Two disconnected knowledge maps, K₁ and K₂, constitute a disjoint knowledge map, K_D, when:

a) a barren node, B, in K₁ corresponds exactly to a frontier node, F, in
K₂ (ie, nodes B and F represent the same distinction); and

b) F's probability distribution is copied directly from B's.

The concept of a disjoint knowledge map can be readily extended to involve more than two constituent knowledge maps with several corresponding nodes. It is important to note that F and B are not the same node and, hence, no statements of conditional independence should be inferred between nodes in K₁ and nodes in K₂.

Also, note that a disjoint knowledge map describes only a portion of the joint probability distribution over the set of outcomes defined by the union of all nodes in the disjoint knowledge map.

See also: Context and Expert.