1: \begin{abstract}

2: The clusters of

3: a distribution are often defined by the

4: connected components of a density level set.

5: However, this definition

6: depends on the user-specified level.

7: We address this issue by

8: proposing a simple, generic algorithm, which uses

9: an almost arbitrary level set estimator to estimate

10: the smallest level at which there are more than one

11: connected components.

12: In the case where this algorithm is fed

13: with histogram-based level set estimates,

14: we provide a finite sample

15: analysis, which is then used to show that the

16: algorithm consistently estimates both the

17: smallest level and the corresponding connected

18: components. We further establish rates of

19: convergence for the two estimation problems, and

20: last but not least, we present a simple,

21: yet adaptive strategy

22: for determining the width-parameter of the

23: involved density estimator in a data-depending

24: way.

25: \end{abstract}