Mean-field theory of a plastic network of integrate-and-fire neurons

Firing rate distributions in plastic networks of spiking neurons

In recurrent networks of leaky integrate-and-fire neurons, the mean-field theory has been instrumental in capturing the statistical properties of neuronal activity, like firing rate distributions. This theory has been applied to networks with either homogeneous synaptic weights and heterogeneous connections per neuron or vice versa. Our work expands mean-field models to include networks with both types of structural heterogeneity simultaneously, particularly focusing on those with synapses that undergo plastic changes. The model introduces a spike trace for each neuron, a variable that rises with neuron spikes and decays without activity, influenced by a degradation rate r p and the neuron's firing rate ν. When the ratio α = ν/r p is significantly high, this trace effectively estimates the neuron's firing rate, allowing synaptic weights at equilibrium to be determined by the firing rates of connected neurons. This relationship is incorporated into our mean-field formalism, providing exact solutions for firing rate and synaptic weight distributions at equilibrium in the high α regime. However, the model remains accurate within a practical range of degradation rates, as demonstrated through simulations with networks of excitatory and inhibitory neurons. This approach sheds light on how plasticity modulates both activity and structure within neuronal networks, offering insights into their complex behavior.

Self-Organization of Microcircuits in Networks of Spiking Neurons with Plastic Synapses

The synaptic connectivity of cortical networks features an overrepresentation of certain wiring motifs compared to simple random-network models. This structure is shaped, in part, by synaptic plasticity that promotes or suppresses connections between neurons depending on their joint spiking activity. Frequently, theoretical studies focus on how feedforward inputs drive plasticity to create this network structure. We study the complementary scenario of self-organized structure in a recurrent network, with spike timing-dependent plasticity driven by spontaneous dynamics. We develop a self-consistent theory for the evolution of network structure by combining fast spiking covariance with a slow evolution of synaptic weights. Through a finite-size expansion of network dynamics we obtain a low-dimensional set of nonlinear differential equations for the evolution of two-synapse connectivity motifs. With this theory in hand, we explore how the form of the plasticity rule drives the evolution of microcircuits in cortical networks. When potentiation and depression are in approximate balance, synaptic dynamics depend on weighted divergent, convergent, and chain motifs. For additive, Hebbian STDP these motif interactions create instabilities in synaptic dynamics that either promote or suppress the initial network structure. Our work provides a consistent theoretical framework for studying how spiking activity in recurrent networks interacts with synaptic plasticity to determine network structure.

Sisyphus effect in pulse-coupled excitatory neural networks with spike-timing-dependent plasticity

The collective dynamics of excitatory pulse-coupled neural networks with spike-timing-dependent plasticity (STDP) is studied. Depending on the model parameters stationary states characterized by high or low synchronization can be observed. In particular, at the transition between these two regimes, persistent irregular low frequency oscillations between strongly and weakly synchronized states are observable, which can be identified as infraslow oscillations with frequencies ≃0.02-0.03 Hz. Their emergence can be explained in terms of the Sisyphus effect, a mechanism caused by a continuous feedback between the evolution of the coherent population activity and of the average synaptic weight. Due to this effect, the synaptic weights have oscillating equilibrium values, which prevents the neuronal population from relaxing into a stationary macroscopic state.

Emergence of slow collective oscillations in neural networks with spike-timing dependent plasticity

The collective dynamics of excitatory pulse coupled neurons with spike-timing dependent plasticity is studied. The introduction of spike-timing dependent plasticity induces persistent irregular oscillations between strongly and weakly synchronized states, reminiscent of brain activity during slow-wave sleep. We explain the oscillations by a mechanism, the Sisyphus Effect, caused by a continuous feedback between the synaptic adjustments and the coherence in the neural firing. Due to this effect, the synaptic weights have oscillating equilibrium values, and this prevents the system from relaxing into a stationary macroscopic state.

Analytical condition for synchrony in a neural network with two periodic inputs

In this study, we apply a mean-field theory to the neural network model with two periodic inputs in order to clarify the conditions of synchronies. This mean-field theory yields a self-consistent condition for the synchrony and enables us to study the effects of synaptic connections for the behavior of neural networks. Then, we obtain a condition of synaptic connections for the synchrony with the cycle time T. The neurons in neural networks receive sensory inputs and top-down inputs from outside of the network. When the network neurons receive two or more inputs, their synchronization depends on the conditions of inputs. We also analyze this case using the mean-field theory. As a result, we clarify the following points: (i) The stronger synaptic connections enhance the shorter synchrony cycle of neurons. (ii) The cycle of the synchrony becomes longer as the cycle of external inputs becomes longer. (iii) The relationships among synaptic weights, the properties of input trains, and the cycle of synchrony are expressed by one equation, and there are two areas for asynchrony. With regard to the third point, the yielded equation is so simple for calculation that it can easily provide us with feasible and infeasible conditions for synchrony.

Event-driven simulations of a plastic, spiking neural network

We consider a fully connected network of leaky integrate-and-fire neurons with spike-timing-dependent plasticity. The plasticity is controlled by a parameter representing the expected weight of a synapse between neurons that are firing randomly with the same mean frequency. For low values of the plasticity parameter, the activities of the system are dominated by noise, while large values of the plasticity parameter lead to self-sustaining activity in the network. We perform event-driven simulations on finite-size networks with up to 128 neurons to find the stationary synaptic weight conformations for different values of the plasticity parameter. In both the low- and high-activity regimes, the synaptic weights are narrowly distributed around the plasticity parameter value consistent with the predictions of mean-field theory. However, the distribution broadens in the transition region between the two regimes, representing emergent network structures. Using a pseudophysical approach for visualization, we show that the emergent structures are of "path" or "hub" type, observed at different values of the plasticity parameter in the transition region.