2022-02-02 11:00:00 | America/New_York

Christopher Alexiev Max Planck Institute of Microstructure Physics

Towards a compact photonic AI accelerator

I will present a brief overview of recent research in integrated photonics for accelerators in computation. I will review the operating principles of the Singular Value Decomposition (SVD) Mach-Zehnder Interferometer (MZI) Mesh Circuit, a particular integrated photonic system used to accelerate matrix vector multiplication, which is important for neural networks. I will then describe our laboratory’s progress toward designing and implementing a rectangular, high-speed, and large-scale 6 input SVD MZI mesh in silicon photonics technology, including the development of algorithms for efficient mesh calibration and electronic controllers for testing in AI-related applications.

Speaker's Bio

Christopher Alexiev is a fourth year undergraduate Engineering Physics student at the University of Toronto. He has completed a year-long internship at the Max Planck Institute of Microstructure Physics, where he developed integrated photonic circuits for accelerating artificial intelligence and new algorithms to calibrate these circuits. In the future he looks forward to using physics, photonics, and artificial intelligence to help end disease.