Chaos-driven dynamics in Morris-Lecar neurons: Implications for real-world classification

Spiking Neural Networks (SNNs), formulated through mathematical models that closely approximate biological neurons, have gained significant attention due to their ability to represent neural dynamics with high fidelity. These models enable the analysis of real biological neuron parameters under varying conditions, among which chaotic neural activity stands out as a crucial factor in cognitive processing. This study, structured in two parts, investigates the classification performance of SNNs composed of different classes of Morris-Lecar neurons and compares them with conventional Artificial Neural Networks (ANNs) of similar architecture. In the second part, the impact of chaotic environmental conditions on the classification performance of these SNNs is examined, revealing how different levels of chaotic input currents influence network behavior. To the best of our knowledge, this is the first study to explore the classification capabilities of an SNN composed of Morris-Lecar neurons under chaotic conditions. In addition to this contribution, we also propose a rectified version of the Morris-Lecar neuron model that supports gradient-based training. Furthermore, we define a novel phenomenon chaotic classification resonance which, to the best of our knowledge, has not been previously reported in the context of SNN-based classification tasks. The findings demonstrate that an SNN incorporating Morris-Lecar neurons can achieve classification accuracy comparable to an ANN of the same architecture activated by the ReLu function. More strikingly, our results indicate that under chaotic conditions, the classification performance of the SNN exhibits a behavior akin to chaotic resonance. Specifically, simulations reveal that this phenomenon termed chaotic classification resonance significantly enhances the classification accuracy of an SNN composed of Class-III Morris-Lecar neurons when an optimal level of chaotic input current is applied.