J. Bell and G. Gupta. Evaluation of self-adjusting binary search tree techniques. Software Practice and Experience, 23(4):369--382, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....tree rotations after every access; they do not store any balance or weight information. From a theoretical standpoint, the splay tree is a beautiful structure, but the large number of tree rotations done after every access makes it less practically e#cient than even AVL trees in many applications [3]. These rotations also make splay trees a poor choice for augmentation with auxiliary structures at internal nodes. Bent, Sleator, and Tarjan [4] and Feigenbaum and Tarjan [9] design biased search trees for arbitrary weights that significantly reduce, but do not eliminate, the number of tree ....

J. Bell and G. Gupta. Evaluation of self-adjusting binary search tree techniques. Software Practice and Experience, 23(4):369--382, 1993.

....adjust themselves to achieve optimal amortized biased access times for access frequency weights. Splay trees store no balance or weight information, but they perform many tree rotations after every access, which makes them less practically ecient than even AVL trees in many applications [3]. These rotations can be particularly deleterious when nodes are augmented with auxiliary structures. Bent, Sleator, and Tarjan [4] and Feigenbaum and Tarjan [9] design biased search trees for arbitrary weights that signi cantly reduce, but do not eliminate, the number of tree rotations needed. ....

J. Bell and G. Gupta. Evaluation of self-adjusting binary search tree techniques. Soft. Prac. Exp., 23(4):369-382, 1993.

....of having a balanced tree structure in con trast to a BST could therefore offset the additional costs that incur at insertion time. Splay trees are another adaptive variant of BST. They have an amortised average cost of O(mlogn) over a sequence of m searches and insertions in a n node tree [2, 16]. The tree is modified at each access, by rotating the accessed node to the root through a series of double rotations. A double rotation re places a node with its grandparent and is efficiently implemented by storing an additional parent pointer in each node. We use non recursive bottom up ....

....accessed node to the root through a series of double rotations. A double rotation re places a node with its grandparent and is efficiently implemented by storing an additional parent pointer in each node. We use non recursive bottom up splaying instead of the top down splaying used in other work [2, 18], since we found that bottom up splaying is more efficient [20] Splaying leads to clustering of commonly accessed nodes towards the root, a beneficial feature for skew input. However, we have found in other work [11, 20] that the potential low expected search costs are offset by the costly ....

J. Bell and G.K. Gupta. An evaluation of self-adjusting binary search tree techniques. Software--Practice and Experience, 23(4):369- 382, 1993.

....than top down splaying. Another contrast was that Pugh s experiments used integer keys. If strings are used as keys, the cost of key comparison is a much greater component of total costs, outweighing savings that might be available. Our results are also in contrast to those of Gopal and Bell [4], whose comparative study of splay trees and other tree structures found that that bottom up splay trees are many times slower than BSTs and that top down splaying is faster than bottom up splaying. These results, however, were determined in a very di#erent test environment. The data set itself ....

J. Bell and G.K. Gupta. An evaluation of self-adjusting binary search tree techniques. Software---Practice and Experience, 23(4):369--382, 1993.

....of having a balanced tree structure in contrast to a BST could therefore o#set the additional costs that incur at insertion time. Splay trees are another adaptive variant of BST. They have an amortised average cost of O(m log n) over a sequence of m searches and insertions in a n node tree [2, 16]. The tree is modified at each access, by rotating the accessed node to the root through a series of double rotations. A double rotation replaces a node with its grandparent and is e#ciently implemented by storing an additional parent pointer in each node. We use non recursive bottom up splaying ....

....the accessed node to the root through a series of double rotations. A double rotation replaces a node with its grandparent and is e#ciently implemented by storing an additional parent pointer in each node. We use non recursive bottom up splaying instead of the top down splaying used in other work [2, 18], since we found that bottom up splaying is more e#cient [20] Splaying leads to clustering of commonly accessed nodes towards the root, a beneficial feature for skew input. However, we have found in other work [11, 20] that the potential low expected search costs are o#set by the costly ....

J. Bell and G.K. Gupta. An evaluation of self-adjusting binary search tree techniques. Software---Practice and Experience, 23(4):369-- 382, 1993.

....on disk) to be updated with minimal disk activity. The other candidate structures are varieties of search tree that maintain strings in sorted order. Tries [4, 16, 22, 29, 35] and ternary search trees [10, 12] are fast, but space intensive. Binary search trees and the major variants splay trees [8, 45], randomised search trees [30] and red black trees use much less space, but are slow. In this paper we propose a new data structure, which we call the burst trie. This structure is a collection of small data structures, which we call containers, that are accessed via a conventional trie, which ....

J. Bell and G.K. Gupta. An evaluation of self-adjusting binary search tree techniques. Software---Practice and Experience, 23(4):369--382, 1993.

....trees. For values within the range 7.5 25 the interval histogram needs to be examined. V. Related Work There has been a large amount of research on analyzing and improving the performance of search data structures both in terms of the operation complexity and the cache performance. Bell Gupta [7] conducted a evaluation of the self adjusting binary tree techniques. They compared three avors of the move to front technique, namely splaying, exchange and move to front with AVL and random binary search trees for di erent operation mixes and data distributions. They used synthetic benchmarks ....

J. Bell and G. Gupta. An evaluation of self-adjusting binary search tree techniques. Software: Practice and Experience, 23(4):369-82, Apr 1993.

.... of requiring more restructuring, especially during accesses (searches) This additional work often makes adaptive trees less practical than balanced trees; for instance, the performance of splay trees [21] is worse than that of AVL trees unless the pattern of accesses is very highly skewed [4]. High overhead, though, is not inherent to adaptation in general. For instance, the move tofront heuristic for sequential lists changes only a constant number of pointers for each operation, yet is very efficient in practice [7] To restructure a tree, we use the promotion operation (see Figure ....

J. Bell and G. Gupta. An evaluation of self-adjusting binary search tree techniques. Software Practice and Experience, 23:369--382, 1993.

....have been generalized to multiway and k ary search trees by Martel [54] and Sherk [63] Some empirical results on self adjusting trees and splay trees in particular have appeared. Moffat et al. 59] give evidence that sorting using splay trees is quite efficient. On the other hand, Bell and Gupta [11] give evidence that on random data that is not particularly skewed, selfadjusting trees are generally slower than standard balanced binary trees. There still remains a great deal of work to be done on empirical evaluation of selfadjusting trees on data drawn from typical real life applications. 4 ....

J. Bell and G. Gupta. Evaluation of self-adjusting binary search tree techniques. Software---Practice & Experience, 23(4):369--382, April 1993.

No context found.

J. Bell and G. Gupta. Evaluation of self-adjusting binary search tree techniques. Software Practice and Experience, 23(4):369--382, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC