Astrocytes Stabilize Neural Simplices Against Noise in STDP-based Spiking Neural Networks

This study investigates synaptic reconfiguration in spiking neural networks (SNNs) under spike-timing-dependent plasticity (STDP) learning and the noise-resilience mechanisms enabled by astrocyte integration. We implemented a neuronal network with connectivity governed by exponential distance distributions. STDP-driven reorganization was analyzed during training on binary digit images under varying noise conditions. Results demonstrate that noise-free inputs produce stimulus-specific neural simplices, while noise corrupts simplex composition through inclusion of non-specific neurons. Integration of astrocytes bidirectionally coupled with neuronal ensembles significantly restored stimulus specificity under noise by modulating synaptic efficacy via calcium-dependent mechanisms. The Bron-Kerbosch algorithm revealed that astrocyte-mediated synaptic gating enhances signal extraction in noisy environments, providing a biologically plausible framework for robust neuromorphic computation. These findings establish astrocytes as critical components in noise-resilient neural architectures.