Tag
This paper proposes using random feature selection as a baseline and empirically shows that many state-of-the-art unsupervised feature selection methods are outperformed by random selection in both performance and efficiency.