TR2017-097

Acceleration of FDTD-based Inverse Design Using a Neural Network Approach

    •  Kojima, K., Wang, B., Kamilov, U., Koike-Akino, T., Parsons, K., "Acceleration of FDTD-based Inverse Design Using a Neural Network Approach", Integrated Photonics Research, Silicon and Nano Photonics (IPR), July 2017.
      BibTeX TR2017-097 PDF
      • @inproceedings{Kojima2017jul,
      • author = {Kojima, Keisuke and Wang, Bingnan and Kamilov, Ulugbek and Koike-Akino, Toshiaki and Parsons, Kieran},
      • title = {Acceleration of FDTD-based Inverse Design Using a Neural Network Approach},
      • booktitle = {Integrated Photonics Research, Silicon and Nano Photonics (IPR)},
      • year = 2017,
      • month = jul,
      • url = {https://www.merl.com/publications/TR2017-097}
      • }
  • MERL Contacts:
  • Research Areas:

    Communications, Signal Processing

Abstract:

Instead of using FDTD simulations for all the inverse design steps, we proposed to use neural network-based fitting to estimate the output of the FDTD simulations, and improve the design. We observed clear acceleration in the improvement of metric.