New Lower Bounds on the Cost of Binary Search Trees

Abstract In this paper we provide new lower bounds on the cost C of binary search trees. The bounds are expressed in terms of the entropy H of the probability distribution, the number of elements and the probability that a search is successful. Most of our lower bounds are derived by means of a new technique which exploits the relation between trees and codes. Our lower bounds compare favorably with known limitations. We also provide an achievable upper bound on the Kraft sum generalized to the internal nodes of a tree. This improves on a previous result.