Fibonacci heaps and their uses in improved network optimization algorithms

Reviewer: Hausi Albert Muller

The authors of this paper define a collection of new data structures called Fibonacci heaps or F-heaps for implementing heaps or priority queues. F-heaps support the delete and delete min operations in O(log n) amortized time and all other standard heap operations in O(1) amortized time (i.e., make heap, insert, find min, meld, decrease key). As stated in the paper's abstract, F-heaps are a direct extension of binomial queues, which support all the heap operations in O(log n) worst-case time. A Fibonacci heap is a set of item-disjoint heap-ordered trees (i.e., the root node contains the minimum key, and the key of any other node is no less than the key of its parent). The structure of the tree is defined by the operations performed on the tree: a node with k children has at least F k+2 descendants including itself, where F 1=1; F 2=1; F k= F k-1+ F k-2; thus, the minimum number of nodes of a tree in an F-heap is a Fibonacci number, hence, the name Fibonacci heaps. Dijkstra uses a similar type of trees, Leonardo trees ( L 1=1; L 2=1; L k= L k?1+ L:- 0Ik?2+1) in his Smoothsort algorithm [1], which is based on Williams's Heapsort algorithm; however, Smoothsort or Leonardo trees are not mentioned in the paper. The paper shows how Fibonacci heaps can be used to obtain improved running times for network optimization algorithms such as Dijkstra's shortest paths and Prim's minimum spanning tree. The new worst-case bounds for graphs of intermediate density ( n:3WiCm:3WiCm 2) are listed in the following table, where n is the number of vertices, m is the number of edges, and &bgr;( m, n)=min{ i &vbm0; log ( i) n?Cm- / n}. :.OC :.HB Single-source shortest path All-pairs shortest path Weighted bipartite matching Minimum spanning tree :.HS :.HS O( n log n+ m) O( n 2 log n+ nm) O( n 2 log n+ nm) O( m &bgr;( m,n)) :.HT :.OE The paper concludes by stating three open problems raised by this research and by listing additional applications of Fibonacci heaps that were explored recently by Gabow, Galil, and Spencer [2]. :.OC :.HB Minimum spanning tree Shortest pair of disjoint paths Optimum branching :.HS :.HS O( m log &bgr;( m,n)) O( n log n+ m) O( n log n+ m) :.HT :.OE F-heaps could be presented in any computer science course discussing algorithms and data structures. The material presented in this well-written paper will certainly find its way into textbooks. Fibonacci heaps are an outstanding research contribution in the area of network optimization algorithms and data structures.