Maxwell’s Equations for Machine Vision and Machine Learning
A materials science and engineering / electrical engineering joint seminar with Prof. Zongfu Yu of University of Wisconsin Madison.
Maxwell’s equations, which describe the flow of electromagnetic wave, were discovered 150 years ago. Today, they are used everywhere in our information age: WiFi antennas, smart phone cameras, as well as internet backbone. As we embrace the age of machine intelligence, Maxwell’s equations could find new life in smart things. We will show a few examples of how Maxwell’s equations could empower future machines.
First, we discuss how Maxwell’s equations is used to accelerate machine learning to allow light to perform artificial intelligence computing inside a nanostructured material with unprecedented speed and energy efficiency. In the second part, we discuss a new class of imaging sensors that are functionalized to perceive scenes more intelligently. They exploit the unique effects of Maxwell’s equations at the nanoscale to measure high-resolution phase front and spectra. Such information is traditionally measured by tabletop setups. Miniaturizing such capability onto a chip allows future machines to make mission-critical decision in many industrial and bio-medical applications. Lastly, we show a fun experiment that combines electromagnetics and reinforcement learning to realize Magneto’s superpower as shown in the X-men movies.