Artificial neural networks, as physical systems, are implemented through hardware substrates that realize interconnected processing units and weighted connections using electronic, optical, or neuromorphic architectures. In conventional digital hardware, they are instantiated as configurations of logic gates, memory arrays, and interconnects on CPUs, GPUs, or specialized accelerators (e.g., TPUs), where weights reside in volatile or non-volatile memory and computation is executed via parallel multiply-accumulate operations. Emerging neuromorphic and analog implementations encode synaptic weights in device conductances (e.g., memristors, phase-change materials) and exploit device physics for in-memory computation, enabling high spatiotemporal parallelism, low-latency signal propagation, and energy-efficient approximation of neural operations in a physically embedded network topology.
An innovative planar spoof plasmonic neural network (SPNN) platform capable of directly detecting and processing terahertz (THz) electromagnetic signals has been unveiled by researchers at City University of Hong Kong (CityUHK) ...
Optics & Photonics
Nov 11, 2024
0
46
EPFL researchers have published a programmable framework that overcomes a key computational bottleneck of optics-based artificial intelligence systems. In a series of image classification experiments, they used scattered ...
Optics & Photonics
Aug 6, 2024
0
119
With the development of artificial intelligence technology, various artificial hardware neural networks such as recursive neural networks, convolutional neural networks, and spiking neural networks have been proposed to meet ...
Optics & Photonics
Aug 6, 2024
0
64
Today's supercomputers consume vast amounts of energy, equivalent to the power usage of thousands of homes. In response, researchers are developing a more energy-efficient form of next-generation supercomputing that leverages ...
General Physics
Aug 5, 2024
0
130
The concept of deep diffractive neural networks (D2NN) was introduced by Professor Ozcan's research team in 2018. D2NN combines the principles of light diffraction with the functionalities of neural networks. It consists ...
Optics & Photonics
Aug 1, 2024
0
61
In Light: Science & Applications journal UCLA researchers introduce an innovative design for diffractive deep neural networks (D2NNs). This new architecture, termed Pyramid-D2NN (P-D2NN), achieves unidirectional image magnification ...
Optics & Photonics
Aug 1, 2024
0
97
Scientists propose a new way of implementing a neural network with an optical system which could make machine learning more sustainable in the future. The researchers at the Max Planck Institute for the Science of Light have ...
Optics & Photonics
Jul 12, 2024
0
260
In recent years, artificial intelligence technologies, especially machine learning algorithms, have made great strides. These technologies have enabled unprecedented efficiency in tasks such as image recognition, natural ...
Optics & Photonics
Apr 2, 2024
0
264
Light can compute functions during its propagation and interaction with structured materials, with high speed and low energy consumption. Achieving universal computing using all-optical neural networks requires optical activation ...
Optics & Photonics
Mar 19, 2024
0
89
Reliable quantum gates are the fundamental component of quantum information processing. However, achieving high-dimensional unitary transformations in a scalable and compact manner with ultrahigh fidelities remains a great ...
Optics & Photonics
Feb 7, 2024
0
172
