We present a novel connection between binary search trees (BSTs) and points in the plane satisfying a simple property. Using this correspondence, we achieve the following results: 1. A surprisingly clean restatement in geometric terms of many results and conjectures relating to BSTs and dynamic optimality. 2. A new lower bound for searching in the BST model, which subsumes the previous two known bounds of Wilber [FOCS’86]. 3. The first proposal for dynamic optimality not based on splay trees. A natural greedy but offline algorithm was presented by Lucas [1988], and independently by Munro [2000], and was conjectured to be an (additive) approximation of the best binary search tree. We show that there exists an equal-cost online algorithm, transforming the conjecture of Lucas and Munro into the conjecture that the greedy algorithm is dynamically optimal. 1

We present the first spatially adaptive data structure that answers approximate nearest neighbor (ANN) queries to points that reside in a geometric space of any constant dimension d. The Lt-norm approximation ratio is O(d 1+1/t), and the running time for a query q is O(d 2 lg δ(p, q)), where p is the result of the preceding query and δ(p, q) is the number of input points in a suitably-sized box containing p and q. Our data structure has O(dn) size and requires O(d 2 n lg n) preprocessing time, where n is the number of points in the data structure. The size of the bounding box for δ depends on d, and our results rely on the Random Access Machine (RAM) model with word size Θ(lg n). 1

Abstract. We present skip-splay, the first binary search tree algorithm known to have a running time that nearly achieves the unified bound. Skip-splay trees require only O(m lg lg n + UB(σ)) time to execute a query sequence σ = σ1...σm. The skip-splay algorithm is simple and similar to the splay algorithm. 1 Introduction and Related Work Although the worst-case access cost for comparison-based dictionaries is Ω(lg n), many sequences of operations are highly nonrandom, allowing tighter, instancespecific running time bounds to be achieved by algorithms that adapt to the input sequence. Splay trees [1] are an example of such an adaptive algorithm

Sleator and Tarjan [39] conjectured that splay trees are dynamically optimal binary search trees (BST). In this context, we study the skip list data structure introduced by Pugh [35]. We prove that for a class of skip lists that satisfy a weak balancing property, the working-set bound is a lower bound on the time to access any sequence. Furthermore, we develop a deterministic self-adjusting skip list whose running time matches the working-set bound, thereby achieving dynamic optimality in this class. Finally, we highlight the implications our bounds for skip lists have on multi-way branching search trees such as B-trees, (ab)-trees, and other variants as well as their binary tree representations. In particular, we show a self-adjusting B-tree that is dynamically optimal both in internal and external memory.

The working-set bound [Sleator and Tarjan, J. ACM, 1985] roughly states that searching for an element is fast if the element was accessed recently. Binary search trees, such as splay trees, can achieve this property in the amortized sense, while data structures that are not binary search trees are known to have this property in the worst case. We close this gap and present a binary search tree called a layered working-set tree that guarantees the working-set property in the worst case. The unified bound [Bădoiu et al., TCS, 2007] roughly states that searching for an element is fast if it is near (in terms of rank distance) to a recently accessed element. We show how layered working-set trees can be used to achieve the unified bound to within a small additive term in the amortized sense while maintaining in the worst case an access time that is both logarithmic and within a small multiplicative factor of the working-set bound. 1

views and conclusions contained in this document are those of the author and should not be interpreted as representing the official policies, either expressed or implied, of any sponsoring institution, the U.S. government or any other entity. Keywords: binary search trees, adaptive algorithms, splay trees, Unified Bound, dynamic A ubiquitous problem in the field of algorithms and data structures is that of searching for an element from an ordered universe. The simple yet powerful binary search tree (BST) model provides a rich family of solutions to this problem. Although BSTs require Ω(lg n) time per operation in the worst case, various adaptive BST algorithms are capable of exploiting patterns in the sequence of queries to achieve tighter, input-sensitive, bounds that can be o(lg n) in many cases. This thesis furthers our understanding of what is achievable in the BST model along two directions. First, we make progress in improving instance-specific lower bounds in the BST model. In particular, we introduce a framework for generating lower bounds on the cost that any BST algorithm must pay to execute a query sequence,

Abstract. We introduce the rank-sensitive priority queue — a data structure that always knows the minimum element it contains, for which insertion and deletion take O(log(n/r)) time, with n being the number of elements in the structure, and r being the rank of the element being inserted or deleted (r = 1 for the minimum, r = n for the maximum). We show how several elegant implementations of rank-sensitive priority queues can be obtained by applying novel modifications to treaps and amortized balanced binary search trees, and we show that in the comparison model, the bounds above are essentially the best possible. Finally, we conclude with a case study on the use of rank-sensitive priority queues for shortest path computation. 1

Abstract. In this paper we present an implicit dictionary with the working set property i.e. a dictionary supporting insert(e), delete(x) and predecessor(x) in O(log n) time and search(x) in O(log ℓ) time, where n is the number of elements stored in the dictionary and ℓ is the number of distinct elements searched for since the element with key x was last searched for. The dictionary stores the elements in an array of size n using no additional space. In the cache-oblivious model the operations insert(e), delete(x) and predecessor(x) cause O(log B n) cache-misses and search(x) causes O(log B ℓ) cache-misses. 1