Based on probability density approximation, a novel unsupervised feature ranking approach was proposed and could be applied to feature selection.

• 摘要依据概率密度逼近提出了一种新的无监督特征排序，应用于特征选择降维。