IMPROVED Decision Network ADAPTATION Method Using BOTTOM Treegraph Extension
Original Publication Date: 1983-Dec-01
Included in the Prior Art Database: 2005-Feb-08
Bottom extension is a sequential adaptation procedure that produces decision networks. In the prior art, the method was implemented as follows. Given a network classifier satisfying (1) and (2) for classes 1,2,...,N, and a new class, N+1, with associated model GN+1(f), the latter is first evaluated for f=F1, the decision function at the root node. If GN+1 (F1)=0 or 1, then the branch having the corresponding label leads to a successor node, which is recorded in a stack. If GN+1(F1)='z', then both branches are traversed, and the two successor nodes are stacked if they are both interior nodes. At the next step a node is removed from the stack and the above process repeated. The process of stacking the successor nodes continues until a terminal node is encountered.