Self-organizing neuromorphic nanowire networks are stochastic dynamical systems DOI Creative Commons
Gianluca Milano, Fabio Michieletti, Carlo Ricciardi

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: April 4, 2024

Abstract Neuromorphic computing aims to develop software and hardware platforms emulating the information processing effectiveness of our brain. In this context, self-organizing neuromorphic nanonetworks have been demonstrated as suitable physical substrates for in materia implementation unconventional paradigms, like reservoir computing. However, understanding relationship between emergent dynamics capabilities still represents a challenge. Here, we demonstrate that nanowire-based networks are stochastic dynamical systems where signals flow relies on intertwined action deterministic random factors. We show through an experimental modeling approach these combine stimuli-dependent trajectories effects caused by noise jumps can be holistically described Ornstein-Uhlenbeck process, providing unifying framework surpassing current approaches (not only nanowire-based) limited either or effects. Since dynamically tuned controlling network’s attractor memory state, results open new perspectives rational development paradigms exploiting in single platform similarly

Language: Английский

Physical reservoir computing with emerging electronics DOI
Xiangpeng Liang, Jianshi Tang, Ya‐Nan Zhong

et al.

Nature Electronics, Journal Year: 2024, Volume and Issue: 7(3), P. 193 - 206

Published: March 12, 2024

Language: Английский

Citations

53
Reconfigurable reservoir computing in a magnetic metamaterial DOI Creative Commons
Ian Vidamour, C. Swindells, G. Venkat

et al.

Communications Physics, Journal Year: 2023, Volume and Issue: 6(1)

Published: Aug. 26, 2023

Language: Английский

Citations

27
Physical Reservoir Computing Based on Nanoscale Materials and Devices DOI

Zhiying Qi,

Linjie Mi,

Haoran Qian

et al.

Advanced Functional Materials, Journal Year: 2023, Volume and Issue: 33(43)

Published: Aug. 31, 2023

Abstract Bioinspired computation systems can achieve artificial intelligence, bypassing fundamental bottlenecks and cost constraints. Computational frameworks suited for temporal/sequential data processing such as recurrent neural networks (RNNs) suffer from problems of high complexity low efficiency. Physical assembled with nanoscale materials devices represent an alternative route to serve the core component physically implanted reservoir computing. In this review, overview development paradigm physical computing (PRC) is provided typical reservoirs constructed nanomaterials nanodevices are described. The based on multiple overcome RNN, show strong robustness, effectively deal tasks improved reliability availability. Finally, challenges perspectives nanomaterial nanodevice‐based PRC a next‐generation machine learning discussed.

Language: Английский

Citations

25
Thin-film transistor for temporal self-adaptive reservoir computing with closed-loop architecture DOI Creative Commons
Ruiqi Chen, Haozhang Yang, Ruiyi Li

et al.

Science Advances, Journal Year: 2024, Volume and Issue: 10(7)

Published: Feb. 16, 2024

Reservoir computing is a powerful neural network-based paradigm for spatiotemporal signal processing. Recently, physical reservoirs have been explored based on various electronic devices with outstanding efficiency. However, the inflexible temporal dynamics of these posed fundamental restrictions in processing signals timescales. Here, we fabricated thin-film transistors controllable dynamics, which can be easily tuned electrical operation and showed excellent cycle-to-cycle uniformity. Based this, constructed adaptive reservoir capable extracting information multiple timescales, thereby achieving improved accuracy human-activity-recognition task. Moreover, by leveraging former output to modify hyperparameters, closed-loop architecture that equips system self-adaptability according current input. The adaptability demonstrated accurate real-time recognition objects moving at diverse speed levels. This work provides an approach systems achieve compound characteristics.

Language: Английский

Citations

17
Experimental Demonstration of Reservoir Computing with Self‐Assembled Percolating Networks of Nanoparticles DOI Creative Commons
Joshua B. Mallinson, Jamie K. Steel, Zachary E. Heywood

et al.

Advanced Materials, Journal Year: 2024, Volume and Issue: 36(29)

Published: April 1, 2024

The complex self-assembled network of neurons and synapses that comprises the biological brain enables natural information processing with remarkable efficiency. Percolating networks nanoparticles (PNNs) are nanoscale systems have been shown to possess many promising brain-like attributes which therefore appealing for neuromorphic computation. Here experiments performed show PNNs can be utilized as physical reservoirs within a nanoelectronic reservoir computing framework demonstrate successful computation several benchmark tasks (chaotic time series prediction, nonlinear transformation, memory capacity). For each task, relevant literature results compiled it is performance compares favorably previously reported from reservoirs. It then demonstrated experimentally used spoken digit recognition state-of-the-art accuracy. Finally, parallel architecture emulated, increases dimensionality richness outputs in further improvements across all tasks.

Language: Английский

Citations

12
In materia implementation strategies of physical reservoir computing with memristive nanonetworks DOI Creative Commons
Gianluca Milano,

Kevin Montano,

Carlo Ricciardi

et al.

Journal of Physics D Applied Physics, Journal Year: 2023, Volume and Issue: 56(8), P. 084005 - 084005

Published: Feb. 1, 2023

Abstract Physical reservoir computing (RC) represents a computational framework that exploits information-processing capabilities of programmable matter, allowing the realization energy-efficient neuromorphic hardware with fast learning and low training cost. Despite self-organized memristive networks have been demonstrated as physical able to extract relevant features from spatiotemporal input signals, multiterminal nanonetworks open possibility for novel strategies implementation. In this work, we report on implementation in materia RC self-assembled networks. Besides showing information processing nanowire networks, show through simulations emergent collective dynamics allows unconventional implementations where same electrodes can be used both inputs outputs. By comparing different digit recognition task, reduction complexity without limiting capabilities, thus providing new insights taking full advantage toward rational design systems.

Language: Английский

Citations

18
In-materio reservoir computing based on nanowire networks: fundamental, progress, and perspective DOI Creative Commons
Renrui Fang,

Woyu Zhang,

Kuan Ren

et al.

Materials Futures, Journal Year: 2023, Volume and Issue: 2(2), P. 022701 - 022701

Published: April 17, 2023

Abstract The reservoir computing (RC) system, known for its ability to seamlessly integrate memory and functions, is considered as a promising solution meet the high demands time energy-efficient in current big data landscape, compared with traditional silicon-based systems that have noticeable disadvantage of separate storage computation. This review focuses on in-materio RC based nanowire networks (NWs) from perspective materials, extending devices applications. common methods used preparing nanowires-based reservoirs, including synthesis nanowires construction networks, are firstly systematically summarized. physical principles memristive memcapacitive junctions then explained. Afterwards, dynamic characteristics reservoirs their capability, well neuromorphic applications NWs-based recognition, classification, forecasting tasks, explicated detail. Lastly, challenges future opportunities facing highlighted, aiming provide guidance further research.

Language: Английский

Citations

14
Mean Field Theory of Self‐Organizing Memristive Connectomes DOI Creative Commons
Francesco Caravelli, Gianluca Milano, Carlo Ricciardi

et al.

Annalen der Physik, Journal Year: 2023, Volume and Issue: 535(8)

Published: June 6, 2023

Biological neuronal networks are characterized by nonlinear interactions and complex connectivity. Given the growing impetus to build neuromorphic computers, understanding physical devices that exhibit structures functionalities similar biological neural is an important step toward this goal. Self-organizing circuits of nanodevices at forefront research in computing, as their behavior mimics synaptic plasticity features circuits. However, effective theory describe lacking. This study provides for first time mean field emergent voltage-induced polymorphism \textit{circuits} a nanowire connectome, showing these can be explained low-dimensional dynamical equation. The equation derived from microscopic dynamics single memristive junction analytical form. We test our model on experiments show it fits both potentiation depression synapse-mimicking applies beyond case formulating general mean-field conductance transitions self-organizing connectomes.

Language: Английский

Citations

14
Ergodicity, lack thereof, and the performance of reservoir computing with memristive networks DOI Creative Commons
Valentina Baccetti, Ruomin Zhu, Zdenka Kuncic

et al.

Nano Express, Journal Year: 2024, Volume and Issue: 5(1), P. 015021 - 015021

Published: Feb. 15, 2024

Abstract Networks composed of nanoscale memristive components, such as nanowire and nanoparticle networks, have recently received considerable attention because their potential use neuromorphic devices. In this study, we explore ergodicity in showing that the performance on machine leaning tasks improves when these networks are tuned to operate at edge between two global stability points. We find lack is associated with emergence memory system. measure level using Thirumalai-Mountain metric, show absence ergodicity, different network systems improved utilized reservoir computers (RC). highlight it also important let system synchronize input signal order for RC exhibit improvements over baseline.

Language: Английский

Citations

5
Reservoir computing using networks of memristors: effects of topology and heterogeneity DOI
Joshua B. Mallinson, Zachary E. Heywood, Ryan. K. Daniels

et al.

Nanoscale, Journal Year: 2023, Volume and Issue: 15(22), P. 9663 - 9674

Published: Jan. 1, 2023

Reservoir computing (RC) has attracted significant interest as a framework for the implementation of novel neuromorphic architectures. Previously attention been focussed on software-based reservoirs, where it demonstrated that reservoir topology plays role in task performance, and functional advantage attributed to small-world scale-free connectivity. However hardware systems, such electronic memristor networks, mechanisms responsible dynamics are very different is largely unknown. Here we compare performance range memristive reservoirs several RC tasks chosen highlight system requirements. We focus percolating networks nanoparticles (PNNs) which self-assembled nanoscale systems exhibit properties. find regular arrays uniform elements limited by their symmetry but this can be broken either heterogeneous distribution properties or topology. The best perfomance across all observed network with memistor These results provide insight into well an overview computational memristors benchmark tasks.

Language: Английский

Citations

11