Modeling of Self-sustained Neuron Population without External Stimulus

\.Ihsan Ertu\u{g}rul Karaka\c{s}; \.Ilkay Ulusoy; Orhan Murat Ko\c{c}ak; \"Ozden \"Ozel

arxiv: 2604.13719 · v1 · submitted 2026-04-15 · 💻 cs.NE · q-bio.NC

Modeling of Self-sustained Neuron Population without External Stimulus

\.Ihsan Ertu\u{g}rul Karaka\c{s} , \"Ozden \"Ozel , \.Ilkay Ulusoy , Orhan Murat Ko\c{c}ak This is my paper

Pith reviewed 2026-05-10 12:04 UTC · model grok-4.3

classification 💻 cs.NE q-bio.NC

keywords self-sustained neural activityHodgkin-Huxley neuronsspike-timing-dependent plasticityrecurrent networkssparse firingautonomous activitystochastic synapses

0 comments

The pith

Recurrent Hodgkin-Huxley networks with plastic and stochastic synapses sustain long-duration sparse autonomous activity after brief initialization.

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

The paper tests whether biophysically detailed networks can generate ongoing neural activity without continuous external drive. It simulates 200 Hodgkin-Huxley neurons with 80 percent connection probability, excitatory and inhibitory spike-timing-dependent plasticity, probabilistic vesicle release, and receptor variability. After a 200-millisecond stimulus to 30 excitatory cells, the network receives no further input yet maintains irregular firing for up to 1800 seconds. Most neurons fire below one spike per second on average, the population rate stays near 1 hertz, and spike timing remains irregular as measured by Fano factors near 1-2. Spontaneous shifts in collective firing patterns also appear over long timescales.

Core claim

A recurrent network of 200 Hodgkin-Huxley neurons that incorporates excitatory and inhibitory STDP together with probabilistic vesicle release and other stochastic synaptic features can maintain autonomous, sparse, irregular activity for thousands of seconds following only a brief 200 ms transient stimulus to a subset of excitatory cells, with no ongoing external input.

What carries the argument

Recurrent connections equipped with spike-timing-dependent plasticity and intrinsic stochasticity in vesicle release and synapse formation, which allow internal dynamics to sustain activity after initialization.

If this is right

Population-mean firing rates remain low and participation increases over longer observation windows.
Fano factors near 1-2 indicate irregular spike timing consistent with observed cortical activity.
Qualitative reorganizations in collective firing patterns emerge spontaneously without external drive.
Sparse firing regimes can persist for at least 1800 seconds in these recurrent networks.

Where Pith is reading between the lines

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

Biological circuits may rely on internal recurrent mechanisms rather than constant external input to maintain baseline activity.
Disruption of the stochastic or plastic components could shift the network into silence or into pathological high-rate states.
Similar self-sustaining dynamics might support resting-state fluctuations observed in larger brain areas.

Load-bearing premise

The specific parameter choices for connection probability, vesicle release probabilities, and STDP rules lie inside biologically plausible ranges that permit self-sustained activity.

What would settle it

A simulation run with identical network size and initialization but with STDP or vesicle-release stochasticity disabled shows that population activity decays to silence within minutes after the brief stimulus.

read the original abstract

Self-sustained neural activity in the absence of ongoing external input is a fundamental feature of nervous system dynamics, yet the conditions under which it can emerge in biophysically grounded network models remain incompletely understood. We studied whether a recurrent network of Hodgkin-Huxley neurons with spike-timing-dependent plasticity and intrinsic stochasticity can maintain autonomous activity after brief transient stimulation. The simulated network comprised 200 neurons (160 excitatory, 40 inhibitory) with 80% connection probability, incorporating excitatory and inhibitory STDP, probabilistic vesicle release, probabilistic synapse formation, receptor variability, and voltage-dependent inhibition. After a brief 200 ms initialization stimulus to 30 excitatory neurons, the network received no further external input. In one 1800 s simulation and two additional 500 s simulations, the network maintained sparse, irregular activity without ongoing drive. In the 1800 s run, 67% of neurons exhibited mean firing rates below 1 Hz, the population mean firing rate was 1.13 +/- 1.34 Hz, participation increased across longer observation windows, and population-mean Fano factors remained near 1-2, consistent with irregular spike timing. Raster activity also showed spontaneous qualitative reorganizations in collective firing patterns over time. These findings suggest that recurrent Hodgkin-Huxley networks with plastic and stochastic synapses can sustain long-duration autonomous activity in a sparse firing regime after brief initialization.

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.

Desk Editor's Note private letter to a colleague

This shows a recurrent HH network with STDP and stochastic synapses sustaining sparse irregular activity after a short stimulus, but the 80% connection density is likely carrying the result.

read the letter

The main thing to know is that this paper simulates a recurrent network of Hodgkin-Huxley neurons with STDP and stochastic synapses that keeps firing sparsely and irregularly for up to half an hour after a brief initial stimulus. The catch is the 80% connection probability, which is much denser than real cortex. They ran one long 1800-second simulation and two shorter ones, showing mean rates around 1 Hz, most neurons below 1 Hz, Fano factors of 1-2 indicating irregularity, and increasing participation over longer windows. The raster plots apparently show spontaneous shifts in collective patterns. That's useful for seeing how such a model behaves over extended periods. The soft spot is the connectivity. At 80%, it's easy for activity to persist via many overlapping paths even with probabilistic release. Typical cortical connection probs are 1-10%, and the paper doesn't report any runs at lower densities or mean-field adjustments to keep total input the same. If the activity vanishes there, this is more of an existence proof under dense wiring than a general mechanism. The abstract also leaves methods at a high level, so without the full parameter table or code it's hard to replicate exactly. This isn't breaking new ground conceptually—recurrent networks with plasticity sustaining activity is known—but the specific HH implementation with added stochasticity adds a concrete example. The authors seem straightforward about what they did. It's aimed at people building detailed biophysical models of spontaneous activity. A reader in that area could get ideas from the setup, but would probably want more controls before building on it. I think it deserves peer review. Referees could push for connectivity ablations and better documentation of the stochastic parts.

Referee Report

2 major / 2 minor

Summary. The manuscript reports that a recurrent network of 200 Hodgkin-Huxley neurons (160 excitatory, 40 inhibitory) with 80% connection probability, incorporating excitatory and inhibitory STDP, probabilistic vesicle release, probabilistic synapse formation, receptor variability, and voltage-dependent inhibition, can sustain sparse irregular autonomous activity for up to 1800 s after a brief 200 ms initialization stimulus to 30 excitatory neurons with no further external input. This is shown in three simulations (one 1800 s and two 500 s) with population mean firing rate 1.13 +/- 1.34 Hz, 67% of neurons below 1 Hz, Fano factors near 1-2, increasing participation over longer windows, and spontaneous reorganizations in collective firing patterns.

Significance. If the result holds under the stated conditions, it provides evidence that biophysically detailed recurrent HH networks with plasticity and multiple stochastic mechanisms can exhibit long-duration self-sustained activity in a sparse regime following transient initialization. This is relevant to models of cortical dynamics without ongoing drive. Strengths include the use of full Hodgkin-Huxley dynamics, multiple long-duration runs, and quantitative measures of irregularity and participation. However, the high connection density and high-level parameter description limit immediate reproducibility and generalization to biological regimes.

major comments (2)

Abstract and Network Setup: the network is constructed with 80% connection probability. Typical biological cortical connectivity is estimated at 1-10%. No ablation studies, parameter sweeps at lower densities, or controls that preserve mean synaptic input while reducing density are reported. If sustained activity requires this atypically dense wiring to maintain recurrent excitation despite probabilistic release and low firing rates, the claim that such networks 'can sustain' long-duration autonomous activity does not establish a general property of recurrent HH networks with STDP and stochastic synapses.
Methods and Simulation Results: the abstract and results describe the model components at a high level without providing complete parameter tables (e.g., exact STDP time constants and learning rates, vesicle release probabilities, receptor variability distributions, or initial conditions). The three reported runs cannot be independently reproduced or tested for sensitivity to these choices, undermining assessment of whether the observed sparse activity is robust or an artifact of specific unstated implementation details.

minor comments (2)

Results: the statement that 'participation increased across longer observation windows' is qualitative; adding explicit metrics (e.g., fraction of active neurons per time bin or cumulative participation curves) with error bars would strengthen the claim.
Abstract: the Fano factor is described as 'near 1-2'; reporting the exact mean and range across neurons or time windows, or providing a supplementary figure, would improve precision and allow direct comparison to experimental irregularity measures.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback on our manuscript. We address each major comment below, indicating where we agree and what revisions we will make to strengthen the work.

read point-by-point responses

Referee: Abstract and Network Setup: the network is constructed with 80% connection probability. Typical biological cortical connectivity is estimated at 1-10%. No ablation studies, parameter sweeps at lower densities, or controls that preserve mean synaptic input while reducing density are reported. If sustained activity requires this atypically dense wiring to maintain recurrent excitation despite probabilistic release and low firing rates, the claim that such networks 'can sustain' long-duration autonomous activity does not establish a general property of recurrent HH networks with STDP and stochastic synapses.

Authors: We agree that 80% connectivity exceeds typical biological estimates and that this choice limits direct generalization. The dense wiring was selected to ensure sufficient recurrent drive given the combination of probabilistic vesicle release, low mean firing rates, and multiple sources of stochasticity, allowing us to test whether self-sustained sparse activity remains possible. The manuscript claims only that such networks 'can sustain' activity under the stated conditions, not that the result is a general property of all recurrent HH networks. We did not perform density sweeps because the study focused on demonstrating the phenomenon with the full set of biophysical and plasticity mechanisms rather than exhaustive parameter exploration. In the revised manuscript we will add an explicit discussion of this limitation, clarify the scope of the claim, and note that exploring sparser regimes while preserving mean input is an important direction for future work. revision: partial
Referee: Methods and Simulation Results: the abstract and results describe the model components at a high level without providing complete parameter tables (e.g., exact STDP time constants and learning rates, vesicle release probabilities, receptor variability distributions, or initial conditions). The three reported runs cannot be independently reproduced or tested for sensitivity to these choices, undermining assessment of whether the observed sparse activity is robust or an artifact of specific unstated implementation details.

Authors: We acknowledge that a high-level description alone is insufficient for full reproducibility. All numerical parameters (STDP time constants and rates, release probabilities, receptor variability distributions, initial conditions, and integration settings) are defined in the Methods section and were used consistently across the reported simulations. To improve accessibility we will insert a dedicated parameter table in the revised Methods section that lists every value together with its source or justification. This addition will enable independent reproduction and sensitivity analyses by other groups. revision: yes

Circularity Check

0 steps flagged

No circularity: direct simulation of network dynamics

full rationale

The paper presents results from numerical integration of a recurrent Hodgkin-Huxley network model incorporating STDP, probabilistic vesicle release, and other stochastic elements. Activity persistence is observed after a brief 200 ms initialization stimulus in explicit 1800 s and 500 s runs, with reported statistics (firing rates, Fano factors) emerging from the simulation itself rather than any algebraic reduction or parameter fit. No derivation chain exists that equates a claimed prediction to its inputs by construction, no self-citation is invoked as a uniqueness theorem, and no ansatz or fitted quantity is relabeled as an independent result. The model parameters (80 % connectivity, STDP rules, etc.) are stated as modeling choices; the outcome is an existence demonstration under those choices, not a tautological re-expression of them.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The model rests on the standard Hodgkin-Huxley equations, published STDP rules, and probabilistic release mechanisms taken from prior literature. No new entities are postulated. Many numerical choices (neuron count, connection density, specific STDP time constants, vesicle probabilities) function as free parameters that were selected to produce the reported behavior.

free parameters (3)

connection probability
Set to 80% without derivation from data; directly controls network sparsity and recurrence strength.
STDP parameters
Excitatory and inhibitory STDP rules and time constants chosen from literature but not re-derived; affect plasticity dynamics.
vesicle release probability
Probabilistic release and receptor variability introduced as stochastic elements; values not specified in abstract.

axioms (2)

standard math Hodgkin-Huxley equations govern single-neuron dynamics
Invoked as the biophysical model for each of the 200 neurons.
domain assumption STDP rules from prior literature apply to both excitatory and inhibitory synapses
Used to implement plasticity without re-derivation.

pith-pipeline@v0.9.0 · 5579 in / 1552 out tokens · 33690 ms · 2026-05-10T12:04:57.180888+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1]

Each neuron was represented by two functional compartments: a soma and an axonal propagation component

Neuron Model The neuronal model employed in this study is based on the Hodgkin-Huxley (HH) formalism. Each neuron was represented by two functional compartments: a soma and an axonal propagation component. The soma represents the main neuronal body in which membrane potential dynamics are generated, whereas the axonal component propagates somatic voltage ...

work page
[2]

Synapse Model Synapses between neurons facilitate signal transmission through the release of excitatory and inhibitory neurotransmitter vesicles from presynaptic to postsynaptic neurons. When certain events happen, a presynaptic neuron can release neurotransmitters, which bind to receptors and affect the conductance of ionic channels in the postsynaptic n...

work page
[3]

We used a learning window of 50ms, meaning learning only occurs for presynaptic and postsynaptic action potentials that happen within 50ms

Spike Timing Dependent Plasticity Our model implements biologically inspired plasticity, called spike timing dependent plasticity, that allows synaptic weights, in our case number of receptors, to be modified based on the relative timing between presynaptic and postsynaptic neurons of a synapse. We used a learning window of 50ms, meaning learning only occ...

work page 1998
[4]

Each neuron had an 80% chance to create a connection with any other neuron, creating around 32000 total synapses

Network Architecture and Simulation The network consisted of 200 neurons, 40 inhibitory and 160 excitatory. Each neuron had an 80% chance to create a connection with any other neuron, creating around 32000 total synapses. Receptor counts were initialized using Gaussian distributions, with AMPA having 120 mean and 12 variance, and GABA having 200 mean and ...

work page arXiv 2021

[1] [1]

Each neuron was represented by two functional compartments: a soma and an axonal propagation component

Neuron Model The neuronal model employed in this study is based on the Hodgkin-Huxley (HH) formalism. Each neuron was represented by two functional compartments: a soma and an axonal propagation component. The soma represents the main neuronal body in which membrane potential dynamics are generated, whereas the axonal component propagates somatic voltage ...

work page

[2] [2]

Synapse Model Synapses between neurons facilitate signal transmission through the release of excitatory and inhibitory neurotransmitter vesicles from presynaptic to postsynaptic neurons. When certain events happen, a presynaptic neuron can release neurotransmitters, which bind to receptors and affect the conductance of ionic channels in the postsynaptic n...

work page

[3] [3]

We used a learning window of 50ms, meaning learning only occurs for presynaptic and postsynaptic action potentials that happen within 50ms

Spike Timing Dependent Plasticity Our model implements biologically inspired plasticity, called spike timing dependent plasticity, that allows synaptic weights, in our case number of receptors, to be modified based on the relative timing between presynaptic and postsynaptic neurons of a synapse. We used a learning window of 50ms, meaning learning only occ...

work page 1998

[4] [4]

Each neuron had an 80% chance to create a connection with any other neuron, creating around 32000 total synapses

Network Architecture and Simulation The network consisted of 200 neurons, 40 inhibitory and 160 excitatory. Each neuron had an 80% chance to create a connection with any other neuron, creating around 32000 total synapses. Receptor counts were initialized using Gaussian distributions, with AMPA having 120 mean and 12 variance, and GABA having 200 mean and ...

work page arXiv 2021