Photonic Cognitive Processor for Next Generation Artificial Intelligence Hardware

Artificial Intelligence (AI) is transforming our lives in the same way as the advent of the Internet and cellular phones has done. However, it takes thousands of CPUs and GPUs, and many weeks to train the neural networks in AI hardware. Traditional CPUs, GPUs, and brain-inspired electronics will not be powerful enough to train the neural networks of the near future. To radically impact the next generation of AI hardware, I propose to develop a fundamental technology: a photonic cognitive processor that uses light (instead of electrons). By employing photonic networks, I will test my processor with standard benchmark tasks on pattern recognition such as MNIST. I propose to use a special purpose GEMM compiler that will efficiently perform small matrix multiplications on enormous matrices. Results will be compared to electronic counterparts in terms of speed, precision, and energy efficiency.

Faculty Supervisor:

Bhavin J. Shastri

Student:

Bicky A. Marquez Alfonzo

Partner:

Huawei Technologies Canada

Discipline:

Engineering

Sector:

Professional, scientific and technical services

University:

Queen's University