Learning the lantern: neural network applications to broadband photonic lantern modeling

Photonic lanterns allow the decomposition of highly multimodal light into a simplified modal basis such as single-moded and/or few-moded. They are increasingly finding uses in astronomy, optics and telecommunications. Calculating propagation through photonic lantern using traditional algorithms takes $\sim 1$ hour per simulation on modern CPU. This paper demonstrates that neural networks can bridge disparate opto-electronic systems, when trained achieve speed-up over 5 orders magnitude. We show this approach be used to model with manufacturing defects well successfully generalising polychromatic data. demonstrate two these network models, propagating seeing performing global optimisation for purposes funnels nullers.