Electro-optic perceptron towards 10^18 MAC/J-efficient photonic neural networks (Conference Presentation)

2020 ◽  
Author(s):  
Rubab Amin ◽  
Mario Miscuglio ◽  
Bhavin J. Shastri ◽  
Paul R. Prucnal ◽  
Volker J. Sorger
Keyword(s):  
Neural Networks ◽  
Electro Optic

scholarly journals Reprogrammable Electro-Optic Nonlinear Activation Functions for Optical Neural Networks

2020 ◽  
Vol 26 (1) ◽  
pp. 1-12 ◽  
Author(s):  
Ian A. D. Williamson ◽  
Tyler W. Hughes ◽  
Momchil Minkov ◽  
Ben Bartlett ◽  
Sunil Pai ◽  
...  
Keyword(s):  
Neural Networks ◽  
Activation Functions ◽  
Optical Neural Networks ◽  
Electro Optic

scholarly journals All-Chalcogenide Programmable All-Optical Deep Neural Networks

2021 ◽  
Author(s):  
Ting Yu ◽  
Xiaoxuan Ma ◽  
Ernest Pastor ◽  
Jonathan George ◽  
Simon Wall ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Autonomous Vehicles ◽  
Structural Phase ◽  
Activation Function ◽  
Navigation Systems ◽  
Optical Neural Networks ◽  
Electro Optic ◽  
All Optical ◽  
On Chip

Abstract Deeplearning algorithms are revolutionising many aspects of modern life. Typically, they are implemented in CMOS-based hardware with severely limited memory access times and inefficient data-routing. All-optical neural networks without any electro-optic conversions could alleviate these shortcomings. However, an all-optical nonlinear activation function, which is a vital building block for optical neural networks, needs to be developed efficiently on-chip. Here, we introduce and demonstrate both optical synapse weighting and all-optical nonlinear thresholding using two different effects in one single chalcogenide material. We show how the structural phase transitions in a wide-bandgap phase-change material enables storing the neural network weights via non-volatile photonic memory, whilst resonant bond destabilisation is used as a nonlinear activation threshold without changing the material. These two different transitions within chalcogenides enable programmable neural networks with near-zero static power consumption once trained, in addition to picosecond delays performing inference tasks not limited by wire charging that limit electrical circuits; for instance, we show that nanosecond-order weight programming and near-instantaneous weight updates enable accurate inference tasks within 20 picoseconds in a 3-layer all-optical neural network. Optical neural networks that bypass electro-optic conversion altogether hold promise for network-edge machine learning applications where decision-making in real-time are critical, such as for autonomous vehicles or navigation systems such as signal pre-processing of LIDAR systems.

Implementation of Neural Networks: Optical Architectures and Devices

1987 ◽  
Vol 109 ◽  
Author(s):  
Kevin J. Malloy ◽  
C. Lee Giles
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Network Models ◽  
The Self ◽  
Optical Systems ◽  
Neural Network Models ◽  
Electro Optic ◽  
Device Technology

ABSTRACTWe introduce neural networks as a new basis for computation. The role optics can play In implementing neural networks is discussed in terms of the requirements for optical systems and devices. A conclusion is drawn that the rapidly evolving knowledge of neural network models argues for a flexible and adaptable device technology. Such a technology is described using the self electro-optic effect device as an example.

Novel Electro-optic Components for Integrated Photonic Neural Networks

2020 ◽  
Author(s):  
Pascal Stark ◽  
Jacqueline Geler-Kremer ◽  
Felix Eltes ◽  
Daniele Caimi ◽  
Jean Fompeyrine ◽  
...  
Keyword(s):  
Neural Networks ◽  
Electro Optic

Epitaxial ferroelectric thin films

1994 ◽  
Vol 52 ◽  
pp. 572-573
Author(s):  
S. G. Ghonge ◽  
E. Goo ◽  
R. Ramesh ◽  
R. Haakenaasen ◽  
D. K. Fork
Keyword(s):  
Single Crystal ◽  
Work Function ◽  
Nonvolatile Memory ◽  
Ferroelectric Properties ◽  
Electrical Contact ◽  
Memory Devices ◽  
Ferroelectric Films ◽  
Si Substrates ◽  
Electro Optic ◽  
Wide Range

Microstructure of epitaxial ferroelectric/conductive oxide heterostructures on LaAIO3(LAO) and Si substrates have been studied by conventional and high resolution transmission electron microscopy. The epitaxial films have a wide range of potential applications in areas such as non-volatile memory devices, electro-optic devices and pyroelectric detectors. For applications such as electro-optic devices the films must be single crystal and for applications such as nonvolatile memory devices and pyroelectric devices single crystal films will enhance the performance of the devices. The ferroelectric films studied are Pb(Zr0.2Ti0.8)O3(PLZT), PbTiO3(PT), BiTiO3(BT) and Pb0.9La0.1(Zr0.2Ti0.8)0.975O3(PLZT).Electrical contact to ferroelectric films is commonly made with metals such as Pt. Metals generally have a large difference in work function compared to the work function of the ferroelectric oxides. This results in a Schottky barrier at the interface and the interfacial space charge is believed to responsible for domain pinning and degradation in the ferroelectric properties resulting in phenomenon such as fatigue.

Sign in / Sign up

Export Citation Format

Share Document