“…In a typical scenario with traditional machine learning approaches, EEG artifacts, e.g., eye/muscle movements and electrical noise, are first removed by band-pass filtering and detrending [7], [8]. Then, time domain features [9], frequency domain features [10], temporalspatial features [11], [12], or nonlinear features [13], can be extracted, selected [14], and finally sent to a classifier, e.g., support vector machine (SVM) [10], [15], logistic regression (LR) [12], or multilayer perceptron (MLP) [16], for classification [17].…”