Self-Organizing Binary Search Trees

We consider heuristics which attempt to maintain a binary search tree in a near optimal form, assuming that elements are requested with fixed, but unknown, independent probabilities. A "move to root" heuristic is shown to yield an expected search time within a constant factor of that of an optimal static binary search tree. On the other hand, a closely related "simple exchange" technique is shown not to have this property. The rate of convergence of the "move to root" heuristic is discussed. We also consider the more general case in which elements not in the tree may have non-zero probability of being requested.

Authors

• Brian Allen
• J. Ian Munro

Keywords

• Binary search trees
• Computer science
• Convergence
• Time measurement
• Performance evaluation
• Cost function
• Heuristic
• Convergence Rate
• Search Optimization
• Tree Search
• Binary Tree
• Binary Search
• Simple Exchange