An efficient GP approach to recognizing cognitive tasks from fNIRS neural signals

This paper presents a new genetic programming (GP) approach to accurately classifying cognitive tasks from non-stationary and noisy fNIRS neural signals. To this end, a new GP that effectively handles multiclass problems is developed. In accordance with multi-tree structure, GP operators are innovated: crossover exchanges every subtree of parents without suffering from any incongruity problem and mutation fine-tunes candidate solutions by a less destructive process. Experimental results verifies the effectiveness of the proposed GP classifier over existing references.