1: \begin{abstract}

2: Convex clustering, a convex relaxation of k-means clustering and hierarchical

3: clustering, has drawn recent attentions since it nicely addresses

4: the instability issue of traditional non-convex clustering methods.

5: Although its computational and statistical properties have been recently

6: studied, the performance of convex clustering has not yet been investigated

7: in the high-dimensional clustering scenario, where the data contains a

8: large number of features and many of them carry no information about

9: the clustering structure. In this paper, we demonstrate that the performance

10: of convex clustering could be distorted when the uninformative features

11: are included in the clustering. To overcome it, we introduce a new

12: clustering method, referred to as \textit{Sparse Convex Clustering},

13: to simultaneously cluster observations and conduct feature selection.

14: The key idea is to formulate convex clustering in a form of regularization,

15: with an adaptive group-lasso penalty term on cluster centers. In order

16: to optimally balance the trade-off between the cluster fitting and

17: sparsity, a tuning criterion based on clustering stability is developed.

18: Theoretically, we obtain a finite sample error bound for our estimator and further establish its variable selection consistency.

19: The effectiveness of the proposed method is examined through a variety of numerical

20: experiments and a real data application.

21: \end{abstract}

22: