In the design of algorithms, partition refinement is a technique for representing a partition of a set as a data structure that allows the partition to be refined by splitting its sets into a larger number of smaller sets. In that sense it is dual to the union-find data structure, which also maintains a partition into disjoint sets but in which the operations merge pairs of sets together.
Partition refinement forms a key component of several efficient algorithms on graphs and finite automata, including DFA minimization, the Coffman–Graham algorithm for parallel scheduling, and lexicographic breadth-first search of graphs.
A partition refinement algorithm maintains a family of disjoint sets Si. At the start of the algorithm, this family contans a single set of all the elements in the data structure. At each step of the algorithm, a set X is presented to the algorithm, and each set Si in the family that contains members of X is split into two sets, the intersection Si ∩ X and the difference Si \ X.
Such an algorithm may be implemented efficiently by maintaining data structures representing the following information:array, ordered by the sets they belong to, and sets may be represented by start and end indices into this array.
- The ordered sequence of the sets Si in the family, in a form such as a doubly linked list that allows new sets to be inserted into the middle of the sequence
- Associated with each set Si, a collection of its elements of Si, in a form such as a doubly linked list or array data structure that allows for rapid deletion of individual elements from the collection. Alternatively, this component of the data structure may be represented by storing all of the elements of all of the sets in a single array, sorted by the identity of the set they belong to, and by representing the collection of elements in any set Si by its starting and ending positions in this array.
- Associated with each element, the set it belongs to.
To perform a refinement operation, the algorithm loops through the elements of the given set X. For each such element x, it finds the set Si that contains x, and checks whether a second set for Si ∩ X has already been started. If not, it creates the second set and add Si to a list L of the sets that are split by the operation. Then, regardless of whether a new set was formed, the algorithm removes x from Si and adds it to Si ∩ X. In the representation in which all elements are stored in a single array, moving x from one set to another may be performed by swapping x with the final element of Si and then decrementing the end index of Si and the start index of the new set. Finally, after all elements of X have been processed in this way, the algorithm loops through L, separating each current set Si from the second set that has been split from it, and reports both of these sets as being newly formed by the refinement operation.
The time to perform a single refinement operations in this way is O(|X|), independent of the number of elements in the famliy of sets and also independent of the total number of sets in the data structure. Thus, the time for a sequence of refinements is proportional to the total size of the sets given to the algorithm in each refinement step.
An early application of partition refinement was in an algorithm by Hopcroft (1971) for DFA minimization. In this problem, one is given as input a deterministic finite automaton, and must find an equivalent automaton with as few states as possible. Hopcroft's algorithm maintains a partition of the states of the input automaton into subsets, with the property that any two states in different subsets must be mapped to different states of the output automaton. Initially, there are two subsets, one containing all the accepting states of the automaton and one containing the remaining states. At each step one of the subsets Si and one of the input symbols x of the automaton are chosen, and the subsets of states are refined into states for which a transition labeled x would lead to Si, and states for which an x-transition would lead somewhere else. When a set Si that has already been chosen is split by a refinement, only one of the two resulting sets (the smaller of the two) needs to be chosen again; in this way, each state participates in the sets X for O(s log n) refinement steps and the overall algorithm takes time O(ns log n), where n is the number of initial states and s is the size of the alphabet.
Partition refinement was applied by Sethi (1976) in an efficient implementation of the Coffman–Graham algorithm for parallel scheduling. Sethi showed that it could be used to construct a lexicographically ordered topological sort of a given directed acyclic graph in linear time; this lexicographic topological ordering is one of the key steps of the Coffman–Graham algorithm. In this application, the elements of the disjoint sets are vertices of the input graph and the sets X used to refine the partition are sets of neighbors of vertices. Since the total number of neighbors of all vertices is just the number of edges in the graph, the algorithm takes time linear in the number of edges, its input size.
Partition refinement also forms a key step in lexicographic breadth-first search, a graph search algorithm with applications in the recognition of chordal graphs and several other important classes of graphs. Again, the disjoint set elements are vertices and the set X represent sets of neighbors, so the algorithm takes linear time.
- Paige, Robert; Tarjan, Robert E. (1987), "Three partition refinement algorithms", SIAM Journal on Computing 16 (6): 973–989, doi:10.1137/0216062, MR 917035.
- Habib, Michel; Paul, Christophe; Viennot, Laurent (1999), "Partition refinement techniques: an interesting algorithmic tool kit", International Journal of Foundations of Computer Science 10 (2): 147–170, doi:10.1142/S0129054199000125, MR 1759929.
- Habib, Michel; Paul, Christophe; Viennot, Laurent (1998), "A synthesis on partition refinement: a useful routine for strings, graphs, Boolean matrices and automata", STACS 98 (Paris, 1998), Lecture Notes in Computer Science 1373, Springer-Verlag, pp. 25–38, doi:10.1007/BFb0028546, MR 1650757.
- Valmari, Antti; Lehtinen, Petri (2008). "Efficient minimization of DFAs with partial transition functions". In Albers, Susanne; Weil, Pascal. 25th International Symposium on Theoretical Aspects of Computer Science (STACS 2008). Leibniz International Proceedings in Informatics (LIPIcs). Dagstuhl, Germany: Schloss Dagstuhl: Leibniz-Zentrum fuer Informatik. pp. 645–656. doi:10.4230/LIPIcs.STACS.2008.1328. ISBN 978-3-939897-06-4. ISSN 1868-8969..
- Knuutila, Timo (2001). "Re-describing an algorithm by Hopcroft". Theoretical Computer Science 250 (1-2): 333–363. doi:10.1016/S0304-3975(99)00150-4. ISSN 0304-3975.
- Hopcroft, John (1971), "An n log n algorithm for minimizing states in a finite automaton", Theory of machines and computations (Proc. Internat. Sympos., Technion, Haifa, 1971), New York: Academic Press, pp. 189–196, MR 0403320.
- Sethi, Ravi (1976), "Scheduling graphs on two processors", SIAM Journal on Computing 5 (1): 73–82, doi:10.1137/0205005, MR 0398156.
- Rose, D. J.; Tarjan, R. E.; Lueker, G. S. (1976), "Algorithmic aspects of vertex elimination on graphs", SIAM Journal on Computing 5 (2): 266–283, doi:10.1137/0205021.
- Corneil, Derek G. (2004), "Lexicographic breadth first search – a survey", Graph-Theoretic Methods in Computer Science, Lecture Notes in Computer Science 3353, Springer-Verlag, pp. 1–19.