pith. sign in

arxiv: 2605.19911 · v1 · pith:5URVSHQZnew · submitted 2026-05-19 · ⚛️ physics.optics · cs.NE· nlin.CD

Reconfigurable Nonlinear Photonic Networks for In-Situ Learning and Memory Formation via Driven-Dissipative Dynamics

Isaac Yorke This is my paper

Pith reviewed 2026-05-20 04:36 UTC · model grok-4.3

classification ⚛️ physics.optics cs.NEnlin.CD
keywords reconfigurable photonic networksdriven-dissipative dynamicsneuromorphic computingin-situ learningmemory formationbistable statesnonlinear photonicsphotonic hardware
0
0 comments X

The pith

A reconfigurable nonlinear photonic network enables in-situ learning and memory formation directly from driven-dissipative dynamics.

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

This paper proposes a Reconfigurable Nonlinear Photonic Decision Network in which computation, memory, and learning arise from the physical dynamics of light rather than from external digital processors. Numerical simulations show that the network adapts its states through local physical rules, balances stability against plasticity using decay and hysteresis, forms and erases memories with bistable photonic configurations, and exhibits fading memory while respecting realistic saturation and dissipation. A sympathetic reader would care because the approach points toward photonic hardware that can learn and remember intrinsically at high speed and low energy.

Core claim

The author proposes the RNPDN as a physically grounded neuromorphic framework in which computation, memory, and learning emerge directly from driven-dissipative dynamics. Numerical simulations establish the simultaneous realization of local physical learning rules enabling adaptive state evolution, a tunable stability-plasticity tradeoff governed by decay and hysteresis mechanisms, controlled memory formation and erasure via bistable photonic states, fading memory, in-situ learning, and hardware-faithful nonlinear dynamics incorporating saturation and dissipation.

What carries the argument

The Reconfigurable Nonlinear Photonic Decision Network (RNPDN), whose driven-dissipative photonic states support adaptive evolution, bistability for memory, and local learning rules.

If this is right

  • Local physical learning rules produce adaptive state evolution inside the photonic layer itself.
  • Decay and hysteresis mechanisms allow external tuning of the stability-plasticity tradeoff.
  • Bistable photonic states enable deliberate formation and erasure of memories.
  • The network supports both transient fading memory and persistent memory within the same hardware.
  • Saturation and dissipation effects remain present, keeping the dynamics faithful to actual photonic components.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If built in hardware, the approach could reduce reliance on hybrid digital-photonic interfaces for neuromorphic tasks.
  • Similar driven-dissipative principles might be explored in other physical platforms such as optomechanical or fluidic systems for intrinsic computation.
  • Varying network size or topology in simulation could reveal scaling limits before hardware fabrication.
  • Adding realistic fabrication noise or loss to the model would test robustness of the reported memory and learning behaviors.

Load-bearing premise

Numerical simulations of driven-dissipative dynamics are assumed to faithfully represent how learning and memory would emerge in real physical photonic hardware without additional external mechanisms.

What would settle it

Fabricating a physical RNPDN device and testing whether it exhibits adaptive state evolution, tunable memory formation, and in-situ learning under driven-dissipative conditions without external training algorithms would confirm or refute the central claim.

Figures

Figures reproduced from arXiv: 2605.19911 by Isaac Yorke.

Figure 2
Figure 2. Figure 2: Bistable switching behaviour of the photonic memory state. The state [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: characterizes the fading memory of the photonic system by plotting the weight evolution under exponential decay, W(t) ≈ e −γt, for different values of decay parameter γ. The decay parameter controls the memory timescale: smaller γ (e.g., 1×10−5 ) yields a slowly decaying weight that retains information over long temporal windows, while larger γ (e.g., 1 × 10−2 ) yields rapid decay, effectively erasing past… view at source ↗
Figure 5
Figure 5. Figure 5: Quantitative demonstration of system memory lifetime vs decay, [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Demonstration of weight saturation as not being a numerical artifact. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Multi-channel operation of the RNPDN. Four independent channels [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Scalable photonic memory array of RNPDN. All four weight [PITH_FULL_IMAGE:figures/full_fig_p007_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Time evolution of the reward signal during In-situ learning. Fast [PITH_FULL_IMAGE:figures/full_fig_p008_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Time evolution of weight during In-situ learning. There is rapid [PITH_FULL_IMAGE:figures/full_fig_p008_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Time evolution of system output. There is high initial value ( [PITH_FULL_IMAGE:figures/full_fig_p008_13.png] view at source ↗
read the original abstract

Photonic neuromorphic computing offers a promising route to overcoming the limitations of conventional von Neumann architectures by exploiting the high bandwidth, low latency, and massive parallelism of optical systems. However, most existing implementations rely on fixed dynamical substrates such as classic reservoir computing, where learning is restricted to external readout layers and memory is limited to transient fading effects. In this work, I propose a Reconfigurable Nonlinear Photonic Decision Network (RNPDN), a physically grounded neuromorphic framework in which computation, memory, and learning emerge directly from driven-dissipative dynamics. Through numerical simulations, I demonstrate the simultaneous realization of key properties: local physical learning rules enabling adaptive state evolution, a tunable stability-plasticity tradeoff governed by decay and hysteresis mechanisms, controlled memory formation and erasure via bistable photonic states, fading memory, in-situ learning, and hardware-faithful nonlinear dynamics incorporating saturation and dissipation. In contrast to conventional approaches, the proposed system enables intrinsic adaptation within the physical layer while supporting both transient and persistent memory. These results establish a unified framework for adaptive photonic information processing and provide a pathway toward scalable and energy-efficient neuromorphic photonic hardware.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a Reconfigurable Nonlinear Photonic Decision Network (RNPDN) in which computation, memory, and learning emerge directly from driven-dissipative dynamics in photonic hardware. Through numerical simulations, it claims simultaneous realization of local physical learning rules for adaptive state evolution, a tunable stability-plasticity tradeoff via decay and hysteresis, controlled memory formation and erasure using bistable photonic states, fading memory, in-situ learning, and hardware-faithful nonlinear dynamics that incorporate saturation and dissipation.

Significance. If the reported simulation results are substantiated with explicit equations and parameters and map faithfully to physical devices, the work would offer a meaningful advance over fixed reservoir-computing approaches in photonic neuromorphic systems by enabling intrinsic adaptation and both transient and persistent memory within the physical layer itself.

major comments (2)
  1. [Abstract] Abstract and main text: The central claims rest entirely on numerical simulations demonstrating emergent learning and memory properties, yet no governing driven-dissipative equations, parameter values, integration scheme, or quantitative results (e.g., time traces, error bars, or specific configurations) are provided, rendering verification of the listed properties impossible.
  2. [Main text] The manuscript assumes that idealized driven-dissipative dynamics without stochastic terms (thermal or shot noise) or device-specific nonlinearities will produce robust in-situ learning and persistent bistable memory; this assumption is load-bearing for the hardware-faithful claim but is not tested or justified against realistic noise models.
minor comments (2)
  1. The acronym RNPDN is defined on first use, but subsequent references would benefit from consistent expansion or a dedicated nomenclature section for clarity.
  2. Figure captions (if present) should explicitly state the parameter values and initial conditions used in each simulation panel to support reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback, which has helped us identify areas where the manuscript's presentation and robustness claims can be strengthened. We address each major comment below and have revised the manuscript to incorporate additional details and analyses.

read point-by-point responses

  1. Referee: [Abstract] Abstract and main text: The central claims rest entirely on numerical simulations demonstrating emergent learning and memory properties, yet no governing driven-dissipative equations, parameter values, integration scheme, or quantitative results (e.g., time traces, error bars, or specific configurations) are provided, rendering verification of the listed properties impossible.

    Authors: We acknowledge that the submitted manuscript did not include explicit governing equations, specific parameter values, the numerical integration scheme, or quantitative simulation outputs such as time traces and error bars. This omission limits independent verification. In the revised manuscript, we have added a new Methods subsection that presents the full driven-dissipative model equations, all numerical parameter values, the integration method (fourth-order Runge-Kutta with adaptive step size), and representative quantitative results including time traces from multiple realizations with error bars. These additions enable direct verification of the claimed emergent properties. revision: yes

  2. Referee: [Main text] The manuscript assumes that idealized driven-dissipative dynamics without stochastic terms (thermal or shot noise) or device-specific nonlinearities will produce robust in-situ learning and persistent bistable memory; this assumption is load-bearing for the hardware-faithful claim but is not tested or justified against realistic noise models.

    Authors: The referee is correct that the original simulations were performed in the deterministic driven-dissipative regime without explicit stochastic noise terms, even though saturation and dissipation were included to approximate hardware behavior. We have not previously tested robustness against thermal or shot noise. In the revision, we now include a dedicated analysis subsection that incorporates stochastic noise models calibrated to typical photonic device parameters. We show through additional simulations that the local learning rules, stability-plasticity tradeoff, and bistable memory formation persist under moderate noise levels, with quantitative metrics (e.g., learning convergence rates and memory retention times) demonstrating only graceful degradation. This directly addresses the hardware-faithful claim. revision: yes

Circularity Check

0 steps flagged

No circularity: proposal via simulation with no derivation chain

full rationale

The manuscript presents a new RNPDN framework whose claimed properties (local learning rules, stability-plasticity tradeoff, bistable memory, in-situ learning) are shown through numerical simulations of driven-dissipative dynamics. No equations, fitted parameters, or self-citations appear that reduce any prediction to its own inputs by construction. The work is framed as a physically grounded proposal rather than a derivation from prior results, so the central claims remain independent of the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The paper adds the RNPDN concept and the assertion that its listed behaviors emerge from driven-dissipative dynamics; all supporting evidence is stated to come from numerical simulations whose details are not supplied.

axioms (1)
  • domain assumption Driven-dissipative nonlinear photonic dynamics can be numerically simulated to produce local learning rules, bistable memory states, and a stability-plasticity tradeoff.
    This premise is invoked to justify the demonstration of all claimed properties through simulation.
invented entities (1)
  • Reconfigurable Nonlinear Photonic Decision Network (RNPDN) no independent evidence
    purpose: To provide a physically grounded substrate in which computation, memory, and learning emerge directly from driven-dissipative dynamics.
    The entity is introduced in the abstract as the central new framework.

pith-pipeline@v0.9.0 · 5734 in / 1585 out tokens · 69071 ms · 2026-05-20T04:36:30.291582+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages

  1. [1]

    Reservoir computing based on a sil- icon microring and time multiplexing for binary and analog operations,

    M. Borghi, S. Biasi, and L. Pavesi, “Reservoir computing based on a sil- icon microring and time multiplexing for binary and analog operations,” Scientific Reports, vol. 11, no. 1, p. 15642, 2021

    work page 2021

  2. [2]

    Deep photonic reservoir computing based on microring resonators,

    Y . Dong, Q. Zhang, Y . Deng, X. Zou, W. Pan, and L. Yan, “Deep photonic reservoir computing based on microring resonators,”Optics & Laser Technology, vol. 196, p. 114614, 2026

    work page 2026

  3. [3]

    On-chip silicon photonic neural networks based on thermally tunable microring resonators for recognition tasks,

    H. Zhang, B. Huang, C. Cheng, B. Jiang, L. Bao, and Y . Xie, “On-chip silicon photonic neural networks based on thermally tunable microring resonators for recognition tasks,” inPhotonics. MDPI, 2025, p. 640

    work page 2025

  4. [4]

    Photonic reservoir computing system for pattern recognition based on an array of four distributed feedback lasers,

    X. Guo, H. Zhou, S. Xiang, Q. Yu, Y . Zhang, Y . Han, and Y . Hao, “Photonic reservoir computing system for pattern recognition based on an array of four distributed feedback lasers,”ACS Photonics, vol. 11, no. 3, pp. 1327–1334, 2024

    work page 2024

  5. [5]

    Reser- voir computing with all-optical non-fading memory in a self-pulsing microresonator network,

    A. Lugnan, S. Biasi, A. Foradori, P. Bienstman, and L. Pavesi, “Reser- voir computing with all-optical non-fading memory in a self-pulsing microresonator network,”Advanced Optical Materials, vol. 13, no. 11, p. 2403133, 2025

    work page 2025

  6. [6]

    Deep learning with coherent nanophotonic circuits,

    Y . Shen, N. C. Harris, S. Skirlo, M. Prabhu, T. Baehr-Jones, M. Hochberg, X. Sun, S. Zhao, H. Larochelle, D. Englundet al., “Deep learning with coherent nanophotonic circuits,”Nature photonics, vol. 11, no. 7, pp. 441–446, 2017

    work page 2017

  7. [7]

    All-optical neural network with nonlinear activation functions,

    Y . Zuo, B. Li, Y . Zhao, Y . Jiang, Y .-C. Chen, P. Chen, G.-B. Jo, J. Liu, and S. Du, “All-optical neural network with nonlinear activation functions,”Optica, vol. 6, no. 9, pp. 1132–1137, 2019

    work page 2019

  8. [8]

    Erneux and P

    T. Erneux and P. Glorieux,Laser dynamics. Cambridge University Press, 2010

    work page 2010

  9. [9]

    Laser theory,

    H. Haken, “Laser theory,”Atoms, Molecules and Lasers, p. 283, 1973

    work page 1973

  10. [10]

    Parallel photonic information processing at gigabyte per second data rates using transient states,

    D. Brunner, M. C. Soriano, C. R. Mirasso, and I. Fischer, “Parallel photonic information processing at gigabyte per second data rates using transient states,”Nature communications, vol. 4, no. 1, p. 1364, 2013

    work page 2013

  11. [11]

    Spatial dissipative structures in passive optical systems,

    L. A. Lugiato and R. Lefever, “Spatial dissipative structures in passive optical systems,”Physical review letters, vol. 58, no. 21, p. 2209, 1987

    work page 1987

  12. [12]

    Introduction to neuromorphic com- puting,

    S. Ramanathan and A. Sengupta, “Introduction to neuromorphic com- puting,”American history, vol. 1861, no. 1900, 1945

    work page 1900

  13. [13]

    Elaydi,An introduction to difference equations

    S. Elaydi,An introduction to difference equations. Springer, 2005

    work page 2005

  14. [14]

    Micrometre-scale silicon electro-optic modulator,

    Q. Xu, B. Schmidt, S. Pradhan, and M. Lipson, “Micrometre-scale silicon electro-optic modulator,”nature, vol. 435, no. 7040, pp. 325– 327, 2005

    work page 2005

  15. [15]

    Rigorous dynamic model of a silicon ring resonator with phase change material for a neuromorphic node,

    A. Lugnan, S. Garc ´ıa-Cuevas Carrillo, C. D. Wright, and P. Bienstman, “Rigorous dynamic model of a silicon ring resonator with phase change material for a neuromorphic node,”Optics Express, vol. 30, no. 14, pp. 25 177–25 194, 2022

    work page 2022

  16. [16]

    Silicon ring resonator with phase-change material as a plastic dynamical node for scalable all-optical neural networks with synaptic plasticity,

    A. Lugnan, S. G.-C. Carrillo, J. Song, S. Aggarwal, F. Br ¨uckerhoff- Pl¨uckelmann, W. H. Pernice, H. Bhaskaran, C. D. Wright, and P. Bien- stman, “Silicon ring resonator with phase-change material as a plastic dynamical node for scalable all-optical neural networks with synaptic plasticity,” in2023 23rd International Conference on Transparent Op- tical N...

    work page 2023

  17. [17]

    Kerr bistability in a 3×3 coupler optical fiber ring resonator,

    Y . Ja, “Kerr bistability in a 3×3 coupler optical fiber ring resonator,” Applied optics, vol. 32, no. 27, pp. 5310–5312, 1993

    work page 1993

  18. [18]

    Multibistability and self-pulsation in nonlinear high-q silicon microring resonators con- sidering thermo-optical effect,

    L. Zhang, Y . Fei, T. Cao, Y . Cao, Q. Xu, and S. Chen, “Multibistability and self-pulsation in nonlinear high-q silicon microring resonators con- sidering thermo-optical effect,”Physical Review A—Atomic, Molecular, and Optical Physics, vol. 87, no. 5, p. 053805, 2013

    work page 2013

  19. [19]

    Phase-controlled non-markovian hysteresis in a nonlinear, non- hermitian silicon microresonator,

    S. Biasi, S. Gretter, B. Aslan, D. Olivieri, R. Franchi, and L. Pavesi, “Phase-controlled non-markovian hysteresis in a nonlinear, non- hermitian silicon microresonator,”Physical Review Applied, vol. 25, no. 2, p. 024041, 2026

    work page 2026

  20. [20]

    Multiscale simulations of ge–sb– se–te phase-change alloys for photonic memory applications,

    H. Zhang, W. Zhou, and W. Zhang, “Multiscale simulations of ge–sb– se–te phase-change alloys for photonic memory applications,”Journal of Materials Informatics, vol. 6, no. 1, pp. N–A, 2026

    work page 2026

  21. [21]

    All-optical set–reset flip–flop based on the passive microring- resonator bistability,

    A. Bahrampour, M. Karimi, M. A. Qamsari, H. R. Nejad, and S. Key- vaninia, “All-optical set–reset flip–flop based on the passive microring- resonator bistability,”Optics Communications, vol. 281, no. 20, pp. 5104–5113, 2008

    work page 2008

  22. [22]

    Symmetric self-electro-optic effect device: Optical set- reset latch,

    A. Lentine, H. Hinton, D. Miller, J. Henry, J. Cunningham, and L. Chirovsky, “Symmetric self-electro-optic effect device: Optical set- reset latch,”Applied Physics Letters, vol. 52, no. 17, pp. 1419–1421, 1988

    work page 1988

  23. [23]

    Circlez: Reliable photometric redshifts for active galactic nuclei computed solely using photometry from legacy survey imaging for desi,

    A. Saxena, M. Salvato, W. Roster, R. Shirley, J. Buchner, J. Wolf, C. Kohl, H. Starck, T. Dwelly, J. Comparatet al., “Circlez: Reliable photometric redshifts for active galactic nuclei computed solely using photometry from legacy survey imaging for desi,”Astronomy & Astro- physics, vol. 690, p. A365, 2024

    work page 2024

  24. [24]

    A refined information processing capacity metric allows an in-depth analysis of memory and nonlinearity trade-offs in neurocomputational systems,

    T. Schulte to Brinke, M. Dick, R. Duarte, and A. Morrison, “A refined information processing capacity metric allows an in-depth analysis of memory and nonlinearity trade-offs in neurocomputational systems,” Scientific reports, vol. 13, no. 1, p. 10517, 2023

    work page 2023

  25. [25]

    The importance of nonlinear den- dritic processing in multimodular memory networks,

    D. Horn, N. Levy, and E. Ruppin, “The importance of nonlinear den- dritic processing in multimodular memory networks,”Neurocomputing, vol. 26, pp. 389–394, 1999

    work page 1999

  26. [26]

    Network mechanisms of working memory: the role of neuronal nonlinearities,

    A. Suarez-Perez, O. Harish, and D. Hansel, “Network mechanisms of working memory: the role of neuronal nonlinearities,”arXiv preprint arXiv:2112.05091, 2021

    work page arXiv 2021

  27. [27]

    A. V . Oppenheim,Discrete-time signal processing. Pearson Education India, 1999

    work page 1999

  28. [28]

    Digital signal processing, upper saddle river,

    J. G. Proakis and D. G. Manolakis, “Digital signal processing, upper saddle river,”New Jersey, vol. 7458, 2007

    work page 2007

  29. [29]

    Gibbs,Optical bistability: controlling light with light

    H. Gibbs,Optical bistability: controlling light with light. Elsevier, 2012

    work page 2012

  30. [30]

    Programmable photonic latch memory,

    F. Ashtiani, “Programmable photonic latch memory,”Optics Express, vol. 33, no. 2, pp. 3501–3510, 2025

    work page 2025

  31. [31]

    Switching between crys- tallization from the glassy and the undercooled liquid phase in phase change material ge2sb2te5,

    J. Pries, S. Wei, M. Wuttig, and P. Lucas, “Switching between crys- tallization from the glassy and the undercooled liquid phase in phase change material ge2sb2te5,”Advanced materials, vol. 31, no. 39, p. 1900784, 2019

    work page 2019

  32. [32]

    Hysteresis in laser cavity-solitons,

    A. Suresh, A. Cutrona, D. Das, S. T. Chu, B. E. Little, R. Moran- dotti, D. J. Moss, J. S. T. Gongora, M. Peccianti, and A. Pasquazi, “Hysteresis in laser cavity-solitons,” in2025 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference (CLEO/Europe-EQEC). IEEE, 2025, pp. 1–1

    work page 2025

  33. [33]

    Investigation of carrier-induced bistability in soi micro-ring resonators for memory applications,

    A. A. Nikitin, I. A. Ryabcev, A. A. Ershov, A. A. Semenov, D. A. Konkin, A. A. Kokolov, L. I. Babak, F. I. Sheyerman, and A. B. Ustinov, “Investigation of carrier-induced bistability in soi micro-ring resonators for memory applications,” inNovel Optical Materials and Applications. Optica Publishing Group, 2022, pp. JW3A–48

    work page 2022

  34. [34]

    Fading memory and the problem of approxi- mating nonlinear operators with volterra series,

    S. Boyd and L. Chua, “Fading memory and the problem of approxi- mating nonlinear operators with volterra series,”IEEE Transactions on circuits and systems, vol. 32, no. 11, pp. 1150–1161, 2003. THIS MANUSCRIPT IS A PREPRINT AND HAS NOT BEEN PEER REVIEWED 10

    work page 2003

  35. [35]

    All-optical reservoir computing,

    F. Duport, B. Schneider, A. Smerieri, M. Haelterman, and S. Massar, “All-optical reservoir computing,”Optics express, vol. 20, no. 20, pp. 22 783–22 795, 2012

    work page 2012

  36. [36]

    W. E. Boyce, R. C. DiPrima, and D. B. Meade,Elementary differential equations and boundary value problems. John Wiley & Sons, 2021

    work page 2021

  37. [37]

    Surrogate gradient learning in spiking neural networks,

    E. O. Neftci, H. Mostafa, and F. Zenke, “Surrogate gradient learning in spiking neural networks,”IEEE Signal Processing Magazine, vol. 36, no. 6, pp. 51–63, 2019

    work page 2019

  38. [38]

    Optimizing memory in reservoir computers,

    T. L. Carroll, “Optimizing memory in reservoir computers,”Chaos: An Interdisciplinary Journal of Nonlinear Science, vol. 32, no. 2, 2022

    work page 2022

  39. [39]

    An introduction to reservoir computing,

    M. t. Vrugt, “An introduction to reservoir computing,”arXiv preprint arXiv:2412.13212, 2024

    work page arXiv 2024

  40. [40]

    The “echo state

    H. Jaeger, “The “echo state” approach to analysing and training recurrent neural networks-with an erratum note,”Bonn, Germany: German na- tional research center for information technology gmd technical report, vol. 148, no. 34, p. 13, 2001

    work page 2001

  41. [41]

    Reservoir computing using dynamic memristors for temporal information process- ing,

    C. Du, F. Cai, M. A. Zidan, W. Ma, S. H. Lee, and W. D. Lu, “Reservoir computing using dynamic memristors for temporal information process- ing,”Nature communications, vol. 8, no. 1, p. 2204, 2017

    work page 2017

  42. [42]

    Memory in reservoirs for high di- mensional input,

    M. Hermans and B. Schrauwen, “Memory in reservoirs for high di- mensional input,” inThe 2010 international joint conference on neural networks (IJCNN). IEEE, 2010, pp. 1–7

    work page 2010

  43. [43]

    The stability-plasticity dilemma: Investigating the continuum from catastrophic forgetting to age-limited learning effects,

    M. Mermillod, A. Bugaiska, and P. Bonin, “The stability-plasticity dilemma: Investigating the continuum from catastrophic forgetting to age-limited learning effects,” p. 504, 2013

    work page 2013

  44. [44]

    Balancing learning plasticity and memory stability: A parameter space strategy for class- incremental learning,

    J. Feng, H. Qiu, L. Zhao, C. Gu, H. Qin, and Y . Xu, “Balancing learning plasticity and memory stability: A parameter space strategy for class- incremental learning,”Neural Networks, vol. 190, p. 107755, 2025

    work page 2025

  45. [45]

    Stability through plasticity: Finding robust memories through representational drift,

    M. Natrajan and J. E. Fitzgerald, “Stability through plasticity: Finding robust memories through representational drift,”Proceedings of the National Academy of Sciences, vol. 122, no. 45, p. e2500077122, 2025

    work page 2025

  46. [46]

    Saturation effects in nonlinear absorption, refraction, and frequency conversion: a review,

    A. S. Rao, “Saturation effects in nonlinear absorption, refraction, and frequency conversion: a review,”Optik, vol. 267, p. 169638, 2022

    work page 2022

  47. [47]

    Absorption saturation and thermo-lensing effects as the sources of nonlinear anisotropic change of refractive index in cr4+: Yag at cw excitation,

    A. V . Kir’yanov and Y . O. Barmenkov, “Absorption saturation and thermo-lensing effects as the sources of nonlinear anisotropic change of refractive index in cr4+: Yag at cw excitation,” inPhotorefractive Effects, Materials, and Devices. Optica Publishing Group, 2003, p. 144

    work page 2003

  48. [48]

    Neural network weights do not converge to stationary points: An invariant measure perspective,

    J. Zhang, H. Li, S. Sra, and A. Jadbabaie, “Neural network weights do not converge to stationary points: An invariant measure perspective,” inInternational Conference on Machine Learning. PMLR, 2022, pp. 26 330–26 346

    work page 2022

  49. [49]

    Stability conditions for nonlinear continuous neural networks with asymmetric connection weights,

    K. Matsuoka, “Stability conditions for nonlinear continuous neural networks with asymmetric connection weights,”Neural networks, vol. 5, no. 3, pp. 495–500, 1992

    work page 1992

  50. [50]

    A linear self- normalizing spike-timing-dependent plasticity rule for online learning in spiking neural networks,

    G. C. Fahey, S. J. Ippolito, and G. I. Matthews, “A linear self- normalizing spike-timing-dependent plasticity rule for online learning in spiking neural networks,”Neurocomputing, p. 133419, 2026

    work page 2026

  51. [51]

    Achieving stable dynamics in neural circuits,

    L. Kozachkov, M. Lundqvist, J.-J. Slotine, and E. K. Miller, “Achieving stable dynamics in neural circuits,”PLoS computational biology, vol. 16, no. 8, p. e1007659, 2020

    work page 2020