bsp trees and KD trees are fundamental data structures for collision detection in walkthrough environments. A basic issue in the construction of these hierarchical data structures is the choice of cutting planes. Rather than base these choices solely on the properties of the scene, we propose using information about how the tree is used in order to determine its structure. We demonstrate how this leads to the creation of bsp trees that are small, do not require much preprocessing time, and respond very efficiently to sequences of collision queries. Categories and Subject Descriptors (according to ACM CCS): I.3.5 [Computer Graphics]: Computational Geometry and Object Modeling I.3.6 [Computer Graphics]: Graphics data structures and data types, Interaction techniques I.3.7 [Computer Graphics]: Virtual reality

The ability to perform efficient collision detection is essential in virtual reality environments and their applications, such as walkthroughs. In this paper we re-explore a classical structure used for collision detection – the binary space partitioning tree. Unlike the common approach, which attributes equal likelihood to each possible query, we assume events that happened in the past are more likely to happen again in the future. This leads us to the definition of self-customized data structures. We report encouraging results obtained while experimenting with this concept in the context of self-customized BSP trees.

We consider a variant of the traditional task of explicitly reconstructing algebraic functions from black box representations. In the traditional setting for such problems, one is given access to an unknown function f that is represented by a black box, or an oracle, which can be queried for the value of f at any input. Given a guarantee that this unknown function f is some nice algebraicfunction, say a polynomial in its input of degree bound d, the goal of the reconstruction problem is to explicitly determine the coefficients of the unknown polynomial. All work on polynomial interpolation, especially sparse ones, are or may be presented in such a setting. The work of Kaltofen and Trager [Computing with polynomials given by black boxes for their evaluations: Greatest common divisors, factorization, separation of numerators and denominators, in Proc. 29th Ann. IEEE Symp. on Foundations of Computer Science, 1988, pp. 296--305], for instance, highlights the utility of this setting, by performing numerous manipulations on polynomials presented as black boxes. The variant considered in this paper differs from the traditional setting in that our black boxes represent several algebraic functions f 1 ,...,f k , where at each input x, the box arbitrarily chooses a subset of f 1 (x),...,f k (x) to output and we do not know which subset it outputs. We show how to reconstruct the functions f 1 ,...,f k from the black box, provided the black box outputs according to these functions "often." This allows us to group the sample points into sets, such that for each set, all outputs to points in the set are from the same algebraic function. Our methods are robust in the presence of a small fraction of arbitrary errors in the black box. Our model and techniques can be applied in the areas of computer vision, machine learning, curve fitting and polynomial approximation, self-correcting programs, and bivariate polynomial factorization.

This paper provides the first systematic investigation of checking approximate numerical computations over subsets of the reals. In most cases, approximate checking is more challenging than exact checking. Problem conditioning, i.e., the measure of sensitivity of the output to slight changes in the input, and the presence of approximation parameters foil the direct transformation of many exact checkers to the approximate setting. Furthermore, approximate checking over the reals is complicated by the lack of nice finite field properties such as the existence of a samplable distribution which is invariant under addition or multiplication by a scalar. We overcome the above problems by using such techniques as testing and checking over similar but distinct distributions, using functions' random and downward self-reducibility properties, and taking advantage of the small variance of the sum of independent identically distributed random variables. We provide approximate checkers for a variety of computations, including matrix multiplication, linear system solution, matrix inversion, and computation of the determinant. We also present an approximate version of Beigel's trick and extend the approximate linear self tester/corrector of [8] and the trigonometric self-Tester/corrector of [5] to more general computations.

The authors consider the task of reconstructing algebraic functions given by black boxes. Unlike traditional settings, they are interested in black boxes which represent several algebraic functions-f/sub 1/, . . ., f/sub k/, where at each input x, the box arbitarrily chooses a subset of f/sub 1/(x), . . ., f/sub k/(x) to output. They show how to reconstruct the functions f/sub 1/,. . ., f/sub k/ from the black box. This allows them to group the same points into sets, such that for each set, all outputs to points in the set are from the same algebraic function. The methods are robust in the presence of errors in the black box. The model and techniques can be applied in the areas of computer vision, machine learning, curve fitting and polynomial approximation, self-correcting programs and bivariate polynomial factorization.>