After reading the problem the example I originally gave here is incorrect.
The graph is simple and directed. Key things to remember are that the edges
leaving a node all have a number
such that
, and that the
sum of the
's on the edges leaving a single node is one. Also, each edge
has a label (a letter is a good example of a label, because then you can
think of a path as containing a word, and the probability as the probability
of getting that word). The labels are not necessarily unique; in fact a
single edge can have more than one identically labelled edge leaving it.
Group into
"clusters" by choosing three points from the original set and
minimizing the distances between each marked point and your chosen point.