Saladi, Rahul2017-11-272017-11-272017-09https://hdl.handle.net/11299/191453University of Minnesota Ph.D. dissertation. September 2017. Major: Computer Science. Advisor: Ravi Janardan. 1 computer file (PDF); xi, 126 pages.The focus of this thesis is the topic of geometric intersection queries (GIQ) which has been very well studied by the computational geometry community and the database community. In a GIQ problem, the user is not interested in the entire input geometric dataset, but only in a small subset of it and requests an informative summary of that small subset of data. Formally, the goal is to preprocess a set A of n geometric objects into a data structure so that given a query geometric object q, a certain aggregation function can be applied efficiently on the objects of A intersecting q. The classical aggregation functions studied in the literature are reporting or counting the objects of A intersecting q. In many applications, the same set A is queried several times, in which case one would like to answer a query faster by preprocessing A into a data structure. The goal is to organize the data into a data structure which occupies a small amount of space and yet responds to any user query in real-time. In this thesis the study of the GIQ problems was conducted from the point-of-view of a computational geometry researcher. Given a model of computation and a GIQ problem, what are the best possible upper bounds (resp., lower bounds) on the space and the query time that can be achieved by a data structure? Also, what is the relative hardness of various GIQ problems and aggregate functions. Here relative hardness means that given two GIQ problems A and B (or, two aggregate functions f(A, q) and g(A, q)), which of them can be answered faster by a computer (assuming data structures for both of them occupy asymptotically the same amount of space)? This thesis presents results which increase our understanding of the above questions. For many GIQ problems, data structures with optimal (or near-optimal) space and query time bounds have been achieved. The geometric settings studied are primarily orthogonal range searching where the input is points and the query is an axes-aligned rectangle, and the dual setting of rectangle stabbing where the input is a set of axes-aligned rectangles and the query is a point. The aggregation functions studied are primarily reporting, top-k, and approximate counting. Most of the data structures are built for the internal memory model (word-RAM or pointer machine model), but in some settings they are generic enough to be efficient in the I/O-model as well.enaggregationdata structuresgeometric algorithmsquery timerange searchingsize of the data structureThesis or Dissertation