Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Photonic Neural Networks: A Compact Review

Mohammad Ahmadi, Hamidreza Bolhasani

TL;DR

This paper surveys photonic neural networks as a route to ultrafast, energy-efficient neural computation, partitioning the literature into experimental, theoretical, and mathematical contributions from 2015–2023. It analyzes core architectures and concepts—Integrated Photonics Neural Networks (IPNN), neuromorphic photonics, optical convolutional networks (CNNs), optical neural networks (ONNs), and three-dimensional photonic interconnects—and reviews progress in integrated photonics and integrated quantum photonics (IQP). The discussion highlights key advantages such as real-time processing and parallelism, alongside challenges like integration with existing electronics, scalability, and fabrication complexity. The work aims to guide future design, standardization, and deployment of photonic neural networks for high-speed AI hardware and quantum photonics applications.

Abstract

It has long been known that photonic science and especially photonic communications can raise the speed of technologies and producing manufacturing. More recently, photonic science has also been interested in its capabilities to implement low-precision linear operations, such as matrix multiplications, fast and effciently. For a long time most scientists taught that Electronics is the end of science but after many years and about 35 years ago had been understood that electronics do not answer alone and should have a new science. Today we face modern ways and instruments for doing tasks as soon as possible in proportion to many decays before. The velocity of progress in science is very fast. All our progress in science area is dependent on modern knowledge about new methods. In this research, we want to review the concept of a photonic neural network. For this research was selected 18 main articles were among the main 30 articles on this subject from 2015 to the 2022 year. These articles noticed three principles: 1- Experimental concepts, 2- Theoretical concepts, and, finally 3- Mathematic concepts. We should be careful with this research because mathematics has a very important and constructive role in our topics! One of the topics that are very valid and also new, is simulation. We used to work with simulation in some parts of this research. First, briefly, we start by introducing photonics and neural networks. In the second we explain the advantages and disadvantages of a combination of both in the science world and industries and technologies about them. Also, we are talking about the achievements of a thin modern science. Third, we try to introduce some important and valid parameters in neural networks. In this manner, we use many mathematic tools in some portions of this article.

Photonic Neural Networks: A Compact Review

TL;DR

This paper surveys photonic neural networks as a route to ultrafast, energy-efficient neural computation, partitioning the literature into experimental, theoretical, and mathematical contributions from 2015–2023. It analyzes core architectures and concepts—Integrated Photonics Neural Networks (IPNN), neuromorphic photonics, optical convolutional networks (CNNs), optical neural networks (ONNs), and three-dimensional photonic interconnects—and reviews progress in integrated photonics and integrated quantum photonics (IQP). The discussion highlights key advantages such as real-time processing and parallelism, alongside challenges like integration with existing electronics, scalability, and fabrication complexity. The work aims to guide future design, standardization, and deployment of photonic neural networks for high-speed AI hardware and quantum photonics applications.

Abstract

It has long been known that photonic science and especially photonic communications can raise the speed of technologies and producing manufacturing. More recently, photonic science has also been interested in its capabilities to implement low-precision linear operations, such as matrix multiplications, fast and effciently. For a long time most scientists taught that Electronics is the end of science but after many years and about 35 years ago had been understood that electronics do not answer alone and should have a new science. Today we face modern ways and instruments for doing tasks as soon as possible in proportion to many decays before. The velocity of progress in science is very fast. All our progress in science area is dependent on modern knowledge about new methods. In this research, we want to review the concept of a photonic neural network. For this research was selected 18 main articles were among the main 30 articles on this subject from 2015 to the 2022 year. These articles noticed three principles: 1- Experimental concepts, 2- Theoretical concepts, and, finally 3- Mathematic concepts. We should be careful with this research because mathematics has a very important and constructive role in our topics! One of the topics that are very valid and also new, is simulation. We used to work with simulation in some parts of this research. First, briefly, we start by introducing photonics and neural networks. In the second we explain the advantages and disadvantages of a combination of both in the science world and industries and technologies about them. Also, we are talking about the achievements of a thin modern science. Third, we try to introduce some important and valid parameters in neural networks. In this manner, we use many mathematic tools in some portions of this article.
Paper Structure (15 sections, 7 equations, 4 figures, 6 tables)

This paper contains 15 sections, 7 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Fundamental Concepts
  3. Roadmap on material-function mapping
  4. Integrated Photonics
  5. Neuromorphic Photonics
  6. Convolutional Neural Network (CNN)
  7. Optical Neural Networks (ONNs)
  8. Three-dimentional waveguide interconnects
  9. Large-scale Optical Neural Networks
  10. Progress and Development of Photonic Neural Networks
  11. Application in Integrated Photonics
  12. Application in IQP (Integrated Quantum Photonic)
  13. Discussion
  14. References
  15. Authors

Figures (4)

  • Figure 1: Schematic for Neural Network includes First Colomn: Input Layers, Second And Third Colomn: Hidden Layers, Final part: Output Layer.
  • Figure 2: Application of DNNs (Deep Neural Networks) for VCSEL (vertical-cavity surface-emitting laser) [71].
  • Figure 3: A schematic for Convolution Neural Network (CNN)[11].
  • Figure 4: A brief history and outlook of Photonic Neural Networks.