Optical computing uses photons instead of electrons to perform calculations. This can significantly improve computational speed and energy efficiency by overcoming the inherent limitations of electrons. The basic principle of optical computing is the interaction of light and matter. Matrix computing has become one of the most widely used and indispensable information processing tools in science and engineering, providing numerous computational tasks for most signal processing, such as discrete Fourier transforms and convolution operations.
As a basic component of Artificial neural network (ANNs), Matrix multiplication Occupies most computational resources. Due to the characteristics of electronic components, performing simple matrix multiplication requires a large number of transistors to work together, but matrix multiplication is basic such as microrings, Mach-Zehnder interferometers (MZIs), diffraction planes, etc. Easy to implement with photonic components. therefore, Optical computing It is orders of magnitude faster and consumes much less power than electronic computing. However, traditional incoherent matrix-vector multiplication focuses on real-value operations and does not work well with complex-value neural networks or discrete Fourier transforms.
Researchers led by Professor Jianji Dong of Huazhong University of Science and Technology (HUST) in China have proposed a photonic complex matrix-vector multiplier chip that can support matrix multiplication of any large scale complex numbers. This chip eliminates the bottleneck of traditional non-coherent optical computing schemes where it is difficult to achieve matrix multiplication of any large complex number. It also enables artificial intelligence applications such as the Discrete Fourier Transform, the Discrete Cosine Transform, the Walsh Transform, and the image convolution.
Their idea is to design matrix factorization and matrix factorization intelligent algorithms for the Microring Array architecture to extend matrix multiplication from real to complex and from small to large. Researchers have succeeded in experimentally demonstrating some typical artificial intelligence applications, demonstrating the potential of photonic composite matrix-vector multiplier chips for artificial intelligence computing applications. A work entitled “Matrix of Large Complex Numbers-Small Microring Arrays Performing Vector Multiplication” was published on April 28, 2022. Frontier of Optoelectronics..
Junwei Cheng et al, a matrix of large complex numbers-a small microring array that performs vector multiplication, Frontier of Optoelectronics (2022). DOI: 10.1007 / s12200-022-00009-4
Higher Education Press Provided
Quote: Matrix multiplication of large complex numbers obtained on May 12, 2022 from https: //phys.org/news/2022-05-small-microring-array-enables-large.html by a small microring array ( 2022, May 12th) will be available
This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. Content is provided for informational purposes only.
Small microring arrays allow matrix multiplication of large complex numbers
Source link Small microring arrays allow matrix multiplication of large complex numbers