Development of photonic structures using Machine Learning
The advent of multifunctional nanophotonic devices enables a wide spectrum of applications in optical metrology, quantum computing, and more. However, with their advantages come unprecedented challenges in terms of the design procedure. To unlock the full potential of these devices to match even complex optical functionalities, machine learning offers elegant solutions. Our group works with a variety of Machine Learning algorithms to enable exotic applications.
TANDEM NETWORKS Combining forward neural networks with designing neural networks, the common issue of many-to-one-mapping can be circumvented.
GENETIC ALGORITHMS Large simulation regions hinder data-driven methods from reaching their full potential. Instead, we utilize biological evolution to optimize a structure.
GENERATIVE MODELS Artificial Intelligence has the potential to outperform human bias by exploring a vast design space. This creativity helps us finding non-intuitive designs.