Showing papers on "Bounding overwatch published in 2001"

Not) Bounding the True Error

Abstract: We present a new approach to bounding the true error rate of a continuous valued classifier based upon PAC-Bayes bounds. The method first constructs a distribution over classifiers by determining how sensitive each parameter in the model is to noise. The true error rate of the stochastic classifier found with the sensitivity analysis can then be tightly bounded using a PAC-Bayes bound. In this paper we demonstrate the method on artificial neural networks with results of a 2 - 3 order of magnitude improvement vs. the best deterministic neural net bounds.

Dual bounding procedures lead to convergent Branch–and–Bound algorithms

Abstract: Branch–and–Bound methods with dual bounding procedures have recently been used to solve several continuous global optimization problems. We improve results on their convergence theory and give a condition that enables us to detect infeasible partition sets from the dual optimal value.

Linear Interval Estimations for Parametric Objects Theory and Application

Abstract: The new concept of parametrized bounding volumes for parametric objects is proposed to replace the common compact bounding volumes like axis aligned bounding boxes and parallelepipeds. Linear Interval Estimations (LIEs) are developed as a realization of the discussed ideas. Two reliable methods for the computation of LIEs are introduced based on a new understanding of the use of affine arithmetics and a special application of Taylor Models. The particular structure of LIEs allows an effective intersection test of LIEs with rays, boxes and other LIEs. The test gives besides of a possible location of the intersection in object space information about affected parts in the parameter spaces of the enclosed objects. A subdivision algorithm for the intersection of two parametric surface patches with remarkable experimental results is presented as a possible application.

On the signal bounding problem in timing analysis

Abstract: In this paper, we study the propagation of slew dependent bounding signals and the corresponding slew problem in static timing analysis. The selection of slew from the latest arriving signal, a commonly used strategy, may violate the rule of monotonic delay. Several methods for generating bounding signals to overcome this difficulty are described. The accuracy and monotonicity of each method is analyzed. These methods can be easily implemented in a static timer to improve the accuracy.

Bounding On-Off Sources -- Variability Ordering and Majorization to the Rescue

Abstract: We consider the problem of bounding the loss rates of the aggregation of on-off sources in a bufferless model by the loss rates associated with the aggregation of i.i.d. on-off sources. We use well known results from the theory of variability orderings to establish a conjecture of Rasmussen et al., a recent upper bound of Mao and Habibi, and to discuss a new conjectured upper bound by these authors.

Spectral Implications of Security Market Data for Models of Dynamic Economies

Abstract: Hansen and Jagannathan (1991) proposed a volatility bound for evaluating asset-pricing models that is a restriction on the volatility of a representative agentIs intertemporal marginal rate of substitution (IMRS). We develop a generalization of their bound that (i) incorporates the serial correlation properties of return data and (ii) allows us to calculate a spectral version of the bound. That is, we develop a bound and then decompose it by frequency; this enables us to judge whether models match important aspects of the data in the long run, at business cycle frequencies, seasonal frequencies, etc. Our generalization is related to the space in which the bounding IMRS lives. Instead of specifying the bounding IMRS to be a linear combination of contemporaneous returns, we let the bounding IMRS live in a linear space of current, past and future returns. We also require the bounding IMRS to satisfy additional restrictions that resemble Euler equations. Our volatility bound not only uses the unconditional first and second moment properties of return data but also the serial correlation properties. Incorporating this additional information results in a tighter bound for two reasons. First, we impose additional orthogonality conditions on our bounding IMRS. Second, our projection is onto a larger space (current, past and future returns). We also show that the spectrum of the model IMRS must exceed the spectrum of our bounding IMRS. Using the serial correlation properties of returns (together with the mean and variance), we are able to derive the spectrum of the bounding IMRS. That is, the lower bound on the spectrum of the model IMRS is completely pinned down by asset return data. This permits a frequency-by-frequency examination of the fundamental component of the model, namely, the Euler equation that links asset returns to the IMRS. In particular, we can identify the frequencies at which an asset-pricing model does not perform well. The researcher can then decide whether or not failures at a particular set of frequencies are troubling. We illustrate our method with four asset pricing models -- time-separable CRRA preferences, state non-separable preferences (Epstein-Zin, 1989, 1991), internal habit formation (Constantinides, 1990), and external habit formation preferences (Campbell and Cochrane, 1999) -- using two data sets, annual data from 1889-1992 and quarterly data spanning 1950:1-1995:4.

Research on Collision Detection for Rigid Object in Deformable Environment

Abstract: Collision Detection for rigid object in deformable environment is very popular in the field of virtual reality such as surgery simulation and so on, but few research has been made. This paper proposes a collision detection method based on fixed direction hull bounding volume tree which provides an effective way for such problem. Fixed direction hulls is a special convex hull whose outward normal of facets comes from a fixed direction set. It overcomes limitations of other bounding volumes and makes a promise between tightness and simpleness. It can be determined that two fixed direction hull bounding volumes do not overlap by checking if one of their bound intervals on direction axis defined by the fixed direction set does not overlap. Thus the intersection test between two bounding volumes need only at most n comparisons ( n is the size of fixed direction set).Updating bounding volumes after rotation is an important problem of hierarchical bounding volume approach. An updating algorithm based on linear programming is proposed in this paper based on the definition and property of fixed direction hulls. This method need not add any additional computation and memory during building bounding volume hierarchies. It can update a bounding volume through 3 n multiplications. Deformation of object is a difficulty in collision detection. This paper analyses two kinds of deformations and proposes resolving method respectively. A bottom up updating method is proposed especially which can compute the bounding volume of parent node through bounding volumes of the two children only by n comparisons.It is proved through experiment that our method can not only solve collision detection between rigid objects efficiently but also solve collision detection between rigid objects and deformable objects.

Laufzeitbestimmung von Realzeitsoftware (Bounding the Execution Time of Software in Realtime Systems)

Abstract: Die Einsatzfelder für Realzeit-Systeme gewinnen zunehmend an Komplexität. Um den gewachsenen Aufgaben gerecht zu werden, wird vermehrt auf leistungsfähige Standardkomponenten für Soft- und Hardware zurückgegriffen. Zentraler Punkt für den Nachweis der Einhaltung der Zeitbedingungen ist die Kenntnis der Ausführungszeiten der Software. Da aufgrund der rasanten Entwicklung in der Hardware eine detailgetreue Modellierung und Simulatorentwicklung zur Bestimmung der Ausführungszeiten praktisch nicht mehr durchzuführen ist, wird ein Ansatz vorgestellt, durch fundierte Messungen zu einer sicheren und realistischen Abschätzung der Ausführungszeit für den ungünstigsten Fall zu kommen. Dabei wird nicht nur die Applikation an sich, sondern auch der Einfluß eines Realzeit-Betriebssystems betrachtet.