Wednesday |
Mapping AI Workloads to a Photonic Matrix MultiplierDr. Gilbert Hendry, Lightelligence, Boston, MA | |
Abstract: With the end of Moore’s Law on the horizon, the computing industry is turning to new technologies and vastly different architectures to meet tomorrow’s demands in areas such as image recognition, natural language processing, recommendation, and scientific simulation. At Lightelligence, we are working on developing hardware to perform vector-matrix multiplication at the speed of light using highly integrated silicon photonics, significantly speeding up many modern workloads that use linear algebra. However, there are many constraints and challenges that come with working with a partially analog computer, including reduced precision and noise, that require attention at all levels of design from algorithms to circuits. In this talk I will discuss how AI and other linear algebra workloads execute on our compute platform and how we envision the future of silicon photonics as compute accelerators. Biography: Gilbert Hendry is the VP of Software at Lightelligence where he works on all things AI, software engineering, computer architecture, and hardware-software codesign. Previously he was at Microsoft and Google working on various software engineering, data science, and machine learning engineering projects. He received a PhD from Columbia University where he worked on silicon photonic network- on-chip design and simulation, and a BS and MS from RIT where he studied Computer Engineering. |
IEEE Photonics Society
Boston Photonics Society Chapter
Boston Chapter of the IEEE Photonics Society
Machine Learning and Optical Systems
Wednesday, October 7, 14, 21, 28 and November 4, 2020, 7:00-9:30 PM
Located at Online Seminar
The workshop expenses have |
For more information on the technical content of the workshop, contact either:
1) Keisuke Kojima, (kojima@merl.com), Chair
2) Ajay Garg, (ajay.sinclair.garg@ieee.org), Co-Chair
3) Dean Tsang, (tsang@ieee.org), Co-Chair
4) Bill Nelson, (w.nelson@ieee.org), Co-Chair
Designed and built by Matthew Emsley, © 2020