Uncertainty Quantification for Inverse Deep Learning Raytracing

Kurzfassung

We present a novel approach that turns an inverse deep learning raytracer (iDLR) incorporating neural networks into a probabilistic estimator, that expands the deterministic prediction by including information about the model's uncertainty for the given sample. Thus, our approach can act as a layer of security for the iDLR during prediction of heliostat surfaces and flux desities.