The mean distance of a (connected) graph is used with two closely related normalizations. The first averages all entries of the graph distance
matrix, including the diagonal entries ; Garijo et al. (2023) denote this normalization
by
and note that it is the arithmetic
mean of the entries of the distance matrix. The second, used for example by Doyle
and Graver (1977) and described by Garijo et al. (2023) as the "most
usual" convention, averages the nonzero distances over unordered pairs of distinct
vertices.
The convention used here and by the GraphData property "MeanDistance" is the all-distance-matrix-entry normalization.
For a graph on
vertices with graph transmission
and Wiener index
, it is given by
|
(1)
| |||
|
(2)
|
Closed forms for some classes of named graphs under the all-distance-matrix-entry convention are given in the following table.
Under the distinct-vertex-pair convention, the mean distance is instead
|
(3)
| |||
|
(4)
| |||
|
(5)
|
for connected graphs with .