Photonics Inverse Design Over Localized Spectral and Spatial Regions

Using a region-specified residual convolutional neural network, we design a variety of nanostructures with desired scattering and absorption responses. This technique allows for single-shot training over different wavelengths of interest and is especially suitable for realizing narrowband responses and/or localized tuning of the response. In this context, we will discuss the impact of the quality of the training dataset on the accuracy of the results over different wavelength regions.