In this algorithm, an effective feature subset is selected according to the maximum correlation and minimum redundancy criteria. This method generalizes the neighborhood entropy function of single-label learning to multi-label learning. Nevertheless, the aforementioned algorithms assume full knowledge of ...