2023-04-26 11:00:00 | America/New_York

Saumil Bandyopadhyay MIT

Thesis Defense: Accelerating artificial intelligence with programmable silicon photonics

Advances in the fabrication of large-scale integrated silicon photonics have sparked interest in optical systems that process information at high speeds with ultra-low energy consumption. Recent demonstrations have shown these systems' ability to accelerate tasks in quantum simulation, artificial intelligence, and signal processing. In this talk, I will discuss work towards scaling up these systems to perform useful computation. I will begin by discussing the development of error correction algorithms for programmable photonic processors, whose capabilities are believed to be limited by fabrication error. By applying deterministic, gate-by-gate error correction, I show that these systems, despite relying on imprecise, analog components, can be efficiently programmed to implement highly accurate computation. I will also discuss my work towards realizing low-loss, alignment-tolerant optical interconnects, facilitating the assembly of complex photonic systems with large channel counts. Finally, I will discuss the design and demonstration of a single-chip, end-to-end photonic processor for deep neural networks (DNNs). This fully-integrated coherent optical neural network (FICONN), which monolithically integrates multiple all-optical processor units for matrix algebra and nonlinear activation functions into a single chip, implements single-shot inference across a DNN with sub-nanosecond latency. We experimentally demonstrate on-chip, in situ training of a DNN, obtaining accuracies comparable to a digital system. Our work lends experimental evidence to proposals for optically-accelerated training, enabling orders of magnitude improvements in the throughput of training data. Moreover, the FICONN opens the path to inference at nanosecond latency and femtojoule per operation energy efficiency.

Speaker's Bio

Saumil Bandyopadhyay received his S.B. and M.Eng. in Electrical Engineering from MIT in 2017 and 2018, respectively. He is a recipient of the NSF Graduate Research Fellowship and is currently with the Quantum Photonics Group at MIT, where he works on integrated silicon photonic systems for computing.