Optimiser update

[EdoardoGiancarli]

Optimiser

Update for the Optimiser used during the IROS procedure to fit the sources (i) camera local-frame position and (ii) counts values.

• updated source model used in the optimisation;
• updated fit procedure: from minimize to curve_fit (from the scipy lib). Also removed the loss metric, since it is computed inside the curve_fit pipeline;
• inserted starting fluence correction wrt the camera coding power value [1];
• inserted narrower source local-frame coords boundaries for the fit (from tests we observed that the true source position does not deviate so much from the source peak position in the sky image --less than 2px--);

[1] Evangelista et al., "Design and performance of the coded mask for the Lunar Electromagnetic Monitor in X-rays (LEM-X)", in preparation (2025)

Tests

The optimiser tests have been commented, at the moment. They will be updated once we decide the final structure for the optimiser procedure.

Notes

With curve_fit, the optimisation procedure proved to be less computational expensive wrt minimize:

• by a factor ~6 wrt non-cached version (used with source fractional shift)
• by a factor ~3 wrt cached version for old source model

The source position and fluence residues are in the worse case similar to the ones obtained with the minimize procedure.
To resume, we improved the optimisation computational cost, while preserving the source parameters fit efficiency.

[peppedilillo]

set default to False

[EdoardoGiancarli]

Okie

[peppedilillo]

does this number depend on the mask? If so it should not be hard coded but either 1. computed from the mask pattern here, 2. red from the mask FITS file

[EdoardoGiancarli]

I agree with the second option, as the coding power is computed from simulations

[peppedilillo]

Move verbose to last argument position.

[EdoardoGiancarli]

roger that