Evan Lerner
The Bell Labs Prize awarded by Nokia Bell Labs is a yearly contest in which researchers from around the world submit proposals for “disruptive innovations that will define the next industrial revolution.” Participants are paired with Nokia Bell Labs researchers to refine their ideas, ultimately presenting them to a panel of experts and industry leaders.
This year, Firooz Aflatouni, associate professor in the Department of Electrical and Systems Engineering, in collaboration with Farshid Ashtiani, a postdoctoral scholar in Aflatouni’s research group won the top prize for their proposal for “Integrated Photonic-mmWave Deep Networks.”
Aflatouni’s expertise is in the design of photonic-electronic chips, which use light to transmit and process information akin to the way traditional computer chips use electricity. The first prize-winning proposal presents a deep neural network that uses a photonic platform to perform computations as light travels through the system, theoretically allowing for image and video recognition at almost the speed of light.
“Our system performs image recognition in the optical and millimeter-wave domains,” says Aflatouni, “which in addition to being very fast, eliminates optical to electrical signal conversion, analog-to-digital conversion, and the need for a large memory module. The goal is not to replace the current deep learning digital hardware platforms, but to introduce a low-cost photonic millimeter-wave co-processor or accelerator to significantly reduce the computation time while maintaining or improving the classification accuracy.”
Such a system could be integrated with cellphones, among other devices, and enable applications like facial recognition or augmented reality at a much faster speed than today’s digital computational platforms.