Matrix Multiplication Using Mode-Division Multiplexing and Coherent Detection (Topic B.2.2 Optoelectronics)

Abstract

The objective of this research program is to perform a proof-of-concept demonstration for a new paradigm for optical matrix-vector multiplications. When performed in the electrical domain, matrix-vector multiplications are decomposed into multiply-accumulate (MAC) operations. In the new paradigm proposed here, multiplications are performed through coherent mixing, and accumulation is performed in the spatial domain using mode-division multiplexing. Therefore, our MAC operations are passive, consume only power associated with data modulation. Combined with wavelength-division multiplexing and 3-dimensional spatial parallelization, this concept can be extended to batch matrix-matrix multiplication, which is a foundational operation in artificial neural networks.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 24, 2019
Source ID
W911NF1910385

Entities

People

  • Guifang Li

Organizations

  • Army Contracting Command
  • United States Army
  • University of Central Florida

Tags

Readers

  • Electronics Engineering
  • Integrated Circuit Design and Technology.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • Microelectronics