MartInverter#

class ctis.inverters.MartInverter(instrument, gamma=None, threshold_convergence=0.001, *, num_iteration=100, intermediate=False)[source]#

Bases: AbstractIterativeInverter

An inversion routine based on the multiplicative algebraic reconstruction technique (MART) Gordon et al. [1970].

For further information, see the discussion Multiplicative Algebraic Reconstruction Technique (MART).

Attributes

`axis_iteration`	The logical axis associated with changing iteration index.
`gamma`	Learning rate, \(\gamma\).
`instrument`	A model of a CTIS instrument which transforms the radiance of an observed scene to photons measured by the sensors.
`intermediate`	Whether to save intermediate solutions.
`num_iteration`	The maximum number of iterations to perform.
`threshold_convergence`	The convergence threshold, \(T\), which halts the iteration.

Methods

`__init__`(instrument[, gamma, ...])
`correlation_residual`(images_observed, ...)	Evaluate the correlation between the predicted images and the residual.
`mean_chi_squared`(images_observed, ...)	Evaluate \(\langle \chi^2 \rangle\) for each observed/predicted image pair.

Inheritance Diagram

Parameters:

instrument (AbstractInstrument)
gamma (None | float)
threshold_convergence (float)
num_iteration (int)
intermediate (bool)

correlation_residual(images_observed, images_predicted)#

Evaluate the correlation between the predicted images and the residual.

Parameters:

images_observed (ScalarArray) – The actual measured images.
images_predicted (ScalarArray) – The images predicted by the inversion.

Return type:

ScalarArray

mean_chi_squared(images_observed, images_predicted)#

Evaluate \(\langle \chi^2 \rangle\) for each observed/predicted image pair.

Parameters:

images_observed (ScalarArray) – The actual measured images.
images_predicted (ScalarArray) – The images predicted by the inversion.

Return type:

ScalarArray

axis_iteration: ClassVar[str] = 'iteration'#: The logical axis associated with changing iteration index.

gamma: None | float = None#

Learning rate, \(\gamma\).

At every iteration, the current correction, \(C\), is replaced by \(C^\gamma\).

If None, \(\gamma = 2 / N\), where \(N\) is the number of channels.

instrument: AbstractInstrument = <dataclasses._MISSING_TYPE object>#: A model of a CTIS instrument which transforms the radiance of an observed scene to photons measured by the sensors.

intermediate: bool = False#

Whether to save intermediate solutions.

This is set to False during normal operation, but can be useful for debugging or demonstration purposes.

num_iteration: int = 100#

The maximum number of iterations to perform.

If convergence is not reached before this number is exceeded, a warning is raised and an unsuccessful result is returned.

threshold_convergence: float = 0.001#

The convergence threshold, \(T\), which halts the iteration.

If \(\langle \chi_{i-1}^2 \rangle - \langle \chi_{i}^2 \rangle < T\), then the algorithm is considered to be converged.