Pith. sign in

REVIEW 4 major objections 5 minor 60 references

From Neurons to Computation: Biological Reservoir Computing for Pattern Recognition

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read The paper claims that cultured biological neurons, interfaced through a multi-electrode array, can act as the untrained reservoir in a reservoir-computing pipeline for pattern recognition.

desk verdict A plausible feasibility demo that needs artifact controls and replication before the central claim holds. read the letter →

arxiv 2505.03510 v2 pith:BI4VJO4Q submitted 2025-05-06 cs.NE cs.AIcs.CV

classification cs.NEcs.AIcs.CV
keywords reservoircomputingbiologicalmulti-electrodearrayculturedneuronspatternrecognitionechostatenetworkspikecountinglinearreadout
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that a network of cultured biological neurons, read through a multi-electrode array, can play the role of the untrained reservoir in reservoir computing. Inputs are delivered as electrical pulses on selected electrodes, and the spike counts observed on the remaining electrodes become a 4096-dimensional feature vector; only a single-layer perceptron is trained. On three tasks of increasing difficulty, point stimuli, oriented bars, and clock-style digits, the biological features support classification accuracies of 98%, 92%, and 95%, respectively. If the claim holds, biological tissue becomes a drop-in nonlinear feature extractor that needs no weight training, with possible energy-efficiency and neuroscience payoffs.

What carries the argument

The central object is the biological reservoir: a cultured network of stem-cell-derived neurons grown on a multi-electrode array and used as a physical echo-state-like system. Inputs are mapped to selected electrode pairs, each stimulus is repeated 25 times with a 10 s rest interval, and per-electrode spikes are detected from the raw recordings; the response of each electrode is the spike count in the 10 ms after the stimulus minus the 10 ms before it. The resulting 4096 spike-count values form a nonlinear, high-dimensional feature representation whose only trained component is a linear readout.

What would settle it

Run the same stimulation protocol on a blank multi-electrode array with no neurons: if the extracted spike counts still separate the classes and a linear classifier trained on them reaches comparable accuracy, the biological computation premise fails. Alternatively, repeat the three tasks on at least two more independently cultured networks; if their accuracies do not reproduce, the claim is culture-specific rather than general.

Watch

Extended reading notes

Core claim

The central claim is that spike-count vectors elicited from a cultured neuronal network by distinct input patterns form a high-dimensional representation in which the patterns are linearly separable enough for a linear classifier to recognize them. The authors call this arrangement biological reservoir computing (BRC). Input patterns are mapped onto selected electrode pairs of a 64 by 64 multi-electrode array, repeated pulses are delivered, and the number of spikes detected on each electrode in a 10 ms window after stimulation forms the feature vector. A single-layer perceptron trained by stochastic gradient descent on 20 responses per class then classifies 5 held-out responses per class, reaching average accuracies of 98% for pointwise stimuli, 92% for oriented bars, and 95% for digit patterns.

Load-bearing premise

The argument rests on the assumption that the spike counts recorded from non-stimulated electrodes are the cultured network's own response to stimulation, not electrical crosstalk or stimulus artifacts, and that the single culture used is representative of cultured neural networks generally.

Editorial extensions

If this is right

  • A linear classifier trained only on the readout can solve spatially overlapping pattern classes, since oriented bars presented at the same location on the array reached 92% average accuracy.
  • Digit-like patterns that are not linearly separable in input space become linearly separable in the biological feature space, with 95% average accuracy.
  • The biological reservoir can be compared directly to an artificial echo state network of the same output dimension, with accuracy in the same range: 98% versus 82% on point stimuli, 92% versus 98% on bars, and 95% versus 97% on digits.
  • The stimulation and recording protocol, 25 repetitions, 10 s inter-stimulus interval, and 10 ms spike-count window, provides a reusable recipe for driving the network and reading its state for downstream classifiers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the reservoir is fixed and only the readout is trained, the same culture could in principle serve several classification tasks at once, with different linear classifiers reading the same spike-count vectors; the paper does not test this, but it is a direct corollary of the architecture.
  • The current protocol discards timing information by counting spikes in a 10 ms window, so replacing or augmenting counts with first-spike latencies or population spike timings may increase separability; this is a testable extension.
  • A decisive reproducibility test would run the same protocol on additional independently cultured networks: if accuracy is stable, the result reflects general biological computation, and if not, the claim concerns this particular culture's connectivity.
  • The comparison with the echo state network suggests the biological and artificial reservoirs occupy different regimes, with the biological one excelling on isolated point stimuli and trailing on overlapping bars; understanding which input statistics favor each substrate could guide when to choose a wet reservoir.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper introduces biological reservoir computing (BRC), in which a cultured neuronal network on a 4096-electrode MEA serves as the untrained nonlinear reservoir. Inputs are mapped to designated stimulation electrodes, and the network response is quantified as the spike-count difference on all electrodes over a 10 ms window (Eq. 2), yielding a high-dimensional feature vector. A single-layer perceptron is trained on these features for three tasks: pointwise stimuli, oriented bars, and digit recognition. The reported accuracies are 98%, 92%, and 95% respectively, with an ESN as a comparison benchmark. The authors conclude that cultured biological neurons can serve as a viable substrate for reservoir computing tasks.

Significance. If the central claim holds, the paper would provide a concrete demonstration of biological reservoir computing with a high-density MEA and would be relevant to neuromorphic and bio-hybrid systems. The paper has several strengths: the stimulation and recording protocol is described in sufficient detail to be reproduced, the response definition in Eq. (2) is explicit, standard spike detection (PTSD) is used, and the task design progresses from simple pointwise stimuli to spatially overlapping bars and digits. The main limitations are experimental controls, statistical power, and generalizability, which currently leave the feasibility claim plausible but not firmly established.

major comments (4)
  1. [Section 3.2, Eq. (2)] The central claim that the 4096-dimensional response vectors encode network-propagated biological computation is not yet supported because no control rules out stimulation artifacts. Since recording and stimulating electrodes share the same MEA, any stimulation-site-dependent residual transient, reference shift, or volume-conduction effect would produce class-discriminative features even in the absence of neural computation. The only stated safeguard is the exclusion of a square region around the stimulated area, but no comparison is reported with and without this exclusion, and no cell-free MEA or pharmacologically silenced culture control is described. I request such control experiments, or an analysis showing that the discriminative structure disappears when the biological signal is blocked, before the biological origin of the features can be accepted.
  2. [Section 5] The conclusion that 'cultured biological neurons can indeed serve as a viable substrate for RC tasks' is a general claim, but the paper does not report how many independent cultures were used. If, as it appears, all experiments come from a single culture, the results cannot support the general feasibility claim because they may reflect culture-specific properties. I ask for replication across multiple independent cultures and a report of inter-culture variability, or a clear statement limiting the conclusion to the specific culture studied.
  3. [Section 3.2, Table 1] With Ntesting = 5 per class, the reported accuracies have very wide sampling variability (e.g., Bar 3: 87% ± 27%, Digit 0: 87% ± 19%), and no statistical test against chance or between systems is provided. A five-sample test set gives coarse resolution; exact binomial confidence intervals should be reported, and a permutation or binomial test should be used to establish above-chance performance for each scenario. This is necessary to support the quantitative accuracy claims in Table 1.
  4. [Section 4, Table 1] The text states that the artificial ESN 'should be considered an upper bound,' yet the BRC system outperforms the ESN on pointwise stimuli (98% vs 82% average). This contradiction suggests that the ESN setup is not a faithful upper-bound benchmark; the single-time-step dynamics and the way input noise is injected are not obviously equivalent to the biological protocol. I ask the authors to either revise the benchmark claim or provide a matched ESN configuration and explain the pointwise discrepancy.
minor comments (5)
  1. [Abstract] The abstract contains 'abiological reservoir computing' where the intended phrase is 'a biological reservoir computing'; this should be corrected.
  2. [Fig. 4 caption] The caption refers to 'the MAE' in the description of electrode colors; this should be 'MEA'.
  3. [Section 3.2] The text states that a 4096-dimensional feature vector is produced, but also that a square region around the stimulated area is excluded; please clarify whether the vector is zero-padded or the dimensionality is reduced, and specify the size of the excluded region.
  4. [Section 2] The related-work discussion would benefit from a more explicit comparison with the closely related studies cited as [13] and [53], which also use biological neurons in reservoir-like frameworks; the current text mentions them only briefly.
  5. [Section 3.1] The visualization tool uses confidence intervals based on a Student's t-estimator, but the number of repetitions N = 25 is small; reporting the exact degrees of freedom or showing the underlying response distributions would improve transparency.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: classification accuracy is measured on held-out biological responses, and no fitted parameter or self-citation is renamed as a prediction.

full rationale

The paper's central claim is experimental: cultured neurons, stimulated through a subset of MEA electrodes, produce spike-count feature vectors (Eq. 2) on which a single-layer perceptron is trained and then tested on withheld responses (Secs. 3.2, 3.3). The only trained parameters are the perceptron weights, fit to N_training = 20 responses per class and evaluated on N_testing = 5 withheld responses; the test responses are not used in fitting. The stimulation parameters (amplitude, pulse width, inter-pulse interval) are described as calibrated for eliciting reliable neuronal responses, not fitted to class labels, so the reported accuracies in Table 1 are not forced by construction. The ESN comparison is an external benchmark rather than an input to the claim. References to prior MEA and RC work (e.g., [47], [13], [53]) are contextual or external; no load-bearing uniqueness theorem or ansatz is imported from the authors' own prior work. The absence of blank-array or silenced-culture controls is a potential experimental confound regarding whether the signal is biological, but it is a validity concern, not a circularity of the derivation chain. Therefore the derivation is self-contained and no circular step is present.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on hand-calibrated stimulation parameters and on domain assumptions inherited from prior neurobiological literature. No code, data, or formal proofs are provided, so the reader must take these assumptions on faith.

free parameters (6)
  • temporal chunk C = 10 ms
    Chosen as about two synaptic transmission cycles (ref 48); directly defines the spike-count feature vector.
  • number of pulses per stimulus N = 25
    Selected to give a distribution of responses for statistics; no sensitivity analysis reported.
  • inter-stimulus interval T = 10 s
    Assumed sufficient for the network to return to resting state (ref 47); hand-selected.
  • stimulus amplitude A = 10 µA for pointwise and bars, 4 µA for digits
    Calibrated to elicit responses while avoiding electrode damage; not derived.
  • pulse width delta = 20 µs
    Part of the calibrated stimulation protocol.
  • number of training epochs = 20
    Chosen empirically to ensure convergence; no validation set or sensitivity analysis.
assumptions (5)
  • domain assumption Cultured neurons derived from stem cells form randomly interconnected networks with complex nonlinear dynamics.
    Stated in Sec. 3; the reservoir property is assumed from neurobiological literature, not measured in this paper.
  • domain assumption The 10 ms temporal chunk captures the stimulus-evoked response (about two synaptic transmission cycles).
    Sec. 3.2, citing ref 48; if the window is wrong, features may miss or dilute the response.
  • domain assumption A 10 s inter-stimulus interval lets the network return to its resting state.
    Sec. 3.2, citing ref 47; carry-over effects could otherwise confound the features.
  • domain assumption The activity recorded on non-stimulated electrodes reflects biological network propagation and not electrical crosstalk.
    Sec. 3.2; no control experiment such as a blank MEA or non-neuronal culture is reported.
  • standard math Standard statistical tools (Student's t-test, SGD, cross-entropy) apply to the recorded spike counts.
    Used in Sec. 3.1 and 3.2; unproblematic but presumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Neurons to Computation: Biological Reservoir Computing for Pattern Recognition." pith.science (2026). https://pith.science/paper/BI4VJO4Q

@misc{pith2026250503510,
  author       = {Pith},
  title        = {Pith review of: From Neurons to Computation: Biological Reservoir Computing for Pattern Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BI4VJO4Q}},
  note         = {Machine review of arXiv:2505.03510}
}
read the original abstract

In this paper, we introduce a paradigm for reservoir computing (RC) that leverages a pool of cultured biological neurons as the reservoir substrate, creating a biological reservoir computing (BRC). This system operates similarly to an echo state network (ESN), with the key distinction that the neural activity is generated by a network of cultured neurons, rather than being modeled by traditional artificial computational units. The neuronal activity is recorded using a multi-electrode array (MEA), which enables high-throughput recording of neural signals. In our approach, inputs are introduced into the network through a subset of the MEA electrodes, while the remaining electrodes capture the resulting neural activity. This generates a nonlinear mapping of the input data to a high-dimensional biological feature space, where distinguishing between data becomes more efficient and straightforward, allowing a simple linear classifier to perform pattern recognition tasks effectively. To evaluate the performance of our proposed system, we present an experimental study that includes various input patterns, such as positional codes, bars with different orientations, and a digit recognition task. The results demonstrate the feasibility of using biological neural networks to perform tasks traditionally handled by artificial neural networks, paving the way for further exploration of biologically-inspired computing systems, with potential applications in neuromorphic engineering and bio-hybrid computing.

Figures

Figures reproduced from arXiv: 2505.03510 by the authors.

Figure 1
Figure 1. Overview of our biological reservoir computing (BRC) approach. A multi-electrode array (MEA) serves as an interface to both stimulate and record activity from a cultured biological neural network. Each input is mapped to a specific subset of MEA electrodes, which deliver the corresponding stimulus to the network. The resulting spiking activity is recorded by a different set of electrodes and assembled into a vector … view at source ↗
Figure 2
Figure 2. Schematic illustration of a cultured biological network interfaced with the MEA device. Each square represents an individual MEA electrode. Input patterns are mapped onto the MEA by assigning elements of the patterns to specific electrodes. Electrical pulses are delivered based on the corresponding input intensities, and the evoked activity of the network is recorded via the remaining electrodes. The resulting spiki… view at source ↗
Figure 3
Figure 3. Visualization of the average network response to N stimuli over a temporal chunk C. Each pixel represents an electrode in the 64 × 64 MEA grid, with the color intensity encoding the average level of response following the stimuli. We set C = 10 ms and N = 25. amount of data available for both training and evaluation. The classifier is trained for 20 epochs, a value determined empirically to ensure convergence of the… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visual representation of the input patterns used across the three experimental scenarios. The first row displays the pointwise stimuli experiments, where single pairs of adjacent electrodes were stimulated. The second row shows the oriented bar experiments with four ca…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 30 canonical work pages

  1. [1]

    Abbott, L.F., van Vreeswijk, C.: Asynchronous states in networks of pulse-coupled oscillators. Phys. Rev. E48, 1483–1490 (Aug 1993). https://doi.org/10.1103/PhysRevE.48.1483

  2. [2]

    Badar, A., Varma, A., Staniec, A., Gamal, M., Magdy, O., Iqbal, H., Arani, E., Zonooz, B.: Highlighting the importance of reducing research bias and carbon emissions in cnns. In: AIxIA 2021 - Advances in ArtificialIntelligence-20thInternationalConferenceoftheItalianAssociationforArtificialIntelligence, Virtual Event, December 1-3, 2021, Revised Selected P...

  3. [3]

    IEEE Trans

    Bianchi, F.M., Scardapane, S., Løkse, S., Jenssen, R.: Reservoir computing approaches for representation and classification of multivariate time series. IEEE Trans. Neural Net- works Learn. Syst. 32(5), 2169–2179 (2021). https://doi.org/10.1109/TNNLS.2020.3001377, https://doi.org/10.1109/TNNLS.2020.3001377

  4. [4]

    Advanced Materials 14(17), 1190– 1193 (2002)

    Bonifazi, P., Fromherz, P.: Silicon chip for electronic communication between nerve cells by non-invasive interfacing and analog–digital processing. Advanced Materials 14(17), 1190– 1193 (2002). https://doi.org/https://doi.org/10.1002/1521-4095(20020903)14:17<1190::AID- ADMA1190>3.0.CO;2-\#

  5. [5]

    Bottou, L.: On-line learning and stochastic approximations, p. 9–42. Cambridge University Press, USA (1999)

  6. [6]

    In: The 2013 International Joint Con- ference on Neural Networks, IJCNN 2013, Dallas, TX, USA, August 4-9, 2013

    Carlson, K.D., Richert, M., Dutt, N.D., Krichmar, J.L.: Biologically plausible models of homeostasis and STDP: stability and learning in spiking neural networks. In: The 2013 International Joint Con- ference on Neural Networks, IJCNN 2013, Dallas, TX, USA, August 4-9, 2013. pp. 1–8. IEEE (2013). https://doi.org/10.1109/IJCNN.2013.6706961 From Neurons to C...

  7. [7]

    https://doi.org/10.1038/nbt.1529, http://dx.doi.org/10.1038/nbt.1529

    Chambers, S.M., Fasano, C.A., Papapetrou, E.P., Tomishima, M., Sadelain, M., Studer, L.: Highly effi- cientneuralconversionofhumanesandipscellsbydualinhibitionofsmadsignaling.NatureBiotechnol- ogy 27(3), 275–280 (Mar 2009). https://doi.org/10.1038/nbt.1529, http://dx.doi.org/10.1038/nbt.1529

  8. [8]

    Nature neuroscience 23(12), 1496—1508 (December 2020)

    Chiaradia, I., Lancaster, M.A.: Brain organoids for the study of human neurobiology at the interface of in vitro and in vivo. Nature neuroscience 23(12), 1496—1508 (December 2020). https://doi.org/10.1038/s41593-020-00730-3

Show all 60 references
  1. [9]

    In: Computer Vision - ECCV 2024 Workshops - Milan, Italy, September 29-October 4, 2024, Proceedings, Part XIV

    Ciampi, L., Lagani, G., Amato, G., Falchi, F.: A biologically-inspired approach to biomedical image seg- mentation. In: Computer Vision - ECCV 2024 Workshops - Milan, Italy, September 29-October 4, 2024, Proceedings, Part XIV. Lecture Notes in Computer Science, vol. 15636, pp....

  2. [10]

    CoRR abs/2412.03192 (2024)

    Ciampi, L., Lagani, G., Amato, G., Falchi, F.: Biologically-inspired semi-supervised semantic segmentation for biomedical imaging. CoRR abs/2412.03192 (2024). https://doi.org/10.48550/ARXIV.2412.03192

  3. [11]

    In: 29th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2021, Online event (Bruges, Belgium), October 6-8, 2021 (2021)

    Cossu, A., Bacciu, D., Carta, A., Gallicchio, C., Lomonaco, V.: Continual learning with echo state networks. In: 29th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2021, Online event (Bruges, Belgium), October 6-8, 202...

  4. [12]

    Frontiers Comput

    Diehl, P.U., Cook, M.: Unsupervised learning of digit recognition using spike-timing-dependent plastic- ity. Frontiers Comput. Neurosci.9, 99 (2015). https://doi.org/10.3389/FNCOM.2015.00099

  5. [13]

    Biosystems 95(2), 90–97 (2009)

    Dockendorf, K.P., Park, I., He, P., Príncipe, J.C., DeMarse, T.B.: Liquid state machines and cultured cortical networks: The separation property. Biosystems 95(2), 90–97 (2009). https://doi.org/https://doi.org/10.1016/j.biosystems.2008.08.001

  6. [14]

    Euro- pean Journal of Neuroscience27(1), 217–227 (2008)

    Durstewitz, D., Deco, G.: Computational significance of transient dynamics in cortical networks. Euro- pean Journal of Neuroscience27(1), 217–227 (2008). https://doi.org/https://doi.org/10.1111/j.1460- 9568.2007.05976.x

  7. [15]

    Neurocomputing114, 3–8 (2013)

    Ferrández, J.M., Lorente, V., de la Paz, F., Fernández, E.: Training biological neural cultures: Towards hebbian learning. Neurocomputing114, 3–8 (2013). https://doi.org/10.1016/J.NEUCOM.2012.09.031

  8. [16]

    Frontiers Comput

    Ferré, P., Mamalet, F., Thorpe, S.J.: Unsupervised feature learning with winner-takes-all based STDP. Frontiers Comput. Neurosci.12, 24 (2018). https://doi.org/10.3389/FNCOM.2018.00024

  9. [17]

    CoRRabs/1712.04323 (2017)

    Gallicchio, C., Micheli, A.: Deep echo state network (deepesn): A brief survey. CoRRabs/1712.04323 (2017)

  10. [18]

    Cam- bridge University Press (2002)

    Gerstner, W., Kistler, W.M.: Spiking Neuron Models: Single Neurons, Populations, Plasticity. Cam- bridge University Press (2002). https://doi.org/10.1017/CBO9780511815706

  11. [19]

    Neuron 91(2), 320– 327 (2016)

    Goel, A., Buonomano, D.V.: Temporal interval learning in cortical cul- tures is encoded in intrinsic network dynamics. Neuron 91(2), 320– 327 (2016). https://doi.org/https://doi.org/10.1016/j.neuron.2016.05.042, https://www.sciencedirect.com/science/article/pii/S0896627316302537

  12. [20]

    Neuron 95(2), 245–258 (Jul 2017)

    Hassabis, D., Kumaran, D., Summerfield, C., Botvinick, M.: Neuroscience-inspired artifi- cial intelligence. Neuron 95(2), 245–258 (Jul 2017). https://doi.org/10.1016/j.neuron.2017.06.011, http://dx.doi.org/10.1016/j.neuron.2017.06.011

  13. [21]

    Huang, G.: An insight into extreme learning machines: Random neurons, random features and kernels. Cogn. Comput.6(3), 376–390 (2014). https://doi.org/10.1007/S12559-014-9255-2

  14. [22]

    Neurocomputing 70(1-3), 489–501 (2006)

    Huang, G., Zhu, Q., Siew, C.K.: Extreme learning machine: Theory and applications. Neurocomputing 70(1-3), 489–501 (2006). https://doi.org/10.1016/J.NEUCOM.2005.12.126

  15. [23]

    https://doi.org/10.1016/J.NEUNET.2019.06.001

    Illing, B., Gerstner, W., Brea, J.: Biologically plausible deep learning - but how far can we go with shal- low networks? Neural Networks118, 90–101 (2019). https://doi.org/10.1016/J.NEUNET.2019.06.001

  16. [24]

    PLoS Comput

    Isomura, T., Kotani, K., Jimbo, Y.: Cultured cortical neurons can perform blind source separation according to the free-energy principle. PLoS Comput. Biol. 11(12) (2015). https://doi.org/10.1371/JOURNAL.PCBI.1004643

  17. [25]

    echo state

    Jaeger, H.: The “echo state” approach to analysing and training recurrent neural networks-with an erratum note. Bonn, Germany: German National Research Center for Information Technology GMD Technical Report148(34), 13 (2001) 14 L. Iannello, L. Ciampi, G. Lagani et al

  18. [26]

    The American Journal of Clinical Nutrition 91(4), 907–912 (2010)

    Javed, F., He, Q., Davidson, L.E., Thornton, J.C., Albu, J., Boxt, L., Krasnow, N., Elia, M., Kang, P., Heshka, S., Gallagher, D.: Brain and high metabolic rate organ mass: contributions to resting energy expenditure beyond fat-free mass1234. The American Journal of Clinical N...

  19. [27]

    In: The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5,

    Journé, A., Rodriguez, H.G., Guo, Q., Moraitis, T.: Hebbian deep learning without feedback. In: The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5,

  20. [28]

    Neuron 110(23), 3952–3969.e8 (2022)

    Kagan, B.J., Kitchen, A.C., Tran, N.T., Habibollahi, F., Khajehnejad, M., Parker, B.J., Bhat, A., Rollo, B., Razi, A., Friston, K.J.: In vitro neurons learn and exhibit sen- tience when embodied in a simulated game-world. Neuron 110(23), 3952–3969.e8 (2022). https://doi.org/ht...

  21. [29]

    Neural Comput

    Kline, D., Berardi, V.L.: Revisiting squared-error and cross-entropy functions for training neural net- work classifiers. Neural Comput. Appl. 14(4), 310–318 (2005). https://doi.org/10.1007/S00521-005- 0467-Y

  22. [30]

    Krotov, D., Hopfield, J.J.: Unsupervised learning by competing hidden units. Proc. Natl. Acad. Sci. USA 116(16), 7723–7731 (2019). https://doi.org/10.1073/PNAS.1820458116

  23. [31]

    In: CBMI 2022: International Conference on Content- based Multimedia Indexing, Graz, Austria, September 14 - 16, 2022

    Lagani, G., Bacciu, D., Gallicchio, C., Falchi, F., Gennaro, C., Amato, G.: Deep features for CBIR with scarce data using hebbian learning. In: CBMI 2022: International Conference on Content- based Multimedia Indexing, Graz, Austria, September 14 - 16, 2022. pp. 136–141. ACM (...

  24. [32]

    In: Machine Learning, Optimization, and Data Science - 7th International Conference, LOD 2021, Grasmere, UK, October 4-8, 2021, Revised Selected Papers, Part I

    Lagani,G.,Falchi,F.,Gennaro,C.,Amato,G.:Trainingconvolutionalneuralnetworkswithcompetitive hebbian learning approaches. In: Machine Learning, Optimization, and Data Science - 7th International Conference, LOD 2021, Grasmere, UK, October 4-8, 2021, Revised Selected Papers, Part...

  25. [33]

    Neural Comput

    Lagani, G., Falchi, F., Gennaro, C., Amato, G.: Comparing the performance of hebbian against back- propagation learning using convolutional neural networks. Neural Comput. Appl.34(8), 6503–6519 (2022). https://doi.org/10.1007/S00521-021-06701-4, https://doi.org/10.1007/s00521-...

  26. [34]

    CoRRabs/2307.16235 (2023)

    Lagani, G., Falchi, F., Gennaro, C., Amato, G.: Spiking neural networks and bio-inspired supervised deep learning: A survey. CoRRabs/2307.16235 (2023). https://doi.org/10.48550/ARXIV.2307.16235

  27. [35]

    CoRR abs/2307.16236 (2023)

    Lagani, G., Falchi, F., Gennaro, C., Amato, G.: Synaptic plasticity models and bio-inspired unsupervised deep learning: A survey. CoRR abs/2307.16236 (2023). https://doi.org/10.48550/ARXIV.2307.16236

  28. [36]

    Neurocomputing 595, 127867 (2024)

    Lagani, G., Falchi, F., Gennaro, C., Fassold, H., Amato, G.: Scalable bio-inspired training of deep neural networks with fasthebb. Neurocomputing 595, 127867 (2024). https://doi.org/10.1016/J.NEUCOM.2024.127867

  29. [37]

    In: Similarity Search and Applications - 15th International Conference, SISAP 2022, Bologna, Italy, October 5-7, 2022, Proceedings

    Lagani, G., Gennaro, C., Fassold, H., Amato, G.: Fasthebb: Scaling hebbian training of deep neural networks to imagenet level. In: Similarity Search and Applications - 15th International Conference, SISAP 2022, Bologna, Italy, October 5-7, 2022, Proceedings. Lecture Notes in C...

  30. [38]

    In: 10th International IEEE/EMBS Conference on Neural Engineering, NER 2021, Virtual Event, Italy, May 4-6, 2021

    Lagani, G., Mazziotti, R., Falchi, F., Gennaro, C., Cicchini, G.M., Pizzorusso, T., Cremisi, F., Amato, G.: Assessing pattern recognition performance of neuronal cultures through accurate simulation. In: 10th International IEEE/EMBS Conference on Neural Engineering, NER 2021, ...

  31. [39]

    Frontiers Comput

    Lazar, A., Pipa, G., Triesch, J.: SORN: a self-organizing recurrent neural network. Frontiers Comput. Neurosci. 3, 23 (2009). https://doi.org/10.3389/NEURO.10.023.2009

  32. [40]

    Lukosevicius, M., Jaeger, H.: Reservoir computing approaches to recurrent neural network training. Comput. Sci. Rev.3(3), 127–149 (2009). https://doi.org/10.1016/J.COSREV.2009.03.005

  33. [41]

    Neural Comput

    Maass, W., Natschläger, T., Markram, H.: Real-time computing without stable states: A new frame- work for neural computation based on perturbations. Neural Comput. 14(11), 2531–2560 (2002). https://doi.org/10.1162/089976602760407955 From Neurons to Computation: Biological Rese...

  34. [42]

    Journal of Neuroscience Methods 177(1), 241–249 (2009)

    Maccione, A., Gandolfo, M., Massobrio, P., Novellino, A., Martinoia, S., Chiap- palone, M.: A novel algorithm for precise identification of spikes in extracellularly recorded neuronal signals. Journal of Neuroscience Methods 177(1), 241–249 (2009). https://doi.org/https://doi....

  35. [43]

    Frontiers Comput

    Marblestone, A.H., Wayne, G., Körding, K.P.: Toward an integration of deep learning and neuroscience. Frontiers Comput. Neurosci.10, 94 (2016). https://doi.org/10.3389/FNCOM.2016.00094

  36. [44]

    Scientific Reports10(1), 1–11 (2020)

    Meloni, I., Sachidanandan, D., Thum, A.S., Kittel, R.J., Murawski, C.: Controlling the behaviour of drosophila melanogaster via smartphone optogenetics. Scientific Reports10(1), 1–11 (2020)

  37. [45]

    PLoS Comput

    Pastore, V.P., Massobrio, P., Godjoski, A., Martinoia, S.: Identification of excitatory-inhibitory links and network topology in large-scale neuronal assemblies from multi-electrode recordings. PLoS Comput. Biol. 14(8) (2018). https://doi.org/10.1371/JOURNAL.PCBI.1006381

  38. [46]

    Nature Neuroscience22(11), 1761–1770 (2019)

    Richards, B., Lillicrap, T., Beaudoin, P., Bengio, Y., Bogacz, R., Christensen, A., Clopath, C., Costa, R., de Berker, A., Ganguli, S., Gillon, C., Hafner, D., Kepecs, A., Kriegeskorte, N., Latham, P., Lind- say, G., Miller, K., Naud, R., Pack, C., Poirazi, P., Roelfsema, P., ...

  39. [47]

    IEEE Trans

    Ruaro, M.E., Bonifazi, P., Torre, V.: Toward the neurocomputer: image processing and pat- tern recognition with neuronal cultures. IEEE Trans. Biomed. Eng. 52(3), 371–383 (2005). https://doi.org/10.1109/TBME.2004.842975

  40. [48]

    An- nual Review of Physiology 61(Volume 61, 1999), 521–542 (1999)

    Sabatini, B.L., Regehr, W.G.: Timing of synaptic transmission. An- nual Review of Physiology 61(Volume 61, 1999), 521–542 (1999). https://doi.org/https://doi.org/10.1146/annurev.physiol.61.1.521

  41. [49]

    In: International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2020, Novi Sad, Serbia, August 24-26, 2020

    Sarli, D.D., Gallicchio, C., Micheli, A.: Gated echo state networks: a preliminary study. In: International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2020, Novi Sad, Serbia, August 24-26, 2020. pp. 1–5. IEEE (2020). https://doi.org/10.1109/INIST...

  42. [50]

    Scardapane, S., Butcher, J.B., Bianchi, F.M., Malik, Z.K.: Advances in biologically inspired reservoir computing. Cogn. Comput.9(3), 295–296 (2017). https://doi.org/10.1007/S12559-017-9469-1

  43. [51]

    Journal of Neuroscience21(22), 8782– 8788 (2001)

    Shahaf, G., Marom, S.: Learning in networks of cortical neurons. Journal of Neuroscience21(22), 8782– 8788 (2001). https://doi.org/10.1523/JNEUROSCI.21-22-08782.2001

  44. [52]

    Nature Neuroscience3(9), 919–926 (Sep 2000)

    Song, S., Miller, K.D., Abbott, L.F.: Competitive hebbian learning through spike-timing-dependent synaptic plasticity. Nature Neuroscience3(9), 919–926 (Sep 2000). https://doi.org/10.1038/78829

  45. [53]

    Proceedings of the National Academy of Sciences120(25) (2023)

    Sumi, T., Yamamoto, H., Katori, Y., Ito, K., Moriya, S., Konno, T., Sato, S., Hirano-Iwata, A.: Biolog- ical neurons act as generalization filters in reservoir computing. Proceedings of the National Academy of Sciences120(25) (2023). https://doi.org/10.1073/pnas.2217008120

  46. [54]

    Sun, Z., Cutsuridis, V., Caiafa, C.F., Solé-Casals, J.: Brain simulation and spiking neural networks. Cogn. Comput.15(4), 1103–1105 (2023). https://doi.org/10.1007/S12559-023-10156-1

  47. [55]

    In: André, E., Koenig, S., Dastani, M., Sukthankar, G

    Tenenbaum, J.: Building machines that learn and think like people. In: André, E., Koenig, S., Dastani, M., Sukthankar, G. (eds.) Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, AAMAS 2018, Stockholm, Sweden, July 10-15, 2018. p. 5....

  48. [56]

    In: 23rd International Conference on Pattern Recognition, ICPR 2016, Cancún, Mexico, December 4-8,

    Wang, Q., Li, P.: D-LSM: deep liquid state machine with unsupervised recurrent reservoir tuning. In: 23rd International Conference on Pattern Recognition, ICPR 2016, Cancún, Mexico, December 4-8,

  49. [57]

    Frontiers Comput

    Yonemura, Y., Katori, Y.: Dynamical predictive coding with reservoir computing per- forms noise-robust multi-sensory speech recognition. Frontiers Comput. Neurosci. 18 (2024). https://doi.org/10.3389/FNCOM.2024.1464603

  50. [58]

    IEEE Trans

    Zhang, Y., Li, P., Jin, Y., Choe, Y.: A digital liquid state machine with biologically inspired learning and its application to speech recognition. IEEE Trans. Neural Networks Learn. Syst.26(11), 2635–2649 (2015). https://doi.org/10.1109/TNNLS.2015.2388544

  51. [2016]

    2652–2657

    pp. 2652–2657. IEEE (2016). https://doi.org/10.1109/ICPR.2016.7900035

  52. [2023]

    OpenReview.net (2023)

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.