1: \begin{abstract}

2: This survey contains a selection of topics unified by the concept of

3: positive semi-definiteness (of matrices or kernels), reflecting natural

4: constraints imposed on discrete data (graphs or networks) or continuous

5: objects (probability or mass distributions). We put emphasis on entrywise

6: operations which preserve positivity, in a variety of guises. Techniques

7: from harmonic analysis, function theory, operator theory, statistics,

8: combinatorics, and group representations are invoked. Some partially

9: forgotten classical roots in metric geometry and distance transforms are

10: presented with comments and full bibliographical references. Modern

11: applications to high-dimensional covariance estimation and regularization

12: are included.

13: \end{abstract}