EC Lyon-DC1: Photonic Convolutional Neural Networks (CNNs)

Supervisors: Prof. Ian O’Connor (EC Lyon), Dr Fabio Pavanello (EC Lyon) and Prof. Arnan Mitchell (RMIT)

A photonic neural network (PNN) is a physical implementation of an artificial NN with photonic components. With respect to conventional NN implementations, PNNs offer several key benefits, such as orders of magnitude improvements in energy consumption, latency and bandwidth density. However, photonics- based CNNs, widely deployed for e.g., image recognition tasks, have been mostly focused on photonic accelerators due the complex task of implementing non-linear functions in photonics. The group of research topics below aims to explore how Lithium Niobate on Insulator (LNOI) platforms can be leveraged to achieve more complex functionalities missing in standard Silicon-based platforms, thus enabling the realization of the whole CNN, rather than only the acceleration portion for input pre-processing (e.g., kernel multiplication) in challenging energy-efficient edge computing applications, or for environment safety-critical applications.

Project 1: Photonic CNN for large-scale image processing

In this project we will consider implementations of high-speed low-latency photonic convolutional neural networks (CNNs) integrated on-chip based on Lithium Niobate on Insulator (LNOI) platforms. We will analyze required performance of key components which will be developed at RMIT. System-level simulations will be carried out with models built on the experimental data. Systems will be fabricated at RMIT and demonstrators will be tested at INL/RMIT for large-scale image processing applications. 

Project 2: Photonic CNN for low-power edge applications

In this project we will consider implementations of low-power energy-efficient photonic convolutional neural networks (CNNs) integrated on-chip based on Lithium Niobate on Insulator (LNOI) platforms. We will analyze required performance of key components which will be developed at RMIT. System-level simulations will be carried out with models built on the experimental data. Systems will be fabricated at RMIT and demonstrators will be tested at INL/RMIT for edge computing applications. 

Project 3: Photonic CNN for safety-critical applications

In this project we will consider implementations of robust photonic convolutional neural networks (CNNs) integrated on-chip based on Lithium Niobate on Insulator (LNOI) platforms. We will analyze required performance of key components which will be developed at RMIT. System-level simulations will be carried out with models built on the experimental data. Systems will be fabricated at RMIT and demonstrators will be tested at INL/RMIT for environment-sensitive safety-critical applications.

 

Reference

EC Lyon-DC1

Research Areas

Photonics, Computing, Neural Networks, Photonic Integrated Circuits, Optoelectronics

Research Host

École Centrale de Lyon, France

PhD awarding institution/s

École Centrale de Lyon & RMIT University

Location

France

Status

Closed Position

RMIT University

Other Positions

Supervisors

Dr. Malte Wagenfeld and Prof. Regina Bernhaupt

PhD awarding institution/s

Eindhoven University of Technology (TU/e), Netherlands and RMIT University, Australia

Location

Netherlands

Status

Closed Position

Supervisors

Dr. Carmen Mendoza Arroyo, Prof. Esther Charlesworth and Dr. Apen Ruiz Martinez (Project 1)

PhD awarding institution/s

Universitat Internacional de Catalunya (UIC) and RMIT University

Location

Spain

Status

Closed Position

Supervisors

Prof. Jesus Cerquides and Associate Prof. Jeffrey Chan and Dr. Azadeh Alavi.

PhD awarding institution/s

Autonomous University of Barcelona (UAB), Spain and RMIT  University, Australia

Location

Spain

Status

Closed Position

RMIT and many of the REDI partners are HSR4R certified
europe-1-1.svg

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 101034328.

Results reflect the author’s view only. The European Commission is not responsible for any use that may be made of the information it contains

Follow us