--- title: Simultaneous Firing & Connection Formation url: https://www.emergentmind.com/topics/simultaneous-firing-and-connection-formation type: topic --- # Simultaneous Firing & Connection Formation Simultaneous firing and connection formation refer to the processes by which excitable elements—biological or artificial—generate synchronized activity, and how these patterns are shaped or enabled by the evolving topology of inter-element connections. In neurobiology, these dynamics underlie phenomena such as synchrony, oscillations, and plasticity in neural networks. In synthetic systems and mathematical models, these mechanisms illuminate the interplay between network connectivity, node excitability, and emergent collective behaviors. ## 1. Dynamical Models of Simultaneous Firing Simultaneous firing in sparsely connected ensembles can be rigorously studied using systems of coupled excitable nodes, such as FitzHugh–Nagumo elements. Each node is modeled by a fast activator $x_j(t)$ and a slow inhibitor $y_j(t)$, with dynamics: \[ \epsilon\,\dot x_j = x_j - \frac{x_j^3}{3} + y_j + I, \qquad \dot y_j = a - x_j, \] where $\epsilon \ll 1$ sets the time-scale separation, $a > 1$ ensures excitability, and $I$ is a constant bias. Nodes are coupled via a time-dependent adjacency matrix $A_{ij}(t)$ encoding the instantaneous network: \[ \epsilon\,\dot x_j = x_j - \tfrac{x_j^3}{3} + y_j + k\sum_{i=1}^N A_{ji}(t)\bigl(x_i - x_j\bigr), \quad \dot y_j = a - x_j. \] Instantaneous degree is strictly constrained: $\sum_{i=1}^N A_{ij}(t)\leq 1$ for all $j$, so that each node is connected to at most one partner at a given time. The coupling strength $k$ is typically set to unity for simulation [1206.6975]. ## 2. Network Rewiring and Connection Formation Protocols Two primary schemes realize time-evolving networks with degree $\leq 1$: - **Fixed Matching:** At all times, exactly $N/2$ undirected links ensure that each node has one partner. At total rate $\lambda N$, pairs of links $(j,j^*), (k,k^*)$ are randomly selected, and endpoints are swapped to form $(j,k), (j^*,k^*)$. - **Underlying Substrate:** A substrate graph $G$ with $L$ possible edges is defined. Nodes can activate dormant edges at rate $\lambda$, deactivating any pre-existing edge to maintain degree $\leq 1$. The instantaneous adjacency $A_{ij}(t)$ is symmetric with row sums $\leq 1$. The reconnection rate $\lambda$ governs the temporal statistics of partner switching, hence the pattern formation dynamics [1206.6975]. In experimental neuroscience, "single-neuron mimicry" allows emulation of a full $N$-node recurrent network using one living neuron. For each network "snapshot," a node's recent inputs are precisely replayed on a biological cell, and its firing or failure is mapped to the network's evolution. Connectivity is encoded in the matrix $W_{ij}$ (weights) and $\tau_{ij}$ (delays), dictating which virtual nodes stimulate one another, and at which times [1601.02189]. ## 3. Synchronization Metrics and Order Parameters Synchrony is quantified via several order parameters: - **Kuramoto Order Parameter ($\rho$):** Using the geometric phase $\phi_j(t)=\arctan\bigl(y_j(t)/x_j(t)\bigr)$, \[ \rho(t)\,e^{i\Psi(t)} = \frac{1}{N}\sum_{j=1}^N e^{i\phi_j(t)};\quad \rho = \langle\rho(t)\rangle_t. \] Static phase coherence corresponds to $\rho\approx 1$. - **Shinomoto–Kuramoto Parameter ($\zeta$):** Measures dynamic, collective pulsations, \[ \zeta = \Bigl\langle \bigl|\rho(t)e^{i\Psi(t)} - \langle\rho(t)e^{i\Psi(t)}\rangle_t\bigr| \Bigr\rangle_t, \] which is nonzero only if the center of mass of phases rotates around the circle. - **Mean Firing Current ($J$):** \[ J = \Bigl\langle \Bigl|\frac{1}{N}\sum_{j=1}^N \dot x_j(t)\Bigr|\Bigr\rangle_t, \] signaling activity regardless of synchrony. These parameters reveal regimes of quiescence ($\rho\approx1$, $\zeta\approx0$, $J\approx0$), partial synchrony, and full, dynamic global firing [1206.6975]. ## 4. Mechanisms and Windows for Globally Synchronized Firing The occurrence of fully synchronized firing (all elements spike in unison) is contingent on the interplay between network evolution rate $\lambda$ and coupling $k$: - **Slow Rewiring ($\lambda\ll\lambda_{c1}$):** Links persist so long that nodes always relax to rest before partners change; no excitation propagates. - **Intermediate Rewiring ($\lambda_{c1}<\lambda<\lambda_{c2}$):** When typically $\lambda \sim (\tau_{\rm left}+\tau_{\rm right})^{-1}$, nodes near rest can suddenly couple to a firing partner, receiving an "impulsive kick" sufficient to trigger their own excursion. This transient drives a spatial-temporal cascade, rapidly synchronizing the entire network. Avalanche-like propagation is observed, with a bell-shaped $\zeta(\lambda)$ and $J(\lambda)$ profile peaking in this window. - **Fast Rewiring ($\lambda\gg\lambda_{c2}$):** Fluctuating couplings self-average; nodes experience only the static, ensemble-mean field and settle to a quiescent fixed point. No synchrony emerges for $k