Flexible Broad-band Optical Device

Oceanit proposes to first is to use radiative transfer techniques pioneered for astrophysics, instead of Monte Carlo simulation. In particular, we used a variant of the Bohm-Vitense method, which expresses the solution as a weighted sum of basis functions, allowing us to convert the radiative transfer equation into a set of linear equations that we can solve with standard linear algebra methods. Second, we propose to use machine learning to invert the radiative transfer. We compute a large number of radiative transfer models and then use these to train a neural net or other machine learning system, using the emitted light as input and the optical properties as output.