Work of my master thesis "Data driven design of optical resonators: using artificial intelligence to gain insight into nanophotonic structures" at Vrije Universiteit Brussel, under supervision of Hannah Pinson and Prof. Vincent Ginis. For this work, I won the Best Physics Master thesis prize of the Belgian Physical Society (BPS). Full thesis can be found on https://arxiv.org/abs/2202.03578. It led to a publication in Nanophotonics:
J. Lenaerts, H. Pinson, V. Ginis, "Artificial Neural Networks for inverse design of resonant nanophotonic components with oscillatory loss landscapes" (2021) Nanophotonics, (10)1, 385-392 https://www.degruyter.com/view/journals/nanoph/ahead-of-print/article-10.1515-nanoph-2020-0379/article-10.1515-nanoph-2020-0379.xml
The notebook contains the code for the 2 steps of inverse design. This is step 1, the training of a neural network to predict the transmission T of the resonator and step 2, using the network to perform gradient descent on the resonator parameters. The networks we trained are found in the folder "Networks Fabry-Pérot", the file 'Total results.csv' therein gives an overview of these networks. In the folder "Inverse design Fabry-Pérot/gifs", you can find some interesting gifs of inverse design on transmissions from the test set, like the one below.
There was only one trained network in the folder "Bragg reflector". This folder also contains a subfolder 'gifs/' with some interesting gifs of inverse design on transmissions from the test set, like the one below.
