In this case, the algorithm is trained using both input features and output labels. It aids in the establishment of a link between variables by estimating how one influences the other.