Investigation of the performance of the pyramidal neuron model in the classification problem

Pyramidal neurons are the most common type of neuron in the cortex, receiving thousands of synaptic inputs from different parts of the brain and sending the largest axon outputs. They have a variety of active conductivities and complex morphologies that support nonlinear high dendritic calculations. In the thesis study, a detailed pyramidal neuron model and perceptron learning algorithm were used to classify two and four-class ECG data. Gray coding was used to generate spikes from ECG data. The classification performance of the subcellular regions of the pyramidal neuron was investigated for the two-class data. Compared with the equivalent single layer perceptron, the pyramidal neuron was found to perform poorly due to the weight constraint. However, a proposed mirroring approach for the inputs has significantly improved the classification performance of the neuron. The highest two class classification accuracy of 95.3% was obtained in the apical region. In the four class classification, the effect of Gray code length and ECG RR sample number on the classification performance of the pyramidal neuron was investigated. The highest classification accuracy was obtained in the basilar region as 98.8%. The results show that the pyramidal neuron can successfully classify real world data.