Cs 229 Final Project: Structure Prediction of Optical Functional Devices with Deep Learning

Improving the efficiency of optical device is important in a variety of applications. However, this process is typically done empirically. In this project, we formulate a systematically way to improve the efficiency of optical device design with machine learning. The optical device we focus on is a silicon (Si) nano-structure. This device is constructed by a series of Si nano-cells, which is represented as the {0, 1} binary series in Fig. 1. In Fig. 1, 1 and 0 indicate the existence and void of Si at a particular location. A beam of light perpendicular to the Si structure is incident onto the Si structure and excites the surrounding electric field. In this project, we use the output electric field to predict its corresponding Si structure.