Figure 3
Multiple sub-solutions. The tree structure allows us to save the information in an efficient way. Each sub-solution corresponds to a subtree and there is no need to duplicate it each time it appears in a solution. In particular, only one node is created for each association and if two different alternatives share this association, the respective (square) nodes will point exactly at the same (circular) node. In this example, the mapping of p into h has the same alternatives (i), (ii) and (iii) as depicted in Figure 2. The association of p with h ′ has local minimum cost of c ′ and can be obtained by two mappings of p 1 and p 2: (iv) p 1 is mapped into h and p 2 is mapped into h z ; and (v) p 1 is mapped into h u and p 2 is mapped into h v . Notice that, h, h ′, h u , h v , h w , h x , h y , and h z are distinct nodes of the host tree.