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
t Synchrony Generation in Recurrent Networks with Frequency-Dependent Synapses
Tsodyks, T., Uziel, A., & Markram, H. (2000) · The Journal of Neuroscience · doi:10.1523/jneurosci.20-01-j0003.2000 · Cited by 377
Self-citationVisual behaviour mediated by retinal projections directed to the auditory pathway
von Melchner, L., Pallas, S. L., & Sur, M. (2000) · Nature · doi:10.1038/35009102 · Cited by 370
Analog computation via neural networks
Siegelmann, H. T., & Sontag, E. D. (1994) · Theoretical Computer Science · doi:10.1016/0304-3975(94)90178-3 · Cited by 355
On the Computational Power of Winner-Take-All
Maass, W. (2000) · Neural Computation · doi:10.1162/089976600300014827 · Cited by 355 · cited 2×
“Theoretically, the class of readout maps that can be implemented in this fashion satises the approximation property AP (Maass, 2000; Auer, Burgsteiner, & Maass, 2001), and is according to theorem 1 in principle sufcient for approximating arbitrary given fading memory lters F.”
BackgroundSelf-citationThe induction of dynamical recognizers
Pollack, J. B. (1991) · Machine Learning · doi:10.1007/bf00114845 · Cited by 294 · cited 1×
“…computation in biological neural systems are based on constructions of arti(cid:142)cial neural networks that simulate Turing machines or other models for digital computation (Pollack, 1991 ; Giles, Miller, Chen, Sun, & Lee, 1992 ; Siegelmann & Sontag, 1994 ; Hyoetyniemi, 1996 ; Moore, 1998).”
MethodModels of Computation: Exploring the Power of Computing
Savage, J. E. (1998) · Cited by 293
What is a moment? Transient synchrony as a collective mechanism for spatiotemporal integration
Hopfield, J. J., & Brody, C. D. (2001) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.98.3.1282 · Cited by 248
Lower Bounds for the Computational Power of Networks of Spiking Neurons
Maass, W. (1996) · Neural Computation · doi:10.1162/neco.1996.8.1.1 · Cited by 226 · cited 2×
“Methods for controlling the dynamics of recurrent neural networks through adaptive mechanisms are reviewed in Pearlmutter (1995). So far, no one has been able to apply these to the case of networks of spiking neurons.”
BackgroundSelf-citationA Model of Corticostriatal Plasticity for Learning Oculomotor Associations and Sequences
Dominey, P., Arbib, M., & Joseph, J.-P. (1995) · Journal of Cognitive Neuroscience · doi:10.1162/jocn.1995.7.3.311 · Cited by 190 · cited 1×
“Special cases of this mechanism have already been reported in Buonomano and Merzenich (1995) and Dominey, Arbib, and Joseph (1995).”
BackgroundWhat is a moment? Transient synchrony as a collective mechanism for spatiotemporal integration
Hopfield, J. J. (2001) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.031567098 · Cited by 76
Maass, W., & Sontag, E. D. (1999) · Neural Computation · doi:10.1162/089976699300016656 · Cited by 67 · cited 1×
“Furthermore, the results of Maass and Sontag (1999) on the impact of noise on the computational power of recurrent neural networks suggest that all of these approaches break down as soon as one assumes that the underlying analog computational units are subject to gaussian or other realistic noise…”
BackgroundSelf-citationCoding and learning of behavioral sequences
Melamed, O., Wulfram Gerstner, B., Maass, W., Tsodyks, M. V., & Markram, H. (2003) · Trends in Neurosciences · doi:10.1016/j.tins.2003.10.014 · Cited by 67
Self-citationNeural Systems as Nonlinear Filters
Maass, W., & Sontag, E. D. (2000) · Neural Computation · doi:10.1162/089976600300015123 · Cited by 66 · cited 1×
“…are the class of all delay (cid:142)lters u ( ) 7! u t 0 ( ) (for t 0 2 R ), the class of all linear (cid:142)lters with impulse responses of the form h ( t ) D e ¡ at with a > 0, and the class of (cid:142)lters de(cid:142)ned by standard models for dynamic synapses (see Maass & Sontag, 2000).”
BackgroundSelf-citationDynamical recognizers: real-time language recognition by analog computers
Moore, C. (1998) · Theoretical Computer Science · doi:10.1016/s0304-3975(97)00028-5 · Cited by 65 · cited 1×
“…computation in biological neural systems are based on constructions of arti(cid:142)cial neural networks that simulate Turing machines or other models for digital computation (Pollack, 1991 ; Giles, Miller, Chen, Sun, & Lee, 1992 ; Siegelmann & Sontag, 1994 ; Hyoetyniemi, 1996 ; Moore, 1998).”
MethodLocal Interactions in Neural Networks Explain Global Effects in Gestalt Processing and Masking
Herzog, M. H., Ernst, U. A., Etzold, A., & Eurich, C. W. (2003) · Neural Computation · doi:10.1162/089976603322297304 · Cited by 54
Dynamical Motifs: Building Blocks of Complex Dynamics in Sparsely Connected Random Networks
Zhigulin, V. P. (2004) · Physical Review Letters · doi:10.1103/physrevlett.92.238701 · Cited by 46
Can excitable media be considered as computational systems?
Holden, A. V., Tucker, J. V., & Thompson, B. C. (1991) · Physica D Nonlinear Phenomena · doi:10.1016/0167-2789(91)90212-r · Cited by 39 · cited 1×
“As an illustration for our general approach to real-time computing consider a series of transient perturbations caused in an excitable medium (see Holden, Tucker, & Thompson, 1991), for example, a liquid, by a sequence of external disturbances (inputs) such as wind, sound, or sequences of peb-bles…”
MethodHäusler, S., Markram, H., & Maass, W. (2003) · Complexity · doi:10.1002/cplx.10089 · Cited by 27
Self-citation
Cited by
Deep learning in neural networks: an overview
Schmidhuber, J. (2015) · Neural Networks · doi:10.1016/j.neunet.2014.09.003 · Cited by 18,570
“In fact, some popular RNN algorithms restricted credit assignment to a single step backwards (Elman, 1990; Jordan, 1986, 1997), also inmore recent studies (Jaeger, 2001, 2004; Maass et al., 2002).”
ResultHarnessing Nonlinearity: Predicting Chaotic Systems and Saving Energy in Wireless Communication
Jaeger, H., & Haas, H. (2004) · Science · doi:10.1126/science.1091277 · Cited by 3,811
Reservoir computing approaches to recurrent neural network training
Lukoševičius, M., & Jaeger, H. (2009) · Computer Science Review · doi:10.1016/j.cosrev.2009.03.005 · Cited by 2,956
“Liquid State Machines (LSMs) [11] are the other pioneering reservoir method, developed independently”
BackgroundMethodInfluentialRecent advances in physical reservoir computing: A review
Tanaka, G., Yamane, T., Héroux, J. B., Nakane, R., Kanazawa, N., Takeda, S., Numata, H., Nakano, D., & Hirose, A. (2019) · Neural Networks · doi:10.1016/j.neunet.2019.03.005 · Cited by 2,146
“…RC is a unified computational framework (Lukoševičius & Jaeger, 2009; Verstraeten, Schrauwen, d’ Haene, & Stroobandt, 2007), derived from independently proposed RNN models, such as echo state networks (ESNs) (Jaeger, 2001) and liquid state machines (LSMs) (Maass, Natschläger, & Markram, 2002).”
BackgroundMethodInfluentialInformation processing using a single dynamical node as complex system
Appeltant, L., Soriano, M. C., Van der Sande, G., Danckaert, J., Massar, S., Dambre, J., Schrauwen, B., Mirasso, C. R., & Fischer, I. (2011) · Nature Communications · doi:10.1038/ncomms1476 · Cited by 1,716
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.