In this repository we evaluate the WaveDiff model and analyse how errors in the SEDs impact on the prediction. This repository contains one folder for each test or experiment which include all the codes and data used to carry out the tests.
Analisis of the impact of the data wavefront-error resolution on the model predictions.
Analisis of the impact of the model wavefront-error resolution on the predictions.
Optimising the model with 10k stars. Five times the number of stars used on WaveDiff.
First study the SED binning errors impact on the PSF generation. Then implement SED interpolation to better approximate the ground truth data.
Tweaking some of the model hyperparameters and adjust the optimisation routine for 10k stars.