The model should not only help to apply corrections based on metadata to raw data, but it should also help to transform correction transformations into parameters that users understand.
Example: The model generates a rough correction transformation. The consuming software performs a numerical optimization of the transformation to match a reference dataset, for example a calibration measurement.
The user then wants to obtain an understandable value back from that optimized transformation and optionally save it as a new version of the relevant metadata . Example: "Your defocus was most likely x and not y" --> Attach "correction_abc" to list of available corrections --> apply "correction_abc" to following analyses.
The model should not only help to apply corrections based on metadata to raw data, but it should also help to transform correction transformations into parameters that users understand.
Example: The model generates a rough correction transformation. The consuming software performs a numerical optimization of the transformation to match a reference dataset, for example a calibration measurement.
The user then wants to obtain an understandable value back from that optimized transformation and optionally save it as a new version of the relevant metadata . Example: "Your defocus was most likely
xand noty" --> Attach "correction_abc" to list of available corrections --> apply "correction_abc" to following analyses.