Do you want to learn how to use algorithms to automatically design and optimize optical devices? This approach is called "inverse design," and has become a very active area of research in recent years. Interestingly, the way that inverse design algorithms are able to efficiently compute gradients (through the adjoint variable method) is mathematically equivalent to the backpropagation algorithm used the machine learning community for training neural networks. Both approaches are instances of automatic differentiation!
In this interactive workshop, we will explore these connections from a practical point of view by showing you how to optimize your very own nanophotonic devices by leveraging machine learning libraries. First, we will provide a brief crash course in optical device simulation. We will then spend most of the time discussing concepts in optimization and inverse design by walking through examples in a notebook format. All code will be made available publicly in advance of the workshop so attendees may follow along as we progress. The goal of this workshop will be to provide attendees with a broad understanding of the concepts involved in inverse design and automatic differentiation, while getting a hands-on feel for code and libraries that they can immediately adapt to their own research projects.
Please sign up - spaces are limited! RSVP: forms.gle/j2k6cZGWq4GhsPye6