1: \begin{abstract}

2: Metric embedding has become a common technique  in the design of

3: algorithms. Its applicability is often dependent on how high the

4: embedding's distortion is. For example, embedding finite metric space into

5: trees may require linear distortion as a function of its size. Using

6: probabilistic metric embeddings, the bound on the distortion

7: reduces to logarithmic in the size.

8: 

9: We make a step in the direction of bypassing the lower bound on

10: the distortion in terms of the size of the metric. We define

11: ``multi-embeddings" of metric spaces in which a point is mapped

12: onto a set of points, while keeping the target metric of

13: polynomial size and preserving the distortion of paths. The

14: distortion obtained with such multi-embeddings into ultrametrics

15: is at most $O(\log \Delta\log\log \Delta)$ where $\Delta$ is the

16: \emph{aspect ratio} of the metric. In particular, for expander

17: graphs, we are able to obtain {\em constant} distortion embeddings

18: into trees in contrast with the $\Omega(\log n)$ lower bound for

19: all previous notions of embeddings.

20: 

21: We demonstrate the algorithmic application of the new embeddings

22: for two optimization problems: \emph{group Steiner tree} and \emph{metrical 

23: task

24: systems}.

25: \end{abstract}

26: