Export Reviews, Discussions, Author Feedback and Meta-Reviews

We thank all reviewers for their detailed and constructive comments. Reviewer 1 and 3 commented that it would be better if the algorithm can handle many (e.g., 90% or 99%) irrelevant features. We certainly agree with this. However, we want to point out that the subspace clustering problem with corrupted feature is an open problem, even when the fraction of irrelevant features are mild. Indeed, as shown in the paper, SSC and LASSO-SSC break down even with few irrelevant features.