A comparison of classification algorithms for Event Related Potentials

We compare six methods for classifying low signal-to-noise Event Related Potential (ERP) data into deceiving and nondeceiving. Three methods perform poorly: Correlation Coefficient (CD), Support Vector Machine with linear Kernel (LSVM) and SVM with radial basis kernel (RBSVM), while another, Amplitude Difference (AD), performs satisfactorily. However, a simple by-individual weight template method performs better still.