Pavlovian Conditioning-Driven Three-Factor STDP-Inspired Learning in FitzHugh-Nagumo Spiking Neural Networks for Classification Tasks

Spiking Neural Networks (SNNs) provide biologically plausible models of computation that closely mimic neural dynamics and plasticity. Traditional supervised training methods often rely on gradient-based backpropagation, limiting biological realism and neuromorphic hardware implementation. This paper proposes a biologically inspired supervised learning rule combining Pavlovian conditioning with three-factor Spike-Timing Dependent Plasticity (STDP) in FitzHugh-Nagumo (FHN) neuron networks. Unlike conventional STDP, our approach employs a reference neuron producing an ideal spike train, guiding synaptic updates without explicit gradients or error functions. We demonstrate the model’s feasibility on simple pattern classification tasks, highlighting stable learning dynamics, high accuracy for binary patterns, and good generalization for grayscale inputs, thus presenting a novel, biologically plausible alternative for supervised SNN training.