Montréal Quantum Photonics Seminar Series
📍 J. Armand Bombardier J-1035, Polytechnique Montréal
🗓️ Thursday, February 29th/2024
🕜 15:30
Liane Bernstein
Quantum Photonics and AI Group at MIT.
Abstract: Artificial deep neural networks (DNNs) have revolutionized tasks such as
automated classification and natural language processing. However, the
exponential growth in DNN model sizes is outpacing improvements in electronic
microprocessor efficiency. This mismatch limits DNN performance and contributes
to soaring data center energy costs. In order to overcome the limits of digital
electronics, analog optical hardware is currently being developed to boost the
speed and energy efficiency of DNNs. Here, I will present our work on large-scale,
programmable optical DNN accelerators. The focus of my talk will be on our
“single-shot” architecture, which enables ultra-low latency using standard DNN
models without retraining, allowing for plug-and-play operation
Bio: Liane Bernstein received
her B.Eng. in Engineering Physics from Polytechnique Montréal in 2016,
specializing in Photonics. In 2018, she was awarded an M.S. in Electrical
Engineering and Computer Science from MIT for “Ultrahigh-Resolution,
Deep-Penetration Spectral-Domain Optical Coherence Tomography”. For her PhD
work in the Quantum Photonics and AI Group at MIT, Liane transitioned to
optical computing, where she developed optical hardware to improve the speed
and energy efficiency of deep learning. Outside the lab, Liane loves to rock
climb and play the flute. Liane was the recipient of the Order of the White
Rose scholarship in 2016.
Contact: nicolas.quesada@polymtl.ca
