Cross-linkages in the Concise Columbia Encyclopedia (1989 2nd ed., online version). A set of data relating to 12025 encyclopedia entries and to 9778 cross-references or links was rebinned 10-fold for computational convenience to produce the 1203 times 978 entries by links array used here. The following image shows part of this array, subsequent to row/column permutation to force nonzero values as far as possible onto the diagonal.

Now we seek to find the most important "knots" or clusters. The following is a filtered version.

Show image of filtered data (0.3 MB)

*mr_detect in.fits out.fits *

*mr_detect -m 9 in.fits out.fits* (poisson few events)

*mr_detect -m 9 -w 2 in.fits out.fits*
(produces ellipses around clusters)

* mr_detect -m 9 -w 2 -L 3 -F 5 in.fits out.fits*
(start and finish scales for cluster detection)

The above figures are from SAOtng.

**To probe further**

"Computational methods for intelligent information access", M.W. Berry, S.T. Dumais and T.A. Letsche, University of Tennessee, 38 pp., 1997, and other papers.

*Image and Data Analysis: The Multiscale Approach*, J-L Starck,
F Murtagh and A Bijaoui, Cambridge University Press, 1998.