An Analytically Solvable Model of Firing Rate Heterogeneity in Balanced State Networks

Alexander Schmidt,Peter Hiemeyer,Fred Wolf
2023-05-23
Abstract:Distributions of neuronal activity within cortical circuits are often found to display highly skewed shapes with many neurons emitting action potentials at low or vanishing rates, while some are active at high rates. Theoretical studies were able to reproduce such distributions, but come with a lack of mathematical tractability, preventing a deeper understanding of the impact of model parameters. In this study, using the Gauss-Rice neuron model, we present a balanced-state cortical circuit model for which the firing rate distribution can be exactly calculated. It offers selfconsistent solutions to recurrent neuronal networks and allows for the combination of multiple neuronal populations, with single or multiple synaptic receptors (e.g. AMPA and NMDA in excitatory populations), paving the way for a deeper understanding of how firing rate distributions are impacted by single neuron or synaptic properties.
Neurons and Cognition
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: in a balanced - state neural network, how to accurately calculate the neuronal firing rate distribution through an analytically solvable model, and understand the influence of individual neuron or synapse characteristics on the firing rate distribution. ### Background and Problem of the Paper The distribution of neural activities in neural cortical circuits usually shows a highly skewed shape. Many neurons fire action potentials at a low or near - zero rate, while some neurons are active at a high frequency. Theoretically, research has been able to reproduce this distribution, but due to the lack of mathematical tractability, these theoretical studies cannot deeply understand the influence of model parameters. This paper uses the Gauss - Rice neuron model to propose a balanced - state cortical circuit model, which can accurately calculate the firing rate distribution, thus providing a deeper perspective for understanding how the firing rate distribution is affected by individual neuron or synapse characteristics. ### Main Contributions 1. **Analytically Solvable Model**: This model provides a self - consistent solution, is applicable to recurrent neural networks, and allows the combination of multiple neuron populations, each of which can have single or multiple synaptic receptors (for example, AMPA and NMDA receptors in an excitatory population). 2. **Accurate Calculation of Firing Rate Distribution**: Through analytical methods, this model can accurately calculate the firing rate distribution, which provides a new way to understand the shape and range of the firing rate distribution. 3. **Analysis of Influencing Factors**: The model reveals how cell and synapse characteristics affect the firing rate distribution, especially for local circuits composed of excitatory and inhibitory neurons, where excitatory synaptic currents show different decay times. ### Method Overview - **Network Topology and Setup**: The research considered a network defined by a directed Erdős - Rényi graph, with a connection probability of \(p=\frac{K}{N}\), where \(K\) is the average number of connections per neuron, and \(N\) is the total number of neurons in the network. - **Input Current Statistics**: Calculate the mean and variance of the input current in the network through composite spike train statistics and time - averaged input current statistics. - **Membrane Potential Statistics**: Using the Gauss - Rice neuron model, derive the joint distribution of the membrane potential and its derivative, and then obtain the neuronal firing rate response. - **Mixed Receptor Types**: The model also considers the mixed situation of multiple synaptic receptors, such as the fast AMPA receptor and the slow NMDA receptor, further improving the applicability of the model. ### Results - **Self - consistent Solution**: The model provides a self - consistent solution that can analytically describe the firing rate distribution. - **Analytical Form of Firing Rate Distribution**: Gives the analytical expression of the firing rate distribution, revealing the peak position and skewness of the firing rate distribution. - **Influence of Parameters**: Analyzes in detail the influence of network parameters (such as synaptic weights, connection probabilities, etc.) on the firing rate distribution. ### Conclusion By introducing the Gauss - Rice neuron model, this research successfully established an analytically solvable balanced - state cortical circuit model, which can accurately calculate the firing rate distribution and deeply understand the influence of individual neuron or synapse characteristics on the firing rate distribution. This model provides an important theoretical basis and tool for future research.