A generative model for computing electromagnetic field solutions
This was a final project for Stanford's CS229 course in machine learning. I designed an unsupervised machine learning model for computing approximate electromagnetic fields in a cavity containing an arbitrary spatial dielectric permit- tivity distribution. The model achieves good predictive performance and is over 10× faster than identically-sized finite-difference frequency-domain simulations, suggesting possible applications for accelerating optical inverse design algorithms. Read the paper here.
If you find this work useful, please cite
Bartlett, Ben, "A 'generative' model for computing electromagnetic field solutions", Stanford CS229 Projects (2018)
or use the BibTeX entry
@article{BartlettEM2018,
author = {Bartlett, Ben},
title = {A "generative" model for computing electromagnetic field solutions},
journal = {Stanford CS229 Projects},
volume = {2018},
number = {233},
year = {2018},
howpublished = "\url{http://cs229.stanford.edu/proj2018/report/233.pdf}",
}