Real-Time Computing Without Stable States: A New Framework for Neural Computation Based on Perturbations
Open accessThe DOI matches its Crossref record; 28 of the 28 references it cites were found in a registry or an index.
Abstract
A key challenge for neural modeling is to explain how a continuous stream of multimodal input from a rapidly changing environment can be processed by stereotypical recurrent circuits of integrate-and-fire neurons in real time. We propose a new computational model for real-time computing on time-varying input that provides an alternative to paradigms based on Turing machines or attractor neural networks. It does not require a task-dependent construction of neural circuits. Instead, it is based on principles of high-dimensional dynamical systems in combination with statistical learning theory and can be implemented on generic evolved or found recurrent circuitry. It is shown that the inherent transient dynamics of the high-dimensional dynamical system formed by a sufficiently large and heterogeneous neural circuit may serve as universal analog fading memory. Readout neurons can learn to extract in real time from the current state of such recurrent neural circuit information about current and past inputs that may be needed for diverse tasks. Stable internal states are not required for giving a stable output, since transient internal states can be transformed by readout neurons into stable target outputs due to the high dimensionality of the dynamical system. Our approach is based on a rigorous computational model, the liquid state machine, that, unlike Turing machines, does not require sequential transitions between well-defined discrete internal states. It is supported, as the Turing machine is, by rigorous mathematical results that predict universal computational power under idealized conditions, but for the biologically more realistic scenario of real-time processing of time-varying inputs. Our approach provides new perspectives for the interpretation of neural coding, the design of experiments and data analysis in neurophysiology, and the solution of problems in robotics and neurotechnology.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Although work in artificial intelligence and machine-learning research has shown that similar or even more complex outcomes can be achieved with a variety of techniques, including deep neural networks with trained weights (Gibney, 2015), reservoir networks with a trained output mapping (Maass et al., 2002;Jaeger & Haas, 2004), networks that are evolved using a genetic algorithm (Iizuka & Di Paolo,”
From A potential mechanism for Gibsonian resonance: behavioral entrainment emerges from local homeostasis in an unsupervised reservoir network (2024) · cites this paper as Maass et al., 2002
In the citing paper
“Finally, we show that a randomly connected circuit model, akin to a “reservoir” network13-15, can reproduce the observed distribution of timescales, provided that the network operates at the critical point (or “edge of chaos”)16-18.”
From A reservoir of time constants for memory traces in cortical neurons (2011) · cites this paper as [13]
In the citing paper
“Reservoir-type networks have been the subject of active research in computational neuroscience and machine learning13-15, but so far experimental support that such networks are adopted by the brain has been lacking.”
From A reservoir of time constants for memory traces in cortical neurons (2011) · cites this paper as [13]
In the citing paper
“Those models predict that the memory of input signals is stored in a large, recurrent and heterogeneous network (reservoir), in a distributed manner, and that a desired output is obtained by a trainable combination of the response signals in the reservoir.”
From A reservoir of time constants for memory traces in cortical neurons (2011) · cites this paper as [13]
Citations
Cites
Wu, Y. (1999) · Technometrics · doi:10.1080/00401706.1999.10485951 · Cited by 27,221
Introduction to the theory of neural computation
(1994) · Neural Networks · doi:10.1016/0893-6080(94)90014-0 · Cited by 6,401
The Synaptic Organization of the Brain
(2004) · Oxford University Press eBooks · doi:10.1093/acprof:oso/9780195159561.001.1 · Cited by 3,546 · cited 2×
“This informationisprocessed by extremely complexbut surprisingly stereotypic microcircuits that can perform a wide spectrum of tasks (Shep-herd, 1988 ; Douglas & Martin, 1998 ; von Melchner, Pallas, & Sur, 2000).”
Dynamics of ongoing activity: explanation of the large variability in evoked cortical responses
Arieli, A., Sterkin, A., Grinvald, A., & Aertsen, A. (1996) · Science · doi:10.1126/science.273.5283.1868 · Cited by 1,660 · cited 1×
“Furthermore, although the input rate may contain salient information that is constant for a particular readout element, it may not be for another (see, for example, Figure 8), indicating that equivalence classes and dynamic stability exist purely from the perspective of the readout elements.”
BackgroundDifferential signaling via the same axon of neocortical pyramidal neurons
Markram, H., Wang, Y., & Tsodyks, M. (1998) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.95.9.5323 · Cited by 1,102 · cited 2×
“In the case of a synaptic connection from a to b , we modeled the synaptic dynamics according to the model proposed in Markram, Wang, and Tsodyks (1998), with the synaptic parameters U (use), D (time constant for depression), and F (time constant for facilitation) randomly chosen from gaussian…”
BackgroundMethodSelf-citationOrganizing principles for a diversity of GABAergic interneurons and synapses in the neocortex
Gupta, A., Wang, Y., & Markram, H. (2000) · Science · doi:10.1126/science.287.5451.273 · Cited by 1,036
Self-citationFading memory and the problem of approximating nonlinear operators with Volterra series
Boyd, S., & Chua, L. (1985) · IEEE Transactions on Circuits and Systems · doi:10.1109/tcs.1985.1085649 · Cited by 991 · cited 2×
“> The proof of this theorem follows from the Stone-Weierstrass approximation theorem, like the proof of theorem 1 in Boyd and Chua (1985).”
BackgroundGradient calculations for dynamic recurrent neural networks: a survey
Pearlmutter, B. A. (1995) · IEEE Transactions on Neural Networks · doi:10.1109/72.410363 · Cited by 582 · cited 1×
“Methods for controllingthe dynamics of recurrent neural networks through adaptive mechanisms are reviewed in Pearlmutter (1995).”
Temporal Information Transformed into a Spatial Code by a Neural Network with Realistic Properties
Buonomano, D. V., & Merzenich, M. M. (1995) · Science · doi:10.1126/science.7863330 · Cited by 449 · cited 1×
“Special cases of this mechanism have already been reported in Buonomano and Merzenich (1995) and Dominey, Arbib, and Joseph (1995).”
BackgroundLearning and Extracting Finite State Automata with Second-Order Recurrent Neural Networks
Giles, C. L., Miller, C. B., Chen, D., Chen, H. H., Sun, G. Z., & Lee, Y. C. (1992) · Neural Computation · doi:10.1162/neco.1992.4.3.393 · Cited by 392
Cited by
Items 401–500 of 1,014
Reservoir riddles: suggestions for echo state network research (extended abstract)
Jaeger, H. (2006) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2005.1556090 · Cited by 96
Computational Models of Timing Mechanisms in the Cerebellar Granular Layer
Yamazaki, T., & Tanaka, S. (2009) · The Cerebellum · doi:10.1007/s12311-009-0115-7 · Cited by 96
“The random projection model is a member of liquid-state machines [54] or echo-state networks [55], which has been recently proposed to provide a new paradigm of modern neural computation (see [56, 57] for recent work).”
BackgroundClassification with a disordered dopant-atom network in silicon
Chen, T., van Gelder, J., van de Ven, B., Amitonov, S. V., de Wilde, B., Ruiz Euler, H.-C., Broersma, H., Bobbert, P. A., Zwanenburg, F. A., & van der Wiel, W. G. (2020) · Nature · doi:10.1038/s41586-019-1901-0 · Cited by 95
Artificial Neuron Based on Integrated Semiconductor Quantum Dot Mode-Locked Lasers
Mesaritakis, C., Kapsalis, A., Bogris, A., & Syvridis, D. (2016) · Scientific Reports · doi:10.1038/srep39317 · Cited by 95
Reservoir computing system with double optoelectronic feedback loops
Chen, Y., Yi, L., Ke, J., Zhao, Y., Yang, Y., Huang, L., Zhuge, Q., & Hu, W. (2019) · Optics Express · doi:10.1364/oe.27.027431 · Cited by 101
Role of delay-times in delay-based photonic reservoir computing [Invited]
Hülser, T., Köster, F., Jaurigue, L., & Lüdge, K. (2022) · Optical Materials Express · doi:10.1364/ome.451016 · Cited by 110
Nokkala, J., Martínez-Peña, R., Giorgi, G. L., Parigi, V., Soriano, M. C., & Zambrini, R. (2020) · arXiv · doi:10.48550/arxiv.2006.04821 · Cited by 94
Training spiking neural networks to associate spatio-temporal input–output spike patterns
Mohemmed, A. W., Schliebs, S., Matsuda, S., & Kasabov, N. K. (2012) · Neurocomputing · doi:10.1016/j.neucom.2012.08.034 · Cited by 100
Polarization Multiplexing Reservoir Computing Based on a VCSEL With Polarized Optical Feedback
Guo, X., Xiang, S., Zhang, Y. H., Lin, L., Wen, A. J., & Hao, Y. (2019) · IEEE Journal of Selected Topics in Quantum Electronics · doi:10.1109/jstqe.2019.2932023 · Cited by 102
Dynamics of random recurrent networks with correlated low-rank structure
Schuessler, F., Dubreuil, A. M., Mastrogiuseppe, F., Ostojic, S., & Barak, O. (2020) · Physical Review Research · doi:10.1103/physrevresearch.2.013111 · Cited by 119
“[19] Wolfgang Maass, Thomas Natschläger, and Henry Markram, “Real-time computing without stable states: A new framework for neural computation based on perturbations,” Neural Computation 14, 2531–2560 (2002).”
BackgroundImplicit, Predictive Timing Draws upon the Same Scalar Representation of Time as Explicit Timing
Piras, F., & Coull, J. T. (2011) · PLoS ONE · doi:10.1371/journal.pone.0018203 · Cited by 96
“Indeed, recent theoretical and experimental work suggests that within the milliseconds range, timing does not rely on clock-like mechanisms or a linear metric of time, but rather is inherently encoded in the state of cortical sensory networks (state dependent networks models) [81]–[83].”
BackgroundRepresenting Objects, Relations, and Sequences
Gallant, S. I., & Okaywe, T. W. (2013) · Neural Computation · doi:10.1162/neco_a_00467 · Cited by 110
“Finally, there is interesting work by Maass, Natschläger, and Markram (2002) on asynchronous modeling of noisy neuron behavior, including recurrent connections, in their Liquid State Machine model.”
BackgroundPredicting non-linear dynamics by stable local learning in a recurrent spiking neural network
Gilra, A., & Gerstner, W. (2017) · eLife · doi:10.7554/elife.28295 · Cited by 100
“Our approach toward learning of recurrent spiking networks is situated at the crossroads of reservoir computing (Jaeger, 2001; Maass et al., 2002; Legenstein et al., 2003; Maass and Markram, 2004; Jaeger and Haas, 2004; Joshi and Maass, 2005; Legenstein and Maass, 2007), FORCE learning (Sussillo and Abbott, 2009, 2012…”
BackgroundResultInfluentialAraujo, F. A., Riou, M., Torrejón, J., Tsunegi, S., Querlioz, D., Yakushiji, K., Fukushima, A., Kubota, H., Yuasa, S., Stiles, M. D., et al. (2020) · Scientific Reports · doi:10.1038/s41598-019-56991-x · Cited by 93
Event detection and localization for small mobile robots using reservoir computing
Antonelo, E. A., Schrauwen, B., & Stroobandt, D. (2008) · Neural Networks · doi:10.1016/j.neunet.2008.06.010 · Cited by 105
“This is a term that groups three similar computing techniques, namely, Echo State Networks [5], Liquid State Machines [6], and BackPropagation DeCorrelation [7].”
Complexity in electro-optic delay dynamics: modelling, design and applications
Larger, L. (2013) · Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences · doi:10.1098/rsta.2012.0464 · Cited by 101
“This novel computational principle was originally proposed independently in the early 2000s by the computer science/neural network computing community [34] and the brain cognitive research community [35].”
BackgroundA neurodynamical model for working memory
Pascanu, R., & Jaeger, H. (2010) · Neural Networks · doi:10.1016/j.neunet.2010.10.003 · Cited by 99
“The RC principle has been independently discovered in cognitive neuroscience (temporal recurrent networks, Dominey, Arbib, and Joseph (1995)), in computational neuroscience (liquid state machines, Maass, Natschläger, and Markram (2002)), and in machine learning (echo state networks, Jaeger (2001)).”
BackgroundShaping dynamics with multiple populations in low-rank recurrent networks
Beiran, M., Dubreuil, A., Valente, A., Mastrogiuseppe, F., & Ostojic, S. (2021) · Neural Computation · doi:10.1162/neco_a_01381 · Cited by 92
Rapid time series prediction with a hardware-based reservoir computer
Canaday, D., Griffith, A., & Gauthier, D. J. (2018) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/1.5048199 · Cited by 98
Microring resonators with external optical feedback for time delay reservoir computing
Donati, G. P., Mirasso, C. R., Mancinelli, M., Pavesi, L., & Argyris, A. (2021) · Optics Express · doi:10.1364/oe.444063 · Cited by 92
Unifying framework for information processing in stochastically driven dynamical systems
Kubota, T., Takahashi, H., & Nakajima, K. (2021) · Physical Review Research · doi:10.1103/physrevresearch.3.043135 · Cited by 109
From circuit motifs to computations: mapping the behavioral repertoire of cortical interneurons
Hangya, B., Pi, H.-J., Kvitsiani, D., Ranade, S. P., & Kepecs, A. (2014) · Current Opinion in Neurobiology · doi:10.1016/j.conb.2014.01.007 · Cited by 92
“Current Opinion in Neurobiology 2014, 26:117–124 heterogeneity can have computational benefits [60–63].”
Gottwald, G. A., & Reich, S. (2021) · Physica D Nonlinear Phenomena · doi:10.1016/j.physd.2021.132911 · Cited by 107
“In the context of dynamical systems the framework of random feature maps was extended to include internal degrees of freedom with their own dynamics in so called echo-state networks [13, 14, 15, 16].”
Natural quantum reservoir computing for temporal information processing
Suzuki, Y., Gao, Q., Pradel, K. C., Yasuoka, K., & Yamamoto, N. (2022) · Scientific Reports · doi:10.1038/s41598-022-05061-w · Cited by 150
“As a special type of RNN, reservoir computing (RC) 11 has been actively studied, which originate from echo state networks (ESNs) 12 and liquid state machines (LSMs) 13 .”
BackgroundA Reproducing Kernel Hilbert Space Framework for Spike Train Signal Processing
Paiva, A. R. C., Park, I. M., & Príncipe, J. C. (2008) · Neural Computation · doi:10.1162/neco.2008.09-07-614 · Cited by 91
“…recently, there has also been a great interest in using spike trains for biologically inspired computation paradigms such as the liquid-state machine (Maass, Natschläger, and Markram, 2002; Maass and Bishop, 1998) or spiking neural networks (Bohte, Kok, and Poutré, 2002; Maass and Bishop, 1998).”
BackgroundMethodA Spiking Neural Network System for Robust Sequence Recognition
Yu, Q., Yan, R., Tang, H., Tan, K. C., & Li, H. (2015) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2015.2416771 · Cited by 90
“We use a similar approach for the perceptron rule as in [47] and [48], where the spatiotemporal spike patterns are transformed into continuous states by a low-pass filter.”
MethodPattern recognition in reciprocal space with a magnon-scattering reservoir
Körber, L., Heins, C., Hula, T., Kim, J.-V., Thlang, S., Schultheiß, H., Faßbender, J., & Schultheiß, K. (2023) · Nature Communications · doi:10.1038/s41467-023-39452-y · Cited by 105
“One promising approach is physical reservoir computing, which exploits the nonlinearity and recurrence of dynamical systems (the reservoir) as a computational resource 1 – 4 .”
Learning Universal Computations with Spikes
Thalmeier, D., Uhlmann, M., Kappen, H. J., & Memmesheimer, R.-M. (2016) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1004895 · Cited by 99
“ition and behavior. In artificial computing, it will allow to propose new, possibly more efficient computing schemes. Randomly connected mesoscopic networks can be a suitable substrate for computations [1, 2, 3, 4], as they can reflect the input in a complicated, nonlinear way and thereby maintain fading memory of past i…”
BackgroundInfluentialLeijnen, S., & Fjodor, V. V. (2020) · International Society for the Study of Information Summit (IS4SI Summit) · doi:10.3390/proceedings2020047009 · Cited by 89
Time scales of memory, learning, and plasticity
Tetzlaff, C., Kolodziejski, C., Markelic, I., & Wörgötter, F. (2012) · Biological Cybernetics · doi:10.1007/s00422-012-0529-z · Cited by 81
Altered tactile processing in children with autism spectrum disorder
Tavassoli, T., Bellesheim, K., Tommerdahl, M., Holden, J. M., Kolevzon, A., & Buxbaum, J. D. (2016) · Autism Research · doi:10.1002/aur.1563 · Cited by 88
A roadmap to integrate astrocytes into Systems Neuroscience
Kastanenka, K. V., Moreno‐Bote, R., De Pittà, M., Perea, G., Eraso‐Pichot, A., Masgrau, R., Poskanzer, K. E., & Galea, E. (2020) · Glia · doi:10.1002/glia.23632 · Cited by 87
“Theoretical models of timing in the brain such as oscillators (Goel & Buonomano, 2014) and liquid state (or liquid computing; Maass et al., 2002) may be useful to explore this idea.”
BackgroundHaluszczynski, A., & Rath, C. W. (2019) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/1.5118725 · Cited by 88
“However, even after trying different parameters and alternatively a setup where also the input states are going into the regression as described by Lukosevicius and Jaeger, 19 the variability can always be observed.”
BackgroundWikner, A., Pathak, J., Hunt, B., Girvan, M., Arcomano, T., Szunyogh, I., Pomerance, A., & Ott, E. (2020) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/5.0005541 · Cited by 98
Toward a formal theory for computing machines made out of whatever physics offers
Jaeger, H., Noheda, B., & van der Wiel, W. G. (2023) · Nature Communications · doi:10.1038/s41467-023-40533-1 · Cited by 130
Morando, S., Jemeï, S., Hissel, D., Gouriveau, R., & Noureddine Zerhouni, A. (2015) · Mathematics and Computers in Simulation · doi:10.1016/j.matcom.2015.06.009 · Cited by 88
Computational analysis of memory capacity in echo state networks
Farkaš, I., Bosák, R., & Gergeľ, P. (2016) · Neural Networks · doi:10.1016/j.neunet.2016.07.012 · Cited by 88
“RC has been used to quantify the STM capacity in several network architectures, such as spiking networks (Legenstein &Maass, 2007; Maass et al., 2002; Wallace, Hamid, & Latham, 2013), continuous-time networks (Büsing, Schrauwen, & Legenstein, 2010; Hermans & Schrauwen, 2010), and discrete-time…”
BackgroundMethodPhysics successfully implements Lagrange multiplier optimization
Vadlamani, S. K., Xiao, T. P., & Yablonovitch, E. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2015192117 · Cited by 108
“xi(t + ∆t) = xi(t)−κ∆t ∂ ∂xi L(x,λ), [19]”
A Perovskite Memristor with Large Dynamic Space for Analog-Encoded Image Recognition
Yang, J., Zhang, F., Xiao, H.-M., Wang, Z., Xie, P., Feng, Z., Wang, J., Mao, J., Zhou, Y., & Han, S. (2022) · ACS Nano · doi:10.1021/acsnano.2c09569 · Cited by 89
Online Training of an Opto-Electronic Reservoir Computer Applied to Real-Time Channel Equalization
Antonik, P., Duport, F., Hermans, M., Smerieri, A., Haelterman, M., & Massar, S. (2016) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2016.2598655 · Cited by 87
“RESERVOIR computing (RC) is a set of methods for designing and training artificial recurrent neural networks [1], [2], which brings a drastic simplification of the system design.”
MethodCompressed Timeline of Recent Experience in Monkey Lateral Prefrontal Cortex
Tiganj, Z., Cromer, J. A., Roy, J. E., Miller, E. K., & Howard, M. W. (2018) · Journal of Cognitive Neuroscience · doi:10.1162/jocn_a_01273 · Cited by 92
Machine learning based on reservoir computing with time-delayed optoelectronic and photonic systems
Chembo, Y. K. (2020) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/1.5120788 · Cited by 93
“This approach finds its roots in the concepts of a liquid-state machine (LSM, see Maass et al. 4 ) and one of echo state networks (ESNs, see Jaeger and Haas 5 ) that were later on uni-fiedundertheconceptofreservoircomputing(orRC,seeVerstraeten et al. 6 ).”
BackgroundReservoir computing as digital twins for nonlinear dynamical systems
Kong, L.-W., Weng, Y., Glaz, B. J., Haile, M., & Lai, Y. (2023) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/5.0138661 · Cited by 102
Zhang, H., & Vargas, D. V. (2023) · IEEE Access · doi:10.1109/access.2023.3299296 · Cited by 106
“Intuitively, the reservoir in LSM is often called liquid, since they follow a metaphor of excited states as ripples on the surface of a pool of water [11].”
BackgroundMethodInfluentialStrong and reliable synaptic communication between pyramidal neurons in adult human cerebral cortex
Hunt, S., Leibner, Y., Mertens, E. J., Barros-Zulaica, N., Kanari, L., Heistek, T. S., Karnani, M., Aardse, R., Wilbers, R., Heyer, D. B., et al. (2022) · Cerebral Cortex · doi:10.1093/cercor/bhac246 · Cited by 89
“We used a previously described Tsodyks–Markram-based model (Tsodyks and Markram 1997; Maass et al. 2002; Barros-Zulaica et al. 2019), to extract utilization of synaptic efficacy (U parameter) from a subset of human and mouse connections (nhuman = 21 connections, nmouse = 21 connections).”
MethodSelf-citationNonlinear System Modeling With Random Matrices: Echo State Networks Revisited
Zhang, B., Miller, D. J., & Wang, Y. (2011) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2011.2178562 · Cited by 85
“[3]) share some features characteristic of models for learning in biological brains and they exhibit superior performance when used as “black-box” time-series models.”
Echo State Networks for data-driven downhole pressure estimation in gas-lift oil wells
Antonelo, E. A., Camponogara, E., & Foss, B. A. (2016) · Neural Networks · doi:10.1016/j.neunet.2016.09.009 · Cited by 91
“Most common RC models are the Echo State Networks (ESNs) [23] when analog neurons are used and Liquid State Machines (LSMs) [28] when spiking neurons are considered as dynamical reservoirs.”
Unsupervised heart-rate estimation in wearables with Liquid states and a probabilistic readout
Das, A. K., Pradhapan, P., Groenendaal, W., Adiraju, P., Rajan, R. T., Catthoor, F., Schaafsma, S., Krichmar, J. L., Dutt, N. D., & Hoof, C. A. V. (2018) · Neural Networks · doi:10.1016/j.neunet.2017.12.015 · Cited by 85
“Following are some characteristics of LSM that are relevant for the heart-rate estimation problem we aim to solve in this work.”
MethodInfluentialGradient based hyperparameter optimization in Echo State Networks
Thiede, L., & Parlitz, U. (2019) · Neural Networks · doi:10.1016/j.neunet.2019.02.001 · Cited by 92
A CMOS-based microelectrode array for interaction with neuronal cultures
Hafizovic, S., Heer, F., Ugniwenko, T., Frey, U., Blau, A. W., Ziegler, C., & Hierlemann, A. R. (2007) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2007.04.006 · Cited by 85
The atoms of neural computation
Marcus, G., Marblestone, A., & Dean, T. (2014) · Science · doi:10.1126/science.1261661 · Cited by 83
““Liquid state machines'' (Maass, Natschläger, & Markram, 2002) are tunable computational elements that are consistent with random internal wiring, and it has been suggested that this framework may provide a canonical, but highly configurable, cortical microcircuit model.”
BackgroundRecurrent Kernel Machines: Computing with Infinite Echo State Networks
Hermans, M., & Schrauwen, B. (2011) · Neural Computation · doi:10.1162/neco_a_00200 · Cited by 101
“Originally, this idea had been separately developed for sigmoid nodes (Jaeger, 2001) and spiking neurons (Maass et al., 2002), where the respective approaches were called Echo State Networks (ESN) and Liquid State Machines (LSM), but there are no limitations to the types of networks one could use…”
BackgroundPhotonic Computing With Single and Coupled Spiking Micropillar Lasers
Pammi, V. A., Alfaro-Bittner, K., Clerc, M. G., & Barbay, S. (2019) · IEEE Journal of Selected Topics in Quantum Electronics · doi:10.1109/jstqe.2019.2929187 · Cited by 84
Trends in recurrence analysis of dynamical systems
Marwan, N., & Kraemer, K. H. (2023) · The European Physical Journal Special Topics · doi:10.1140/epjs/s11734-022-00739-8 · Cited by 94
“Reservoir computing (e.g. liquid state machines, echo state networks) is a specific approach of recurrent neural networks to predict the future states of a dynamical system based on time series without a model [137, 138].”
BackgroundAlexander, K., Vaerenbergh, T. V., Fiers, M., Mechet, P., Dambre, J., & Bienstman, P. (2013) · Optics Express · doi:10.1364/oe.21.026182 · Cited by 84
STDP in recurrent neuronal networks
Gilson, M., Burkitt, A. N., & van Hemmen, L. J. (2010) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2010.00023 · Cited by 93
“For instance, in the liquid-state machine a recurrently connected network behaves as a reservoir that performs many arbitrary operations on the inputs, which allows simple supervised training to discriminate between different classes of input (Maass et al., 2002).”
Fading memory and time series prediction in recurrent networks with different forms of plasticity
Lazăr, A., Pipa, G., & Triesch, J. (2007) · Neural Networks · doi:10.1016/j.neunet.2007.04.020 · Cited by 94
“In recent years, a number of recurrent network architectures for time series processing have been proposed that rely on a fixed, randomly connected reservoir of units (Maass et al., 2002; Jäeger & Haas, 2004).”
An ensemble quadratic echo state network for non‐linear spatio‐temporal forecasting
McDermott, P. L., & Wikle, C. K. (2017) · Stat · doi:10.1002/sta4.160 · Cited by 84
“, analog forecasting – see McDermott and Wikle (2016), for a recent example) often utilize embeddings, in which lagged values of the inputs are used in the forecast, as motivated by the theorem of Takens et al. (1981). This is far from standard practice in the ESN literature.”
BackgroundMethodA review: Photonics devices, architectures, and algorithms for optical neural computing
Xiang, S., Han, Y., Song, Z., Guo, X., Zhang, Y., Ren, Z., Wang, S., Ma, Y., Zou, W., Ma, B., et al. (2021) · Journal of Semiconductors · doi:10.1088/1674-4926/42/2/023105 · Cited by 87
Neural cytoskeleton capabilities for learning and memory
Priel, A., Tuszynski, J. A., & Woolf, N. J. (2010) · Journal of Biological Physics · doi:10.1007/s10867-009-9153-0 · Cited by 78
“The basic structure of an LSM is composed of an excitable medium (i.e., a “liquid”), with the output function mapping the current liquid state, as shown in Fig.”
BackgroundMorphological Computation and Morphological Control: Steps Toward a Formal Theory and Applications
Füchslin, R. M., Dzyakanchuk, A., Flumini, D., Hauser, H., Hunt, K. J., Luchsinger, R. H., Reller, B., Scheidegger, S., & Walker, R. (2013) · Artificial Life · doi:10.1162/artl_a_00079 · Cited by 83
“For example, it is well established that the dynamics of artificial neural networks can be used to perform complex computations [14, 20].”
BackgroundNumerical demonstration of neuromorphic computing with photonic crystal cavities
Laporte, F., Katumba, A., Dambre, J., & Bienstman, P. (2018) · Optics Express · doi:10.1364/oe.26.007955 · Cited by 88
Information Processing Capacity of Spin-Based Quantum Reservoir Computing Systems
Martínez‐Peña, R., Nokkala, J., Giorgi, G. L., Zambrini, R., & Soriano, M. C. (2020) · Cognitive Computation · doi:10.1007/s12559-020-09772-y · Cited by 121
“cess the input information [10,11].”
BackgroundSackesyn, S., Ma, C., Dambre, J., & Bienstman, P. (2021) · Optics Express · doi:10.1364/oe.435013 · Cited by 82
“Reservoir computing was originally introduced as an efficient way to train a recurrent neural network [17,18], but it has caught a lot of attention as an easy-to-implement machine learning method in new hardware platforms.”
MethodA θ–γ Oscillation Code for Neuronal Coordination during Motor Behavior
Igarashi, J., Isomura, Y., Arai, K., Harukuni, R., & Fukai, T. (2013) · Journal of Neuroscience · doi:10.1523/jneurosci.2126-13.2013 · Cited by 96
“, 2012) and the “general-purpose neocortical circuit,” such as transient memory storage and its readout (Maass et al., 2002; Jaeger and Haas, 2004; Sussillo and Abbott, 2009).”
BackgroundNeuromorphic learning, working memory, and metaplasticity in nanowire networks
Loeffler, A., Álvarez, A. D., Zhu, R., Ganesh, N., Shine, J. M., Nakayama, T., & Kuncic, Z. (2023) · Science Advances · doi:10.1126/sciadv.adg3289 · Cited by 91
Bouazizi, S., & Ltifi, H. (2023) · Decision Support Systems · doi:10.1016/j.dss.2023.114126 · Cited by 92
Nonlinear dynamics of the brain: emotion and cognition
Rabinovich, M. I., & Muezzinoglu, M. K. (2010) · Physics-Uspekhi · doi:10.3367/ufne.0180.201004b.0371 · Cited by 108
A New Supervised Learning Algorithm for Spiking Neurons
Xu, Y., Zeng, X., & Zhong, S. (2013) · Neural Computation · doi:10.1162/neco_a_00450 · Cited by 85
“Because every input of the classification problem is a single spike train, we employ the concept of reservoir computing with a reservoir network (Maass et al., 2002; Jaeger, Maass, & Principe, 2007; Ponulak & Kasiński, 2010) to perform the experiments.”
MethodOverview of facts and issues about neural coding by spikes
Cessac, B., Paugam‐Moisy, H., & Viéville, T. (2009) · Journal of Physiology-Paris · doi:10.1016/j.jphysparis.2009.11.002 · Cited by 87
“…networks are currently implemented through non specific network architectures, such as Echo State Networks (Jaeger, 2003) and Liquid Sate Machines (Maass et al., 2002) or other models (Paugam-Moisy et al., 2008) that are called ‘‘reservoir computing” (see Verstraeten et al. (2007) for a merge and…”
Echo State Networks used for Motor Control
Salmen, M., & Plöger, P. G. (2006) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/robot.2005.1570399 · Cited by 81
“Neural networks can serve as an universal dynamical system representation [8] and they have a great potential for black-box modeling especially for non-linear systems.”
BackgroundA Model for Real-Time Computation in Generic Neural Microcircuits
Maass, W., Natschläger, T., & Markram, H. (2002) · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · Cited by 99
Self-citationA Spike-Timing-Based Integrated Model for Pattern Recognition
Hu, J., Tang, H., Tan, K. C., Li, H., & Shi, L. (2012) · Neural Computation · doi:10.1162/neco_a_00395 · Cited by 82
“One approach is to employ multiple-layer and recurrent neural structures, such as a liquid state machine (Maass, Natschl¨ager, & Markram, 2002), so as to increase the computational capability of the system and absorb the influence of multiple spikes.”
MethodOptimizing a quantum reservoir computer for time series prediction
Kutvonen, A., Fujii, K., & Sagawa, T. (2020) · Scientific Reports · doi:10.1038/s41598-020-71673-9 · Cited by 106
Hybrid quantum-classical reservoir computing of thermal convection flow
Pfeffer, P., Heyder, F., & Schumacher, J. (2022) · Physical Review Research · doi:10.1103/physrevresearch.4.033176 · Cited by 99
Parallel and deep reservoir computing using semiconductor lasers with optical feedback
Hasegawa, H., Kanno, K., & Uchida, A. (2022) · Nanophotonics · doi:10.1515/nanoph-2022-0440 · Cited by 89
“There has been a considerable rise in the demand and subsequent interest in reservoir computing in the past two decades [7, 8].”
BackgroundBiological neurons act as generalization filters in reservoir computing
Sumi, T., Yamamoto, H., Katori, Y., Ito, K., Moriya, S., Konno, T., Sato, S., & Hirano‐Iwata, A. (2023) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2217008120 · Cited by 93
“The strength of reservoir computing is its ability to process time-series data using the transient dynamics of the reservoir (19).”
Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks
Goudar, V., & Buonomano, D. V. (2018) · eLife · doi:10.7554/elife.31134 · Cited by 81
Metastable attractors explain the variable timing of stable behavioral action sequences
Recanatesi, S., Pereira, U., Murakami, M., Mainen, Z. F., & Mazzucato, L. (2021) · Neuron · doi:10.1016/j.neuron.2021.10.011 · Cited by 87
Deep Randomized Neural Networks
Gallicchio, C., & Scardapane, S. (2020) · Studies in computational intelligence · doi:10.1007/978-3-030-43883-8_3 · Cited by 103
Robust Optimization and Validation of Echo State Networks for learning chaotic dynamics
Racca, A., & Magri, L. (2021) · Neural Networks · doi:10.1016/j.neunet.2021.05.004 · Cited by 96
Thin-film transistor for temporal self-adaptive reservoir computing with closed-loop architecture
Chen, R., Yang, H., Li, R., Yu, G., Zhang, Y., Dong, J., Han, D., Zhou, Z., Huang, P. C., Liu, L., et al. (2024) · Science Advances · doi:10.1126/sciadv.adl1299 · Cited by 81
“As a variant of RNN, reservoir computing (RC) is proposed by replacing the complex recursive network with relatively fixed nonlinear reservoirs ( 9 , 10 ).”
BackgroundFace-selective neurons maintain consistent visual responses across months
McMahon, D. B. T., Jones, A. P., Bondar, I. V., & Leopold, D. A. (2014) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1318331111 · Cited by 84
“And indeed, both theoretical (4) and experimental results from motor cortex (5, 6) demonstrate that a network composed of unstable units may nonetheless exhibit stable performance.”
BackgroundA comparative study of Reservoir Computing strategies for monthly time series prediction
wyffels, F., & Schrauwen, B. (2010) · Neurocomputing · doi:10.1016/j.neucom.2010.01.016 · Cited by 92
“When spiking neurons are used, one often speaks of Liquid State Machines [15].”
MethodDoborjeh, M., Wang, G., Kasabov, N. K., Kydd, R. R., & Russell, B. (2015) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2015.2503400 · Cited by 83
“Many methods for SNN have been already created: Encoding continuous data, such as image and speech data, into trains of spikes [12]–[14]; spatiotemporal data learning [11], [15], [16]; SNN reservoir computing and liquid state machines (LSMs) [17], [18]; classification systems [19], [20].”
BackgroundMethodMachine-learning inference of fluid variables from data using reservoir computing
Nakai, K., & Saiki, Y. (2018) · Physical review. E · doi:10.1103/physreve.98.023111 · Cited by 89
Reproducible Sequence Generation In Random Neural Ensembles
Huerta, R., & Rabinovich, M. (2004) · Physical Review Letters · doi:10.1103/physrevlett.93.238104 · Cited by 77
Overcoming Catastrophic Interference using Conceptor-Aided Backpropagation
He, X., & Jaeger, H. (2018) · International Conference on Learning Representations · Cited by 115
Minimal approach to neuro-inspired information processing
Soriano, M. C., Brunner, D., Escalona-Morán, M., Mirasso, C. R., & Fischer, I. (2015) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2015.00068 · Cited by 84
“The neuro-scientific reality of the reservoir computing has already been envisioned in the work of Maass et al. (2002).”
BackgroundSynthesis of neural networks for spatio-temporal spike pattern recognition and processing
Tapson, J. C., Cohen, G., Afshar, S., Stiefel, K. M., Buskila, Y., Wang, R. M., Hamilton, T. J., & André, V. S. (2013) · Frontiers in Neuroscience · doi:10.3389/fnins.2013.00153 · Cited by 78
Movement Generation with Circuits of Spiking Neurons
Joshi, P., & Maass, W. (2005) · Neural Computation · doi:10.1162/0899766054026684 · Cited by 90
“Whereas traditional models for neural computation focused on constructions of neural implementations of Turing machines or other off-line computational models, more recent results have demonstrated that biologically more realistic neural microcircuit models consisting of spiking neurons and dynamic synapses are well s…”
BackgroundMethodInfluentialSelf-citationModel-free tracking control of complex dynamical trajectories with machine learning
Zhai, Z.-M., Moradi, M., Kong, L.-W., Glaz, B. J., Haile, M., & Lai, Y. (2023) · Nature Communications · doi:10.1038/s41467-023-41379-3 · Cited by 84
Stable Output Feedback in Reservoir Computing Using Ridge Regression
wyffels, F., Schrauwen, B., & Stroobandt, D. (2008) · Lecture notes in computer science · doi:10.1007/978-3-540-87536-9_83 · Cited by 100
“These techniques were originally introduced as Liquid State Machines [1] or Echo State Networks [2], but are now commonly referred to as Reservoir Computing [3].”
MethodAn approach to reservoir computing design and training
Ferreira, A. A., Ludermir, T. B., & de Aquino, R. R. B. (2013) · Expert Systems with Applications · doi:10.1016/j.eswa.2013.01.029 · Cited by 85
“In 2001, a new approach to RNN design and training was proposed independently under the name of Liquid State Machines [19] and under the name of Echo State Networks [13].”
MethodOn Learning Navigation Behaviors for Small Mobile Robots With Reservoir Computing Architectures
Antonelo, E. A., & Schrauwen, B. (2014) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2014.2323247 · Cited by 84
Dynamic Flux Tubes Form Reservoirs of Stability in Neuronal Circuits
Monteforte, M., & Wolf, F. (2012) · Physical Review X · doi:10.1103/physrevx.2.041007 · Cited by 71
“[22] Wolfgang Maass, Thomas Natschläger, and Henry Markram, Real-Time Computing without Stable States: A New Framework for Neural Computation Based on Perturbations, Neural Comput.”
Rats adopt the optimal timescale for evidence integration in a dynamic environment
Piet, A. T., El Hady, A., & Brody, C. D. (2018) · Nature Communications · doi:10.1038/s41467-018-06561-y · Cited by 76
Patel, D., Canaday, D., Girvan, M., Pomerance, A., & Ott, E. (2021) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/5.0042598 · Cited by 86
Stock market index prediction based on reservoir computing models
Wang, W., Tang, Y., Xiong, J. J., & Zhang, Y. (2021) · Expert Systems with Applications · doi:10.1016/j.eswa.2021.115022 · Cited by 83
Reservoir computing with diverse timescales for prediction of multiscale dynamics
Tanaka, G., MATSUMORI, T., Yoshida, H., & Aihara, K. (2022) · Physical Review Research · doi:10.1103/physrevresearch.4.l032014 · Cited by 82
Metadata
- Authors
- Maass, W., Natschläger, T., & Markram, H.
- Affiliations
- Graz University of Technology (AT)
- École Polytechnique Fédérale de Lausanne (CH)
- Corresponding author
- Maass, Wolfgang
- Natschläger, Thomas
- Markram, Henry
- Venue
- Neural Computation
- Publisher
- MIT Press
- Published
- 2002-11-01
- Volume
- 14
- Issue
- 11
- Pages
- 2531-2560
- ISSN
- 0899-7667 (ISSN-L)
- 1530-888X
- OpenAlex
- W2103179919
- Cited by
- 4,025
- References
- 28
- Citations by year
- 2026: 286 · 2025: 351 · 2024: 307 · 2023: 309 · 2022: 276 · 2021: 303 · 2020: 239 · 2019: 244 · 2018: 217 · 2017: 173 · 2016: 155 · 2015: 135 · 2014: 142 · 2013: 139 · 2012: 125
- FWCI
- 18.37
- Type
- journal article
- Language
- en
- Open access
- green
- infoscience.epfl.ch
- Locations
- Neural Computation (publishedVersion)
- PubMed (publishedVersion)
- Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion)
- Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion, cc-by-nc-nd)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Reservoir Computing
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
- Keywords
- liquid state machines, recurrent neural circuits, transient dynamics, high-dimensional dynamical systems, fading memory, integrate-and-fire neurons, real-time processing, time-varying inputs, neural coding, statistical learning theory, computational power, attractor networks
- MeSH terms
- Action Potentials, Computer Simulation, Computer Systems, Computers, Models, Neurological, Neurons, Neural Networks, Computer
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.