Sürekli Zaman Uzayında Vuru Üreten Nöral A˘gın Öngörüsel Davranı¸sı

Predictive coding is a theory developed to explain how the brain predicts statistical regularities present in nature and eliminates redundant stimuli between cortical layers. A common method involves modeling a feedback system with neurons to process error signals. While treating neuron behavior as a filter provides computational benefits in some studies, modeling the behavior exhibited by a neuron exposed to synaptic stimulus, especially after the hyperpolarization process, along with its learning mechanism, as a continuous-time dynamic system, more effectively represents the nonlinear dynamic behavior of the neuron. Therefore, this study simulates a feedback-based Spiking Neural Network with a realistic neuron model that performs predictive inference and processes error signals. As a result of training, the postsynaptic neuron has successfully become selective to the initial input.