Abstract
We introduce a new formalism for evaluating analytically the cross-correlation structure of a finite-size firing-rate network with recurrent connections. The analysis performs a first-order perturbative expansion of neural activity equations that include three different sources of randomness: the background noise of the membrane potentials, their initial conditions, and the distribution of the recurrent synaptic weights. This allows the analytical quantification of the relationship between anatomical and functional connectivity, i.e. of how the synaptic connections determine the statistical dependencies at any order among different neurons. The technique we develop is general, but for simplicity and clarity we demonstrate its efficacy by applying it to the case of synaptic connections described by regular graphs. The analytical equations so obtained reveal previously unknown behaviors of recurrent firing-rate networks, especially on how correlations are modified by the external input, by the finite size of the network, by the density of the anatomical connections and by correlation in sources of randomness. In particular, we show that a strong input can make the neurons almost independent, suggesting that functional connectivity does not depend only on the static anatomical connectivity, but also on the external inputs. Moreover we prove that in general it is not possible to find a mean-field description à la Sznitman of the network, if the anatomical connections are too sparse or our three sources of variability are correlated. To conclude, we show a very counterintuitive phenomenon, which we call stochastic synchronization, through which neurons become almost perfectly correlated even if the sources of randomness are independent. Due to its ability to quantify how activity of individual neurons and the correlation among them depends upon external inputs, the formalism introduced here can serve as a basis for exploring analytically the computational capability of population codes expressed by recurrent neural networks.
Introduction
Understanding how the physical "wiring" of the brain (anatomical connectivity) relates to the statistical dependencies between firing neurons (functional connectivity) is a central challenge in neuroscience. Historically, researchers have relied on mean-field approximations, which assume the number of neurons ($N$) approaches infinity. Under these "thermodynamic limit" conditions, neurons often become independent—a phenomenon known as the "propagation of chaos." However, biological networks are finite, often sparse, and subject to multiple noise sources. This paper addresses the need for analytical tools that can quantify correlations in realistic, finite-size recurrent networks without the computational burden of brute-force simulations.
Methodology
The authors develop a formalism based on a first-order perturbative expansion of neural activity equations. The model utilizes a standard firing-rate framework where the membrane potential of a neuron is determined by recurrent synaptic weights, an external input, and a non-linear activation function.
Unlike models that consider a single noise source, this framework incorporates three: background membrane potential noise (Brownian motion), variability in initial conditions, and randomness in synaptic weight distributions. By treating the standard deviations of these noise sources as small parameters, the non-linear activation function is expanded around a stationary solution. This linearizes the complex stochastic system, allowing for closed-form solutions using spectral analysis on regular graph topologies, such as ring lattices ($C_{10}$), fully connected graphs ($K_{10}$), and 4D hypercubes ($Q_4$).
Key Insights and Results
The analytical formulas derived through this perturbative approach showed remarkable agreement with numerical simulations (Euler–Maruyama and Monte Carlo methods), with a percentage-relative error typically below 3.5% for standard noise levels.
Dynamic Functional Connectivity
A core finding is that functional connectivity is not a static map of anatomy. The study demonstrates that strong external inputs can "saturate" neurons, effectively reducing their gain and making them act independently, regardless of physical synaptic links. This highlights that functional connectivity is a dynamic property shaped by external stimuli.
Failure of Mean-Field Assumptions
The paper proves that traditional mean-field theories, such as Sznitman’s, fail if anatomical connections are too sparse or if noise sources are correlated. In such scenarios, neurons remain significantly correlated even as network size grows, contradicting the assumption of independence in the thermodynamic limit.
Stochastic Synchronization
Perhaps the most counterintuitive discovery is "stochastic synchronization." By precisely tuning network parameters—specifically moving the largest eigenvalue of the system's Jacobian toward zero—neurons can become perfectly correlated ($Corr = 1$) even when their noise sources are entirely independent.
Conclusion
This new formalism provides a powerful analytical basis for exploring the computational capabilities of population codes in recurrent neural networks. By quantifying how individual neural activity and inter-neuronal correlations depend on external inputs and network size, it offers a more nuanced understanding of brain function than traditional mean-field models. Future work will aim to extend this technique to irregular networks, oscillatory states, and time-dependent structural changes like synaptic plasticity.
Summary updated