2) (y, x) coordinates on the ellipse perimeter where interpolation occurs.

Sampling proceeds in one shot:

  1. N = Ellipse.total_points_from(pixel_scale) perimeter angles are generated.

  2. Multipole perturbations (if any) are applied in-place.

  3. The mask interpolator is evaluated at every point; positions that sit on (or partly on) masked pixels are marked as NaN via xp.where(keep, points, nan).

The output shape is fixed at (N, 2) regardless of mask geometry — masked rows are NaN, unmasked rows hold their (y, x) coordinates. Downstream reductions (residual_map, chi_squared, noise_normalization) use nanmean / nansum and skip NaN positions correctly.

The algorithm is identical for self.use_jax False and True; only the numerics backend (np vs jax.numpy) differs.

Return type:

The (N, 2) (y, x) coordinates with NaN at masked positions.

property data_interp: ArrayIrregular#

Returns the data values of the dataset that the ellipse fits, which are computed by overlaying the ellipse over the 2D data and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse.

The (y,x) coordinates on the ellipse where the interpolation occurs are computed in the points_from_major_axis property of the Ellipse class, with the documentation describing how these points are computed.

If the interpolation of an ellipse point uses one or more masked values, this point is not reliable, therefore the value is converted to np.nan and not used by other fitting quantities.

Returns:

  • The data values of the ellipse fits, computed via a 2D interpolation of where the ellipse

  • overlaps the data.

property noise_map_interp: ArrayIrregular#

Returns the noise-map values of the dataset that the ellipse fits, which are computed by overlaying the ellipse over the 2D noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse.

The (y,x) coordinates on the ellipse where the interpolation occurs are computed in the points_from_major_axis property of the Ellipse class, with the documentation describing how these points are computed.

If the interpolation of an ellipse point uses one or more masked values, this point is not reliable, therefore the value is converted to np.nan and not used by other fitting quantities.

Returns:

  • The noise-map values of the ellipse fits, computed via a 2D interpolation of where the ellipse

  • overlaps the noise-map.

property signal_to_noise_map_interp: ArrayIrregular#

Returns the signal-to-noise-map of the dataset that the ellipse fits, which is computed by overlaying the ellipse over the 2D data and noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse.

Returns:

  • The signal-to-noise-map values of the ellipse fits, computed via a 2D interpolation of where

  • the ellipse overlaps the data and noise-map.

property model_data: ArrayIrregular#

Returns the model-data, which is the data values where the ellipse overlaps the data minus the mean value of these data values.

By subtracting the mean of the data values from each data value, the model data quantifies how often there are large variations in the data values over the ellipse.

For example, if every data value subtended by the ellipse are close to one another, the difference between the data values and the mean will be small.

Conversely, if some data values are much higher or lower than the mean, the model data will be large.

Return type:

The model data values of the ellipse fit, which are the data values minus the mean of the data values.

property residual_map: ArrayIrregular#

Returns the residual-map of the fit, which is the data minus the model data and therefore the same as the model data.

Return type:

The residual-map of the fit, which is the data minus the model data and therefore the same as the model data.

property normalized_residual_map: ArrayIrregular#

Returns the normalized residual-map of the fit, which is the residual-map divided by the noise-map.

The residual map and noise map are computed by overlaying the ellipse over the 2D data and noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse. See the documentation of the residual_map and noise_map properties for more details.

Return type:

The normalized residual-map of the fit, which is the residual-map divided by the noise-map.

property chi_squared_map: ArrayIrregular#

Returns the chi-squared-map of the fit, which is the normalized residual-map squared.

The normalized residual map is computed by overlaying the ellipse over the 2D data and noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse. See the documentation of the normalized_residual_map property for more details.

Return type:

The chi-squared-map of the fit, which is the normalized residual-map squared.

property chi_squared: float#

The sum of the chi-squared-map, which quantifies how well the model data represents the data and noise-map.

The chi-squared-map is computed by overlaying the ellipse over the 2D data and noise-map and performing a 2D interpolation at discrete (y,x) coordinates on the ellipse. See the documentation of the chi_squared_map property for more details.

Return type:

The chi-squared of the fit.

property noise_normalization#

The noise normalization term of the log likelihood, which is the sum of the log noise-map values squared.

Return type:

The noise normalization term of the log likelihood.

property log_likelihood#

The log likelihood of the fit, which quantifies how well the model data represents the data and noise-map.

Return type:

The log likelihood of the fit.

property figure_of_merit: float#

The figure of merit of the fit, which is passed by the Analysis class to the non-linear search to determine the best-fit solution.

Return type:

The figure of merit of the fit.