Fluctuation-response relations and response-response relations for membrane voltage and spike train of stochastic integrate-and-fire neurons

Kolja Klett, Benjamin Lindner · 2025

Read Fluctuation-response relations and response-response relations for membrane voltage and spike train of stochastic integrate-and-fire neurons by Kolja Klett, Benjamin Lindner (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-07434.txt.

Introduction

Kolja Klett and Benjamin Lindner published a paper in 2025 on the mathematics of simple model neurons that fire spikes at random. Such neurons both jitter on their own and answer outside signals, and earlier work tied those two behaviors together with exact relations. The authors widen that family of relations.

They add formulas covering the voltage inside the cell, including its nonlinear range below firing. For the simplest leaky model they work out closed forms for the voltage spectrum and its response to input. A second kind of relation links the spike train's response to the voltage's response in a broader setting. A version with an adaptation current and colored noise is also solved. These tools let modelers infer hidden responses from visible noise.

Catalog record

CREATORKolja Klett, Benjamin Lindner
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-07434

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Fluctuation-response relations and response-response relations for membrane voltage and spike train of stochastic integrate-and-fire neurons Kolja Klett, Benjamin Lindner Abstract Neurons display spontaneous spiking (in the absence of stimulus signals) as well as a characteristic response to time-dependent external stimuli. In a simple but important class of stochastic neuron models, the integrate-and-fire model with Gaussian current noise, both aspects can mathematically be related via fluctuation-response relations (FRRs) as has been shown recently Lindner 2022a . Here we extend the class of FRRs to include the susceptibilities of the membrane voltage and subthreshold voltage nonlinearity as well as the power spectrum of the membrane voltage. For a simple but often considered IF model, the leaky IF model with white Gaussian noise, we exploit the FRRs and derive explicit expressions for the power spectrum and susceptibility of the subthreshold membrane voltage. We also put forward a relation between the response functions of the spike train and membrane voltage, a response-response relation (RRR) that holds true for a more general setting than considered in most parts of the paper. For the generalized IF model with an adaptation current and colored Gaussian noise we derive an FRR and an RRR. We briefly discuss useful applications of the derived FRRs and RRRs. I Introduction Nerve cells (neurons) in the brain encode information on sensory signals in sequences of action potentials, called spike trains Kandel et al. 2000 ; Rieke et al. 1996 ; Dayan and Abbott 2001 . Often, however, neurons also display a considerable spontaneous spiking activity in the absence of sensory stimulation and they also show spiking variability in the presence of stimuli (trial-to-trial variability). Both aspects are indicative of intrinsic and extrinsic noise sources and necessitate stochastic models of neural firing which have by now a long history in computational neuroscience Holden 1976 ; Ricciardi 1977 ; Tuckwell 1989 ; Gerstner et al. 2014 ; Gabbiani and Cox 2017 . These models take fluctuations into account and can reproduce both the firing statistics of stimulus-free neurons as well as the response and trial-to-trial variability in presence of a stimulus. An important class of phenomenological models that have been applied both to experimental data of single cells as well as in theoretical studies of neural networks is the integrate-and-fire (IF) neuron model Fourcaud and Brunel 2002 ; Burkitt 2006a ; Burkitt 2006b . The calculation of the characteristics of spontaneous firing (e.g. the spike-train power spectrum) and of the response to time-dependent stimulation (e.g. the susceptibility) is a nontrivial problem in the theory of stochastic processes because of the unavoidable nonlinearity in the mathematical description of spike generation. There are a couple of exact results for the power spectrum and susceptibility of integrate-and-fire models with white Gaussian noise Lindner et al. 2002 ; Brunel et al. 2001 ; Lindner and Schimansky-Geier 2001 , with colored telegraph noise Droste and Lindner 2017a , and with white shot noise Richardson and Swarbrick 2010 ; Droste and Lindner 2017b ; however, generally, such results are difficult to derive, especially for biophysically more realistic models that include spike-frequency adaptation, short-term plasticity, and/or temporally correlated (colored) noise. In statistical physics, it is known that the spontaneous activity and the response to perturbations of stochastic systems are not independent but often connected by fluctuation-dissipation theorems (FDTs), also referred to as fluctuation-response relations (FRRs) Marconi et al. 2008 . In recent years, different research groups have attempted to find similar relationships in neural systems by different methods and at very different levels (single neuron, neural population, network of networks). Investigating whole brain regions using electroencephalogram recording and functional magnetic resonance imaging (fMRI), multiple groups found a complex interplay of the spontaneous and evoked activity He 2013 ; Huang et al. 2017 . FDTs were found for symmetrically coupled brain regions in a model of coupled stochastic Stuart-Landau oscillators near criticality fitted to whole-brain fMRI data: the violation of the equilibrium FDT were associated with complex cognitive tasks and used to quantify deviations from equilibrium dynamics in the brain Deco et al. 2023 . At the population level, the spontaneous fluctuations and response of a linearized noisy version of the Wilson-Cowan model were compared with experimental measurements of spontaneous and evoked activity using magnetoencephalography Sarracino et al. 2020 . This method has been expanded to account for imbalanced networks in Nandi et al. 2023 . For a network of leaky IF neurons in discrete time, a theoretical framework has been proposed based on the probability density of the spike times and been used to derive approximate FDTs. The applicability of this approach to more general time-continuous IF models (which were shown to best describe spiking in real neurons Jolivet et al. 2008 ) appears to be difficult. Further, a recently developed universal description of a multidimensional stochastic oscillator led to yet another simple FRR, although the latter holds only for an abstract complex-valued variable to which the system variables have to be transformed Pérez-Cervera et al. 2023 . For stochastically spiking neurons, in particular for the simple yet phenomenologically very successful IF model, it has turned out that spontaneous firing statistics and response characteristics in terms of the spike train and the membrane potential are comparatively simple related. For different types of IF models with Gaussian noise, in Lindner 2022a one of us derived FRRs that connect the susceptibility of the firing-rate response to the spectral statistics of spike train and subthreshold membrane voltage in the absence of a stimulus signal (i.e. only intrinsic noise is present). By the same approach, FRRs have also been found for IF models with refractory period Puttkammer and Lindner 2024 and shot noise Stubenrauch and Lindner 2024 . In this work, we will use the methods from Lindner 2022a to derive relations for different IF neurons that focus on the spontaneous and response statistics of the subthreshold membrane voltage, a standard observable in many neuroscientific experiments. In addition to FRRs for the subthreshold membrane voltage of the different models, which we confirm by comparison with stochastic simulations, we will also derive response-response relations (RRRs) that connect the susceptibilities of spike train and membrane voltage. All these relations may turn out to be useful (i) to estimate voltage statistics from spike-train statistics (or the other way around), (ii) to extract intrinsic noise statistics (power spectrum), and (iii) to validate modeling approaches to experimental data involving the IF framework. Our paper is structured as follows. We start by introducing the IF model as well as typical measures of fluctuations and response characteristics in sec. II . Then, we derive FRR and RRR for a simple leaky IF model with white Gaussian noise in sec. III . In sec. IV , we extend these relations to the case of a more general IF model with adaptation and colored noise. We conclude by summarizing our main results and discussing possible applications of the derived relations in sec. V . II Models and measures In the IF model the dynamics of the subthreshold membrane potential v are described by a Langevin equation \dot{v}=F(v,a)+\eta(t)+\varepsilon s(t) (1) where F captures the deterministic behavior of v and may also include a dynamical adaptation variable a(t) that follows its own dynamics (see below sec.

Read the complete text