Optimal Decomposition Strategy For Tree Edit Distance

An ordered labeled tree is a tree where the left-to-right order among siblings is significant. Given two ordered labeled trees, the edit distance between them is the minimum cost edit operations that convert one tree to the other. In this thesis, we present an algorithm for the tree edit distance problem by using the optimal tree decomposition strategy. By combining the vertical compression of trees with optimal decomposition we can significantly reduce the running time of the algorithm. We compare our method with other methods both theoretically and experimentally. The test results show that our strategies on compressed trees are by far the best decomposition strategy, creating the least number of relevant sub-problems.