Metamaterial-Based Analog Recurrent Neural Network Toward Machine Intelligence

Jiang, Tianxi; Li, Tianqi; Huang, Hongbin; Peng, Zhike; He, Qingbo

doi:10.1103/physrevapplied.19.064065

Cited by 8 publications

(3 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recurrent neural networks (RNNs) [275,276] are a class of supervised neural networks whose task is to deal with data sequences and to recognize patterns in these sequences or time series of data. They are often used in language processing and speech recognition.…”

Section: Recurrent Neural Network (Rnn)mentioning

confidence: 99%

“…The same team leader with most of the authors from the work cited in the previous paragraph proposed in 2023 a novel all-optical neural meta-transformer [365] whose metaatoms make use of structural birefringence and polarization rotation to ensure complete control over the tailoring of full Fourier components, in this manner ensuring arbitrary control of all learnable parameters in diffractive optical neural computing. Also in 2023, Jian et al proposed the concept of a metasurface-based hardware implementation of an analog recurrent neural network for mechanical (acoustic) vibrations [276]. This represents expanding the area of DNN metasurfaces beyond wave electromagnetics.…”

Section: Metasurfaces With Hardware-integrated Optical Neural Networkmentioning

confidence: 99%

See 1 more Smart Citation

Synergy between AI and Optical Metasurfaces: A Critical Overview of Recent Advances

Jakšić

2024

Photonics

View full text Add to dashboard Cite

The interplay between two paradigms, artificial intelligence (AI) and optical metasurfaces, nowadays appears obvious and unavoidable. AI is permeating literally all facets of human activity, from science and arts to everyday life. On the other hand, optical metasurfaces offer diverse and sophisticated multifunctionalities, many of which appeared impossible only a short time ago. The use of AI for optimization is a general approach that has become ubiquitous. However, here we are witnessing a two-way process—AI is improving metasurfaces but some metasurfaces are also improving AI. AI helps design, analyze and utilize metasurfaces, while metasurfaces ensure the creation of all-optical AI chips. This ensures positive feedback where each of the two enhances the other one: this may well be a revolution in the making. A vast number of publications already cover either the first or the second direction; only a modest number includes both. This is an attempt to make a reader-friendly critical overview of this emerging synergy. It first succinctly reviews the research trends, stressing the most recent findings. Then, it considers possible future developments and challenges. The author hopes that this broad interdisciplinary overview will be useful both to dedicated experts and a general scholarly audience.

show abstract

Section: Recurrent Neural Network (Rnn)mentioning

confidence: 99%

Section: Metasurfaces With Hardware-integrated Optical Neural Networkmentioning

confidence: 99%

Synergy between AI and Optical Metasurfaces: A Critical Overview of Recent Advances

Jakšić

2024

Photonics

View full text Add to dashboard Cite

show abstract

“…14,15 Similarly, optical/mechanical resonators have been engineered to emulate analog recurrent neural networks (RNNs), with their wave dynamics forming a mathematical isomorphism to digital RNNs. 16,17 These physical learning platforms represent a new class of computing materials known as Physical Neural Networks (PNNs). [18][19][20][21][22] However, it is important to emphasize that replicating the architectures of digital neural networks is not the sole approach to designing a PNN, as illustrated in Fig.…”

Section: Introductionmentioning

confidence: 99%

Equilibrium learning: inverse design of intelligent mechanical machines for prescribed behavior control

Chen,

Miao,

et al. 2024

Active and Passive Smart Structures and Integrated Systems XVIII

View full text Add to dashboard Cite

The exploration of intelligent machines has recently spurred the development of physical neural networks, a class of intelligent metamaterials capable of learning, whether in silico or in situ, from observed data. In this letter, we introduce 'equilibrium learning', a novel physical learning rule designed for lattice-based mechanical neural networks (MNNs) to achieve target performance. This approach leverages the steady states of nodes for back-propagation, efficiently updating the learning degrees of freedom. One-dimensional MNNs, trained with equilibrium learning in silico, can exhibit the desired behaviors on demand function as intelligent mechanical machines. The approach is then employed for the precise morphing control of two-dimensional MNNs subjected to shear or uniaxial loads. Moreover, the MNN is trained to execute classical machine learning tasks such as regression, and preprogrammed bandgap control, establishing it as a versatile platform for physical learning. Our approach presents an efficient pathway for the design of lattice-based mechanical metamaterials for a wide range of static and dynamic target functionalities, positioning them as powerful engines for physical learning.

show abstract

Mechanical Neural Networks with Explicit and Robust Neurons

Mei,

Zhou,

Chen

2024

Advanced Science

View full text Add to dashboard Cite

Mechanical computing provides an information processing method to realize sensing‐analyzing‐actuation integrated mechanical intelligence and, when combined with neural networks, can be more efficient for data‐rich cognitive tasks. The requirement of solving implicit and usually nonlinear equilibrium equations of motion in training mechanical neural networks makes computation challenging and costly. Here, an explicit mechanical neuron is developed of which the response can be directly determined without the need of solving equilibrium equations. A training method is proposed to ensure the robustness of the neuron, i.e., insensitivity to defects and perturbations. The explicitness and robustness of the neurons facilitate the assembly of various network structures. Two exemplified networks, a robust mechanical convolutional neural network and a mechanical recurrent neural network with long short‐term memory capabilities for associative learning, are experimentally demonstrated. The introduction of the explicit and robust mechanical neuron streamlines the design of mechanical neural networks fulfilling robotic matter with a level of intelligence.

show abstract

Metamaterial-Based Analog Recurrent Neural Network Toward Machine Intelligence

Cited by 8 publications

References 35 publications

Synergy between AI and Optical Metasurfaces: A Critical Overview of Recent Advances

Synergy between AI and Optical Metasurfaces: A Critical Overview of Recent Advances

Equilibrium learning: inverse design of intelligent mechanical machines for prescribed behavior control

Mechanical Neural Networks with Explicit and Robust Neurons

Contact Info

Product

Resources

About