1: \begin{abstract} \small\baselineskip=9pt

2: Accurate estimation of Intrinsic Dimensionality (ID) is of crucial

3: importance in many data mining and machine learning tasks, including

4: dimensionality reduction, outlier detection, similarity search and subspace

5: clustering. However, since their convergence generally requires sample sizes (that

6: is, neighborhood sizes) on the order of hundreds of points, existing ID

7: estimation methods may have only limited usefulness for applications in

8: which the data consists of many natural groups of small size. In this

9: paper, we propose a local ID estimation strategy stable even for `tight'

10: localities consisting of as few as 20 sample points. The estimator applies

11: MLE techniques over all available pairwise distances among the members of

12: the sample, based on a recent extreme-value-theoretic model of intrinsic

13: dimensionality, the Local Intrinsic Dimension (LID). Our experimental

14: results show that our proposed estimation technique can achieve notably

15: smaller variance, while maintaining comparable levels of bias, at much

16: smaller sample sizes than state-of-the-art estimators.

17: \end{abstract}

18: