This research study contributes to unraveling the black box problem of ML models. Its main objective is to demonstrate that ML can accurately distinguish between brain states under different experimental conditions based on large feature sets of functional connectivity. Additionally, the study shows that reducing and interpreting input features helps in understanding the neuronal processes underlying differences between brain states. Specifically, the study examines how prolonged repetition of a cognitive task (or prolonged cognitive load) reconfigures the prestimulus functional network structure to adapt to task performance.
(DOI: 10.1063/5.0070493)