Improving a cortical pyramidal neuron model’s classification performance on a real-world ecg dataset by extending inputs

Abstract

Pyramidal cells are the most prevalent neuronal type in the cortex, receiving thousands of synaptic inputs from all over the brain, and sending the largest axon outputs. They have a variety of active conductivities and complex morphologies that support highly nonlinear dendritic calculations. There has been a growing interest in understanding the classification abilities of pyramidal neurons. The perceptron learning algorithm, one of the foundations of machine learning, uses the highly simplified mathematical abstraction of a neuron, and it is unclear to what extent real biophysical neurons can perform perceptron like learning. In this article, we investigated the performance of a pyramidal neuron model in the classification problem of a two-class ECG dataset for different synaptic regions by using the perceptron learning method. The main purpose of this study is to reveal what role the soma, basilar and apical dendrites play in a classification problem. We concluded that when the synaptic receptor locations are selected close to the soma, classification performance close to Perceptron Learning in a Cortical Pyramidal Neuron Model the single layer perceptron can be obtained. The results indicated that the pyramidal neuron can successfully classify real-world data.