The Setting
Take a finite algebra and give the edge between two of its elements the color , the principal congruence of the pair, an element of . A one-variable translation of an operation of never raises a color, , and it may collapse an edge to a loop. Which edges can such maps reach from a few given ones, and which elements do the reached edges link? The same questions make sense for an arbitrary coloring, with every self-map that never raises a color in place of the translations.
Colorings
For a set with a relation , a lattice with its order for instance, an -coloring of a set of vertices is any map ; the value is the color of the edge , and loops count as edges.
The Definition
Definition 1. Let be an -coloring of a set . A self-map , not necessarily injective, is admissible when (1)✦ For a finite set of seed edges, the marked edges of the graph on are the images of seeds under admissible self-maps: (2)✦ Two vertices and are connected in , written , when they lie in one class of the equivalence relation on generated by : for some there are vertices with (3)✦
In the paper, the definition reads: an -colored complete graph is a finite nonempty vertex set with a symmetric and , the least element of , loops being edges; a placement into another such graph is admissible when for all ; for a nonempty set of edges, is the undirected graph on whose marked adjacencies are the pairs with and admissible, collapsed pairs allowed; and .
What Is Checked
Admissibility (1), the marked edges (2) and connectedness in (3) are defined over any set of vertices, any set with a relation , any -coloring of and, for the last two, any finite set of ordered pairs of vertices as seeds.