pith. sign in

arxiv: 2606.27529 · v1 · pith:VZD5MRTZnew · submitted 2026-06-25 · 🧬 q-bio.NC

Modelling chronic stress as an excitatory-inhibitory perturbation in recurrent working-memory networks

Mauricio A Diaz , Manuela A. Beyer , Janina Hesse This is my paper

Pith reviewed 2026-06-29 00:47 UTC · model grok-4.3

classification 🧬 q-bio.NC
keywords chronic stressexcitatory-inhibitory balanceworking memoryrecurrent networksinhibitory dominanceresiliencegeneralizationprefrontal cortex
0
0 comments X

The pith

Stronger inhibitory-to-excitatory synapses recover all three signatures of chronic stress in working-memory networks.

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

The paper tests eight possible ways chronic stress could alter excitatory-inhibitory balance inside recurrent networks that perform a working-memory task. Only one change—making inhibitory synapses onto excitatory neurons stronger—produces the three observed effects at once: inhibitory dominance, weaker excitatory drive, and poorer task performance. The same networks, when trained from the start under this stress rule, keep their performance and stay in the same dynamical regime whether stress is present or not. This resilience, however, comes with narrower generalization to memory demands outside the training range.

Core claim

Among eight candidate synaptic or activity modulations, only stronger inhibitory-to-excitatory synapses simultaneously produce inhibitory dominance, excitatory hypofunction, and impaired working-memory performance. Networks trained under this mechanism maintain performance and remain in the same dynamical subspace and energetic regime with or without stress, yet show reduced generalization when the task demands longer memory intervals than those seen during training; the resilience-generalization trade-off holds across stress levels and network sizes.

What carries the argument

Recurrent networks trained on a working-memory task, with eight candidate operators that modulate synaptic strength or neuronal activity to model chronic stress.

If this is right

  • Resilient networks preserve task performance under stress and stay within the same dynamical subspace and energetic regime.
  • Resilient networks generalize less well to working-memory tasks that require longer retention intervals than those used in training.
  • The resilience-generalization trade-off remains across different stress magnitudes and network sizes.
  • Resilience training produces a more specialized solution tuned to the trained regime.

Where Pith is reading between the lines

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

  • The model predicts that chronic stress biases circuits toward rigid, less adaptable solutions that resemble habit-like behavior.
  • Interventions that selectively weaken inhibitory-to-excitatory synapses could restore both performance and flexibility after stress exposure.
  • Similar stress operators could be tested in other recurrent circuits to see whether the same mechanism explains dysfunction outside working memory.
  • The observed trade-off suggests a computational reason why stressed animals show reduced behavioral flexibility on novel problems.

Load-bearing premise

The three experimental signatures plus the working-memory task are enough to single out the right mechanism among the eight candidates.

What would settle it

Finding that biological prefrontal circuits under chronic stress do not show strengthened inhibitory-to-excitatory synapses while still displaying the three signatures, or that another of the eight operators matches the signatures equally well under additional biological constraints.

Figures

Figures reproduced from arXiv: 2606.27529 by Janina Hesse, Manuela A. Beyer, Mauricio A Diaz.

Figure 2
Figure 2. Figure 2: Stress training preserves task performance under stress but introduces a delay-generalisation trade-off. (a) Psychometric accuracy as a function of stimulus evidence, |∆| = |S2 − S1|, for naïve (blue) and resilient (orange) networks. Left: without test-time stress, the two classes are nearly identical across evidence levels, confirming that stress training leaves baseline performance intact. Right: under S… view at source ↗
Figure 3
Figure 3. Figure 3: Stress-induced changes in task performance and network dynamics. Each cell shows the signed percentage change relative to the unperturbed baseline (δ = 0, σstress = 0; Eq. 11), with diagonally split cells: upper-left triangle at δT = 0.5 (within training range) and lower-right triangle at δmax = 1.0 (beyond training range). Rows are grouped into biologically targeted protocols and population-agnostic contr… view at source ↗
Figure 4
Figure 4. Figure 4: Stress resilience produces a context-dependent delay-generalisation trade-off. The figure is a 2 × 2 array of blocks. Top blocks (a, b) report raw accuracy; bottom blocks (c, d) report the naïve-minus-resilient difference ∆a = aNaive − aResilient. Left blocks (a, c) use within-distribution delays (400–900 ms, the trained range); right blocks (b, d) use long out-of-distribution delays (OOD-high, > 900 ms). … view at source ↗
Figure 5
Figure 5. Figure 5: Stress training preserves recurrent population geometry and constrains energetic collapse under perturbation. Naïve networks (blue) and stress-trained resilient networks (orange) were analysed under inhibitory￾to-excitatory weight perturbations, S ↑ [WI→E], of strength δ. (a) Representative PCA projections of recurrent rate trajectories for a matched seed pair, shown in the top three baseline PCs. Solid li… view at source ↗
Figure 6
Figure 6. Figure 6: Stress resilience is associated with a reorganization of recurrent network topology. Density (a) and weighted directed reciprocity (b) for trained recurrent connectivity matrices. Each point represents one trained network (n = 200 networks per condition). Black circles indicate group means ± SEM. Statistical comparisons use two-sided Mann–Whitney U tests, with *** indicating p < 0.001. 10 [PITH_FULL_IMAGE… view at source ↗
read the original abstract

Stress is an adaptive response coordinated by neural and physiological systems. While acute stress can enhance survival, chronic stress drives structural brain changes, cognitive dysfunction, and increased psychiatric risk. At the cellular level, chronic stress shifts the excitatory-inhibitory (E/I) balance of prefrontal pyramidal neurons toward inhibitory dominance, yet the mechanisms underlying these alterations are still unknown. We here investigate possible mechanisms causing inhibitory dominance using recurrent neuronal networks trained on a working memory task. Chronic stress is modelled as a modulation in synaptic strength or neuronal activity, systematically comparing eight candidate operators against three experimentally motivated signatures of stress-induced prefrontal dysfunction: inhibitory dominance, excitatory hypofunction, and impaired task performance. These signatures are all recovered by a single stress mechanism, stronger inhibitory-to-excitatory synapses. Contrasting naive networks with resilient networks trained under the stress mechanism, we find that resilience training not only preserves task performance under stress, but also confines the network to the same dynamical subspace and energetic regime with and without stress. This resilience comes at a cost: resilient networks generalise less well when the task requires longer memory than seen during training, indicating that resilient networks find a specialised solution tuned to the trained regime. This trade-off between resilience and generalization performance persists across stress magnitude and network size, offering a computational analogue of the shift toward rigid, habit-like behaviour reported in animal following chronic stress.

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

1 major / 2 minor

Summary. The paper models chronic stress as one of eight candidate E/I perturbations (synaptic strength or activity modulations) in recurrent networks trained on a working-memory task. It reports that only stronger inhibitory-to-excitatory synapses simultaneously reproduce the three experimental signatures of inhibitory dominance, excitatory hypofunction, and impaired task performance. Networks trained under this mechanism are then shown to be resilient (preserving performance, dynamical subspace, and energetic regime under stress) while exhibiting reduced generalization to longer memory delays, with the resilience-generalization trade-off persisting across stress magnitudes and network sizes.

Significance. If the central identification holds, the work supplies a mechanistic account linking a specific synaptic operator to multiple stress-induced prefrontal phenotypes and supplies a computational analogue for the resilience-rigidity trade-off observed in stressed animals. The systematic enumeration of eight operators and the subsequent analysis of subspace confinement and generalization cost are positive features that go beyond single-mechanism fitting.

major comments (1)
  1. [Results, operator comparison] Results (mechanism comparison): the uniqueness claim—that only the I-to-E strengthening operator recovers all three signatures while the other seven do not—rests on the assumption that the three signatures supply independent constraints. Because inhibitory dominance and excitatory hypofunction are both direct, correlated consequences of any E/I shift in the same recurrent circuit, and task impairment follows from the resulting dynamics, the reported analysis does not demonstrate that the signatures are sufficiently orthogonal to exclude alternative operators that could be tuned to match the same three observables. A quantitative measure of signature independence or an additional, biologically motivated signature would be required to support the uniqueness conclusion.
minor comments (2)
  1. [Abstract / Methods] The abstract and methods should explicitly state the precise definitions of the eight operators (e.g., which synapses are scaled and by what functional form) so that the comparison can be reproduced without ambiguity.
  2. [Figures] Figure legends for the resilience and generalization panels should report the exact number of networks, random seeds, and statistical tests used to support the claim that the trade-off 'persists across stress magnitude and network size.'

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the strength of our uniqueness claim. We address the major comment below.

read point-by-point responses

  1. Referee: Results (mechanism comparison): the uniqueness claim—that only the I-to-E strengthening operator recovers all three signatures while the other seven do not—rests on the assumption that the three signatures supply independent constraints. Because inhibitory dominance and excitatory hypofunction are both direct, correlated consequences of any E/I shift in the same recurrent circuit, and task impairment follows from the resulting dynamics, the reported analysis does not demonstrate that the signatures are sufficiently orthogonal to exclude alternative operators that could be tuned to match the same three observables. A quantitative measure of signature independence or an additional, biologically motivated signature would be required to support the uniqueness conclusion.

    Authors: We agree that inhibitory dominance and excitatory hypofunction are mechanistically linked through E/I balance and that task impairment is a downstream consequence. However, the eight operators represent distinct biological perturbations (specific synaptic weight changes vs. activity modulations), and our systematic parameter sweeps show that only I-to-E strengthening simultaneously matches the quantitative experimental signatures (direction and magnitude of E/I shift plus performance drop) reported in the stress literature. Other operators either produce mismatched E/I ratios, fail to impair performance at observed levels, or require implausible parameter values outside biological ranges. While the signatures are not fully orthogonal, their combination still discriminates among the operators in our enumeration. We will add a supplementary figure quantifying pairwise correlations among the three signatures across all operators and a discussion paragraph addressing the referee's concern about independence. This constitutes a partial revision. revision: partial

Circularity Check

0 steps flagged

No significant circularity; mechanism identification uses independent experimental signatures

full rationale

The paper compares eight candidate operators against three experimentally motivated signatures (inhibitory dominance, excitatory hypofunction, impaired task performance) drawn from external literature. The identification of stronger inhibitory-to-excitatory synapses as the sole matching mechanism does not reduce to a self-definitional loop, a fitted parameter renamed as prediction, or a self-citation chain. Resilience and generalization results are obtained by direct simulation of networks trained under the identified mechanism; they are consequences rather than inputs. No equations or steps in the provided derivation exhibit the enumerated circularity patterns. The analysis remains self-contained against the external signatures.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard domain assumptions in computational neuroscience about RNNs modeling prefrontal function and on the selection of eight candidate operators, some of which likely involve implicit scaling parameters chosen to match signatures.

free parameters (1)
  • perturbation strength
    The magnitude of synaptic or activity modulation used to model chronic stress, selected to reproduce the three signatures.
axioms (1)
  • domain assumption Recurrent networks trained on a working memory task adequately capture prefrontal E/I balance relevant to stress-induced dysfunction.
    Invoked to evaluate task performance and dynamical subspace confinement.

pith-pipeline@v0.9.1-grok · 5777 in / 1314 out tokens · 49096 ms · 2026-06-29T00:47:28.144890+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

58 extracted references · 52 canonical work pages

  1. [1]

    A comprehensive overview on stress neurobiology: Basic concepts and clinical implications

    Lívea Dornela Godoy, Matheus Teixeira Rossignoli, Polianna Delfino-Pereira, Norberto Garcia-Cairasco, and Eduardo Henrique de Lima Umeoka. A comprehensive overview on stress neurobiology: Basic concepts and clinical implications. 12:127. doi:10.3389/fnbeh.2018.00127. URL http://dx.doi.org/10.3389/fnbeh. 2018.00127

    work page doi:10.3389/fnbeh.2018.00127 2018

  2. [2]

    Understand- ing the relationships between physiological and psychosocial stress, cortisol and cognition

    Katharine Ann James, Juliet Ilena Stromin, Nina Steenkamp, and Marc Irwin Combrinck. Understand- ing the relationships between physiological and psychosocial stress, cortisol and cognition. 14:1085950. doi:10.3389/fendo.2023.1085950. URLhttp://dx.doi.org/10.3389/fendo.2023.1085950. 18 Chronic stress as an E/I perturbation in RNNsA PREPRINT

    work page doi:10.3389/fendo.2023.1085950 2023

  3. [3]

    Stress signalling pathways that impair prefrontal cortex structure and function

    Amy F T Arnsten. Stress signalling pathways that impair prefrontal cortex structure and function. 10:410–422. doi:10.1038/nrn2648. URLhttp://dx.doi.org/10.1038/nrn2648

    work page doi:10.1038/nrn2648

  4. [4]

    Neurocognitive effects of stress: a metaparadigm perspective

    Eun Joo Kim and Jeansok J Kim. Neurocognitive effects of stress: a metaparadigm perspective. 28:2750–2763. doi:10.1038/s41380-023-01986-4. URLhttp://dx.doi.org/10.1038/s41380-023-01986-4

    work page doi:10.1038/s41380-023-01986-4

  5. [5]

    Stress-induced impairment of a working memory task: role of spiking rate and spiking history predicted discharge

    David M Devilbiss, Rick L Jenison, and Craig W Berridge. Stress-induced impairment of a working memory task: role of spiking rate and spiking history predicted discharge. 8:e1002681. doi:10.1371/journal.pcbi.1002681. URL http://dx.doi.org/10.1371/journal.pcbi.1002681

    work page doi:10.1371/journal.pcbi.1002681

  6. [6]

    Chronic stress causes frontostriatal reorganization and affects decision-making

    Eduardo Dias-Ferreira, João C Sousa, Irene Melo, Pedro Morgado, Ana R Mesquita, João J Cerqueira, Rui M Costa, and Nuno Sousa. Chronic stress causes frontostriatal reorganization and affects decision-making. 325: 621–625. doi:10.1126/science.1171203. URLhttp://dx.doi.org/10.1126/science.1171203

    work page doi:10.1126/science.1171203

  7. [7]

    Effects of chronic stress on cognitive function - from neurobiology to intervention

    Milena Girotti, Sarah E Bulin, and Flavia R Carreno. Effects of chronic stress on cognitive function - from neurobiology to intervention. 33:100670. doi:10.1016/j.ynstr.2024.100670. URL http://dx.doi.org/10. 1016/j.ynstr.2024.100670

    work page doi:10.1016/j.ynstr.2024.100670 2024

  8. [8]

    Stress effects on neuronal structure: Hippocampus, amygdala, and prefrontal cortex

    Bruce S McEwen, Carla Nasca, and Jason D Gray. Stress effects on neuronal structure: Hippocampus, amygdala, and prefrontal cortex. 41:3–23. doi:10.1038/npp.2015.171. URL http://dx.doi.org/10.1038/npp.2015. 171

    work page doi:10.1038/npp.2015.171 2015

  9. [9]

    Neurobiology of chronic stress-related psychiatric disorders: Evidence from molecular imaging studies

    Margaret T Davis, Sophie E Holmes, Robert H Pietrzak, and Irina Esterlis. Neurobiology of chronic stress-related psychiatric disorders: Evidence from molecular imaging studies. 1:247054701771091. doi:10.1177/2470547017710916. URLhttp://dx.doi.org/10.1177/2470547017710916

    work page doi:10.1177/2470547017710916

  10. [10]

    The evolving neurobiology of early-life stress

    Matthew T Birnie and Tallie Z Baram. The evolving neurobiology of early-life stress. 113:1474–1490. doi:10.1016/j.neuron.2025.02.016. URLhttp://dx.doi.org/10.1016/j.neuron.2025.02.016

    work page doi:10.1016/j.neuron.2025.02.016 2025

  11. [11]

    Influence of glutamate and GABA transport on brain excita- tory/inhibitory balance

    Sheila Ms Sears and Sandra J Hewett. Influence of glutamate and GABA transport on brain excita- tory/inhibitory balance. 246:1069–1083. doi:10.1177/1535370221989263. URL http://dx.doi.org/10. 1177/1535370221989263

    work page doi:10.1177/1535370221989263

  12. [12]

    Keeping excitation-inhibition ratio in balance

    Sergei Kirischuk. Keeping excitation-inhibition ratio in balance. 23:5746. doi:10.3390/ijms23105746. URL http://dx.doi.org/10.3390/ijms23105746

    work page doi:10.3390/ijms23105746

  13. [13]

    Brain-wide changes in excitation-inhibition balance of major depressive disorder: a systematic review of topographic patterns of GABA- and glutamater- gic alterations

    Yu-Ting Hu, Zhong-Lin Tan, Dusan Hirjak, and Georg Northoff. Brain-wide changes in excitation-inhibition balance of major depressive disorder: a systematic review of topographic patterns of GABA- and glutamater- gic alterations. 28:3257–3266. doi:10.1038/s41380-023-02193-x. URL http://dx.doi.org/10.1038/ s41380-023-02193-x

    work page doi:10.1038/s41380-023-02193-x

  14. [14]

    Prefrontal excitatory/inhibitory balance in stress and emotional disorders: Evidence for over-inhibition

    Chloe E Page and Laurence Coutellier. Prefrontal excitatory/inhibitory balance in stress and emotional disorders: Evidence for over-inhibition. 105:39–51. doi:10.1016/j.neubiorev.2019.07.024. URL http://dx.doi.org/10. 1016/j.neubiorev.2019.07.024

    work page doi:10.1016/j.neubiorev.2019.07.024 2019

  15. [15]

    Chronic stress increases prefrontal inhibition: A mechanism for stress-induced prefrontal dysfunction

    Jessica M McKlveen, Rachel L Morano, Maureen Fitzgerald, Sandra Zoubovsky, Sarah N Cassella, Jessie R Scheimann, Sriparna Ghosal, Parinaz Mahbod, Benjamin A Packard, Brent Myers, Mark L Baccei, and James P Herman. Chronic stress increases prefrontal inhibition: A mechanism for stress-induced prefrontal dysfunction. 80:754–764. doi:10.1016/j.biopsych.2016....

    work page doi:10.1016/j.biopsych.2016.03.2101 2016

  16. [16]

    Chronic stress alters synaptic in- hibition/excitation balance of pyramidal neurons but not PV interneurons in the infralimbic and prelim- bic cortices of C57BL/6J mice

    Diana Rodrigues, Cátia Santa, Bruno Manadas, and Patrícia Monteiro. Chronic stress alters synaptic in- hibition/excitation balance of pyramidal neurons but not PV interneurons in the infralimbic and prelim- bic cortices of C57BL/6J mice. 11:ENEURO.0053–24.2024. doi:10.1523/ENEURO.0053-24.2024. URL http://dx.doi.org/10.1523/ENEURO.0053-24.2024

    work page doi:10.1523/eneuro.0053-24.2024 2024

  17. [17]

    Repeated stress causes cognitive impairment by suppressing glutamate receptor expression and function in prefrontal cortex

    Eunice Y Yuen, Jing Wei, Wenhua Liu, Ping Zhong, Xiangning Li, and Zhen Yan. Repeated stress causes cognitive impairment by suppressing glutamate receptor expression and function in prefrontal cortex. 73:962–977. doi:10.1016/j.neuron.2011.12.033. URLhttp://dx.doi.org/10.1016/j.neuron.2011.12.033

    work page doi:10.1016/j.neuron.2011.12.033 2011

  18. [18]

    Circuit- and laminar-specific regulation of medial prefrontal neurons by chronic stress

    Wei-Zhu Liu, Chun-Yan Wang, Yu Wang, Mei-Ting Cai, Wei-Xiang Zhong, Tian Liu, Zhi-Hao Wang, Han- Qing Pan, Wen-Hua Zhang, and Bing-Xing Pan. Circuit- and laminar-specific regulation of medial prefrontal neurons by chronic stress. 13:90. doi:10.1186/s13578-023-01050-2. URL http://dx.doi.org/10.1186/ s13578-023-01050-2

    work page doi:10.1186/s13578-023-01050-2

  19. [19]

    Inhibition of prefrontal cortex parvalbumin interneurons mitigates behavioral and physiological sequelae of chronic stress in male mice

    Nawshaba Nawreen, Kristen Oshima, James Chambers, Marissa Smail, and James P Herman. Inhibition of prefrontal cortex parvalbumin interneurons mitigates behavioral and physiological sequelae of chronic stress in male mice. 27:2361238. doi:10.1080/10253890.2024.2361238. URL http://dx.doi.org/10.1080/ 10253890.2024.2361238. 19 Chronic stress as an E/I pertur...

    work page doi:10.1080/10253890.2024.2361238 2024

  20. [20]

    Effects of altered excitation-inhibition balance on decision making in a cortical circuit model

    Norman H Lam, Thiago Borduqui, Jaime Hallak, Antonio Roque, Alan Anticevic, John H Krystal, Xiao-Jing Wang, and John D Murray. Effects of altered excitation-inhibition balance on decision making in a cortical circuit model. 42:1035–1053. doi:10.1523/JNEUROSCI.1371-20.2021. URL http://dx.doi.org/10.1523/ JNEUROSCI.1371-20.2021

    work page doi:10.1523/jneurosci.1371-20.2021 2021

  21. [21]

    Choice selective inhibition drives stability and competition in decision circuits

    James P Roach, Anne K Churchland, and Tatiana A Engel. Choice selective inhibition drives stability and competition in decision circuits. 14:147. doi:10.1038/s41467-023-35822-8. URL http://dx.doi.org/10. 1038/s41467-023-35822-8

    work page doi:10.1038/s41467-023-35822-8

  22. [22]

    Training excitatory-inhibitory recurrent neural networks for cognitive tasks: A simple and flexible framework

    H Francis Song, Guangyu R Yang, and Xiao-Jing Wang. Training excitatory-inhibitory recurrent neural networks for cognitive tasks: A simple and flexible framework. 12:e1004792. doi:10.1371/journal.pcbi.1004792. URL http://dx.doi.org/10.1371/journal.pcbi.1004792

    work page doi:10.1371/journal.pcbi.1004792

  23. [23]

    Training dynamically balanced excitatory-inhibitory networks

    Alessandro Ingrosso and L F Abbott. Training dynamically balanced excitatory-inhibitory networks. 14:e0220547. doi:10.1371/journal.pone.0220547. URLhttp://dx.doi.org/10.1371/journal.pone.0220547

    work page doi:10.1371/journal.pone.0220547

  24. [25]

    Recurrent neural network models for working memory of continuous variables: activity manifolds, connectivity patterns, and dynamic codes

    Christopher J Cueva, Adel Ardalan, Misha Tsodyks, and Ning Qian. Recurrent neural network models for working memory of continuous variables: activity manifolds, connectivity patterns, and dynamic codes. doi:10.48550/arXiv.2111.01275. URLhttp://arxiv.org/abs/2111.01275

    work page doi:10.48550/arxiv.2111.01275

  25. [26]

    A recurrent neural network model of prefrontal brain activity during a working memory task

    Emilia P Piwek, Mark G Stokes, and Christopher Summerfield. A recurrent neural network model of prefrontal brain activity during a working memory task. 19:e1011555. doi:10.1371/journal.pcbi.1011555. URL http: //dx.doi.org/10.1371/journal.pcbi.1011555

    work page doi:10.1371/journal.pcbi.1011555

  26. [27]

    Synapses mediate the effects of different types of stress on working memory: a brain-inspired spiking neural network study

    Chengcheng Du, Yinqian Sun, Jihang Wang, Qian Zhang, and Yi Zeng. Synapses mediate the effects of different types of stress on working memory: a brain-inspired spiking neural network study. 19:1534839. doi:10.3389/fncel.2025.1534839. URLhttp://dx.doi.org/10.3389/fncel.2025.1534839

    work page doi:10.3389/fncel.2025.1534839 2025

  27. [28]

    Neuromodulators generate multiple context-relevant behaviors in recurrent neural networks

    Ben Tsuda, Stefan C Pate, Kay M Tye, Hava T Siegelmann, and Terrence J Sejnowski. Neuromodulators generate multiple context-relevant behaviors in recurrent neural networks. 38:292–327. doi:10.1162/NECO.a.1489. URL https://dx.doi.org/10.1162/NECO.a.1489

    work page doi:10.1162/neco.a.1489

  28. [29]

    Noisy recur- rent neural networks

    Soon Hoe Lim, N Benjamin Erichson, Liam Hodgkinson, and Michael W Mahoney. Noisy recur- rent neural networks. 34:5124–5137. URL https://proceedings.neurips.cc/paper/2021/hash/ 29301521774ff3cbd26652b2d5c95996-Abstract.html

    2021

  29. [30]

    Adversarial weight perturbation helps robust generalization

    Dongxian Wu, Shu-Tao Xia, and Yisen Wang. Adversarial weight perturbation helps robust generalization. doi:10.5555/3495724.3495973. URLhttp://dx.doi.org/10.5555/3495724.3495973

    work page doi:10.5555/3495724.3495973

  30. [31]

    Random noise promotes slow heterogeneous synaptic dynamics important for robust working memory computation

    Nuttida Rungratsameetaweemana, Robert Kim, Thiparat Chotibut, and Terrence J Sejnowski. Random noise promotes slow heterogeneous synaptic dynamics important for robust working memory computation. 122: e2316745122. doi:10.1073/pnas.2316745122. URLhttp://dx.doi.org/10.1073/pnas.2316745122

    work page doi:10.1073/pnas.2316745122

  31. [32]

    Effect in the spectra of eigenvalues and dynamics of RNNs trained with excitatory-inhibitory constraint

    Cecilia Jarne and Mariano Caruso. Effect in the spectra of eigenvalues and dynamics of RNNs trained with excitatory-inhibitory constraint. 18:1323–1335. doi:10.1007/s11571-023-09956-w. URL http://dx.doi.org/ 10.1007/s11571-023-09956-w

    work page doi:10.1007/s11571-023-09956-w

  32. [33]

    Functional implications of dale’s law in balanced neuronal network dynamics and decision making

    Victor J Barranca, Asha Bhuiyan, Max Sundgren, and Fangzhou Xing. Functional implications of dale’s law in balanced neuronal network dynamics and decision making. 16:801847. doi:10.3389/fnins.2022.801847. URL http://dx.doi.org/10.3389/fnins.2022.801847

    work page doi:10.3389/fnins.2022.801847 2022

  33. [34]

    Task representations in neural networks trained to perform many cognitive tasks

    Guangyu Robert Yang, Madhura R Joglekar, H Francis Song, William T Newsome, and Xiao-Jing Wang. Task representations in neural networks trained to perform many cognitive tasks. 22:297–306. doi:10.1038/s41593- 018-0310-2. URLhttp://dx.doi.org/10.1038/s41593-018-0310-2

    work page doi:10.1038/s41593-

  34. [35]

    Artificial neural networks for neuroscientists: A primer

    Guangyu Robert Yang and Xiao-Jing Wang. Artificial neural networks for neuroscientists: A primer. 107:1048–

  35. [36]

    URLhttp://dx.doi.org/10.1016/j.neuron.2020.09.005

    doi:10.1016/j.neuron.2020.09.005. URLhttp://dx.doi.org/10.1016/j.neuron.2020.09.005

    work page doi:10.1016/j.neuron.2020.09.005 2020

  36. [37]

    Synaptic basis of cortical persistent activity: the importance of NMDA receptors to working memory

    X J Wang. Synaptic basis of cortical persistent activity: the importance of NMDA receptors to working memory. 19:9587–9603. doi:10.1523/jneurosci.19-21-09587.1999. URL http://dx.doi.org/10.1523/JNEUROSCI. 19-21-09587.1999

    work page doi:10.1523/jneurosci.19-21-09587.1999 1999

  37. [38]

    Synaptic mechanisms and network dynamics underlying spatial working memory in a cortical network model

    A Compte, N Brunel, P S Goldman-Rakic, and X J Wang. Synaptic mechanisms and network dynamics underlying spatial working memory in a cortical network model. 10:910–923. doi:10.1093/cercor/10.9.910. URL http://dx.doi.org/10.1093/cercor/10.9.910. 20 Chronic stress as an E/I perturbation in RNNsA PREPRINT

    work page doi:10.1093/cercor/10.9.910

  38. [39]

    Russo, and Marianne B

    Raffael Kalisch, Scott J. Russo, and Marianne B. Müller. Neurobiology and systems biology of stress resilience. Physiological Reviews, 104(3):1205–1263, 2024. doi:10.1152/physrev.00042.2023

    work page doi:10.1152/physrev.00042.2023 2024

  39. [40]

    Neurobiological basis of stress resilience

    Eric J Nestler and Scott J Russo. Neurobiological basis of stress resilience. 112:1911–1929. doi:10.1016/j.neuron.2024.05.001. URLhttp://dx.doi.org/10.1016/j.neuron.2024.05.001

    work page doi:10.1016/j.neuron.2024.05.001 1911

  40. [41]

    Stimulus-driven and spontaneous dynamics in excitatory- inhibitory recurrent neural networks for sequence representation

    Alfred Rajakumar, John Rinzel, and Zhe S Chen. Stimulus-driven and spontaneous dynamics in excitatory- inhibitory recurrent neural networks for sequence representation. 33:2603–2645. doi:10.1162/neco_a_01418. URLhttp://dx.doi.org/10.1162/neco_a_01418

    work page doi:10.1162/neco_a_01418

  41. [42]

    Predictive coding is a consequence of energy efficiency in recurrent neural networks

    Abdullahi Ali, Nasir Ahmad, Elgar de Groot, Marcel Antonius Johannes van Gerven, and Tim Christian Ki- etzmann. Predictive coding is a consequence of energy efficiency in recurrent neural networks. 3:100639. doi:10.1016/j.patter.2022.100639. URLhttp://dx.doi.org/10.1016/j.patter.2022.100639

    work page doi:10.1016/j.patter.2022.100639 2022

  42. [43]

    Sussillo and O

    David Sussillo and Omri Barak. Opening the black box: low-dimensional dynamics in high-dimensional recurrent neural networks. 25:626–649. doi:10.1162/NECO_a_00409. URL http://dx.doi.org/10.1162/NECO_a_ 00409

    work page doi:10.1162/neco_a_00409

  43. [44]

    Dynamical system approach to explainability in recurrent neural networks

    Alexis Dubreuil. Dynamical system approach to explainability in recurrent neural networks. URL https://www. semanticscholar.org/paper/Dynamical-system-approach-to-explainability-in-Dubreuil/ d08515aefa1a9df330a33106e788e21daa14c135

  44. [45]

    Considerations in using recurrent neural networks to probe neural dynamics

    Jonathan C Kao. Considerations in using recurrent neural networks to probe neural dynamics. 122:2504–2521. doi:10.1152/jn.00467.2018. URLhttp://dx.doi.org/10.1152/jn.00467.2018

    work page doi:10.1152/jn.00467.2018 2018

  45. [46]

    A neural network walks into a lab: towards using deep nets as models for human behavior

    Wei Ji Ma and Benjamin Peters. A neural network walks into a lab: towards using deep nets as models for human behavior. doi:10.48550/arXiv.2005.02181. URLhttp://arxiv.org/abs/2005.02181

    work page doi:10.48550/arxiv.2005.02181 2005

  46. [47]

    General principles of neuronal co-transmission: Insights from multiple model systems

    Erik Svensson, John Apergis-Schoute, Geoffrey Burnstock, Michael P Nusbaum, David Parker, and Helgi B Schiöth. General principles of neuronal co-transmission: Insights from multiple model systems. 12:117. doi:10.3389/fncir.2018.00117. URLhttp://dx.doi.org/10.3389/fncir.2018.00117

    work page doi:10.3389/fncir.2018.00117 2018

  47. [48]

    The neuro-symphony of stress

    Marian Joëls and Tallie Z Baram. The neuro-symphony of stress. 10:459–466. doi:10.1038/nrn2632. URL http://dx.doi.org/10.1038/nrn2632

    work page doi:10.1038/nrn2632

  48. [49]

    The neocortical circuit: themes and variations

    Kenneth D Harris and Gordon M G Shepherd. The neocortical circuit: themes and variations. 18:170–181. doi:10.1038/nn.3917. URLhttp://dx.doi.org/10.1038/nn.3917

    work page doi:10.1038/nn.3917

  49. [50]

    Dale’s principle

    P Strata and R Harvey. Dale’s principle. 50:349–350. doi:10.1016/s0361-9230(99)00100-8. URL http: //dx.doi.org/10.1016/s0361-9230(99)00100-8

    work page doi:10.1016/s0361-9230(99)00100-8

  50. [51]

    Neuronal correlates of parametric working memory in the prefrontal cortex

    R Romo, C D Brody, A Hernández, and L Lemus. Neuronal correlates of parametric working memory in the prefrontal cortex. 399:470–473. doi:10.1038/20939. URLhttp://dx.doi.org/10.1038/20939

    work page doi:10.1038/20939

  51. [52]

    Neuronal population coding of parametric working memory

    Omri Barak, Misha Tsodyks, and Ranulfo Romo. Neuronal population coding of parametric working memory. 30:9424–9430. doi:10.1523/JNEUROSCI.1875-10.2010. URL http://dx.doi.org/10.1523/JNEUROSCI. 1875-10.2010

    work page doi:10.1523/jneurosci.1875-10.2010 2010

  52. [53]

    JAX: composable transformations of Python+NumPy programs

    James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Yash Katariya, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and Qiao Zhang. JAX: composable transformations of Python+NumPy programs. URLhttp://github.com/jax-ml/jax

  53. [54]

    Array programming with NumPy,

    Charles R Harris, K Jarrod Millman, Stéfan J van der Walt, Ralf Gommers, Pauli Virtanen, David Cournapeau, Eric Wieser, Julian Taylor, Sebastian Berg, Nathaniel J Smith, Robert Kern, Matti Picus, Stephan Hoyer, Marten H van Kerkwijk, Matthew Brett, Allan Haldane, Jaime Fernández del Río, Mark Wiebe, Pearu Peterson, Pierre Gérard-Marchant, Kevin Sheppard, ...

    work page doi:10.1038/s41586-020-2649-2

  54. [55]

    SciPy1.0: FundamentalalgorithmsforscientificcomputinginPython,

    Pauli Virtanen, Ralf Gommers, Travis E Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, Stéfan J van der Walt, Matthew Brett, Joshua Wilson, K Jarrod Millman, Nikolay Mayorov, Andrew R J Nelson, Eric Jones, Robert Kern, Eric Larson, C J Carey, ˙Ilhan Polat, Yu Feng, Eric W Moore, ...

    work page doi:10.1038/s41592-019-0686-2

  55. [56]

    Data Structures for Statistical Computing in Python

    Wes McKinney. Data structures for statistical computing in python. In Stéfan van der Walt and Jarrod Millman, editors,Proceedings of the 9th Python in Science Conference, pages 56–61. doi:10.25080/Majora-92bf1922-00a. URLhttp://dx.doi.org/10.25080/Majora-92bf1922-00a. 21 Chronic stress as an E/I perturbation in RNNsA PREPRINT

    work page doi:10.25080/majora-92bf1922-00a

  56. [57]

    Scikit-learn: Machine learning in python

    F Pedregosa, G Varoquaux, A Gramfort, V Michel, B Thirion, O Grisel, M Blondel, P Prettenhofer, R Weiss, V Dubourg, J Vanderplas, A Passos, D Cournapeau, M Brucher, M Perrot, and E Duchesnay. Scikit-learn: Machine learning in python. 12:2825–2830

  57. [58]

    statsmodels: Econometric and statistical modeling with python

    Skipper Seabold and Josef Perktold. statsmodels: Econometric and statistical modeling with python. In9th Python in Science Conference

  58. [59]

    Matplotlib: A 2D Graphics Environment

    J D Hunter. Matplotlib: A 2D graphics environment. 9:90–95. doi:10.1109/MCSE.2007.55. URL http: //dx.doi.org/10.1109/MCSE.2007.55. 22

    work page doi:10.1109/mcse.2007.55 2007