The Compact Opto-electronic Integrated Neural Network (COIN) coprocessor is a hardware-based neural network that uses light for passing information. It employs arrays of light emitters, arrays of photodetectors and thresholding circuits in silicon, and optical interconnection elements, stacked by layers into one system. We are exploring development at the component level before committing to the integrated chip-based system; this project involves designing and training models of the COIN. We expect to demonstrate the COIN coprocessor’s value for image processing & recognition applications and to reveal approaches to use and pitfalls to avoid for advanced iterations of the COIN.

“I’ve taken machine learning and micro/nano fabrication courses; I’m ready to continue cultivating my understanding and interest in the dynamic field of machine learning, and I’m ready to gain advanced, long-term research experience to prepare myself for my master’s. By the end of SuperUROP, I hope to publish a productive paper and to get my name and work out there.”