Catálogo de publicaciones - libros

Many approaches to production scheduling are based on the step by step improvement of a given schedule. For these, finding a “good” initial schedule may critically influence performance. In the following, an extremely fast and simple method is presented for finding such a schedule in the case of a continuous-time single machine problem with the goal of minimizing setup and holding costs. In the case where there are no setups, this method always finds the optimal solution. A benchmark illustrates the possibilities of the method.

After the detection of the first extrasolar planet (exoplanet) more than one decade ago we currently know about more than 200 planets around other stars and there are about twenty multi-planet systems. Today’s most commonly used technique for identifying such exoplanets is based on radial velocity measurements of the star. Due to the increasing time span and accuracy of the respective observations, the measured data samples will more and more contain indications of multiple planetary companions. Unfortunately, especially in multi-planet systems, the determination of these exoplanets’ orbital elements is a rather complex and computationally expensive data analysis task. It is usually formulated as an optimization problem in which the orbital parameters yielding minimal residues w.r.t. the measured data are to be found. Very likely, improved algorithms for (approximately) solving this problem will enable the detection of more complex systems. In this paper we describe a specialized evolution strategy for approaching this problem.

Receiver operating characteristic (ROC) curves are a plot of a ranking classifier’s true-positive rate versus its false-positive rate, as one varies the threshold between positive and negative classifications across the continuum. The area under the ROC curve offer a measure of the discriminatory power of machine learning algorithms that is independent of class distribution, via its equivalence to Mann-Whitney U-statistics. This measure has recently been extended to cover problems of discriminating three and more classes. In this case, the area under the curve generalizes to the volume under the ROC surface.

In this paper, we show how a multi-class classifier can be trained by directly maximizing the volume under the ROC surface. This is accomplished by first approximating the discrete U-statistic that is equivalent to the volume under the surface in a continuous manner, and then maximizing this approximation by gradient ascent.

Model transformation has applications in many areas and usually is to certain properties. In model-driven software development, e.g., abstract models are transformed into more specific ones while preserving the behavior of the overall specified system. In this paper we demonstrate how constraint-based preservation, which we have initially developed in the context of digital archiving, can be applied to systems specification. We demonstrate how our approach helps to preserve communication flows inside a component-based system. In particular, formal allow us to prove invariance of certain significant properties of a component interaction specification if the transformation process adheres to a given set of such constraints.

Automated modelling is of increasing relevance in empirical social sciences because of the increasing availability of potentially important variables. The availability of many variables causes uncertainty about which variables should be included in parsimonious models for the explanation of phenomena in social sciences and which variables should be excluded. Given a large number of potentially informative variables this paper argues that the use of genetic algorithms for Bayesian model selection allows the efficient automated identification of an optimal subset of variables. The advantages of using a genetic algorithm as a method for automated modelling is exemplified by the identification of previously unknown but important causal relationships for long-run inflation, the share spent on defence and political rights on the basis of a cross-country data set.

In real-world very large scale integration (VLSI) silicon implementations of 3GPP wireless modems modularity of system on chip (SoC) architecture and efficiency in terms of power and die size are crucial criteria. This is particularly true for the signal processing subsystem as the data flow volume in the modem’s physical layer is proportionally high.

One architectural method to deal with modularity and power efficiency is to focus on well defined communication structures inbetween the dedicated signal processing units and well formatted data objects stored locally in intermediate buffers to keep data transport effort low. Contrary to that, die size reduction often implies the usage of centralized and shared entities like memories and busses.

This paper presents an example for data communication structure and data format of the radio frame buffers in a UMTS/HSDPA/MBMS wireless modem receive path and how an object oriented system design approach combines those optimization strategies according to the initially mentioned criteria.

Low Density Parity Check (LDPC) codes are considered to be used in many future communication systems. We will show in this work how the performance of a communication system is influenced when different bit to symbol mapping schemes of a higher order modulation (e.g.: 16-QAM) are combined with LDPC coding. To evaluate and optimize bit-to-symbol mapping methods, a modified version of density evolution is proposed. In addition, simulation results are shown which justify the use of density evolution as a method for qualitative analysis of different symbol mapping schemes in conjunction with LDPC coding. We propose novel mapping methods and show performance results in combination with LDPC coding.

Ultrasound imaging segmentation is a common method used to help in the diagnosis in multiple medical disciplines. This medical image modality is particularly difficult to segment and analyze since the quality of the images is relatively low, because of the presence of speckle noise. In this paper we present a set of techniques, based on texture findings, to increase the quality of the images. We characterize the ultrasound image texture by a vector of responses to a set of Gabor filters. Also, we combine front-propagation and active contours segmentation methods to achieve a fast accurate segmentation with the minimal expert intervention.

This paper presents a basic concept, development and implementation of software tool designated for programming of Field Programmable Gate Arrays. However, the programming scheme does not employ typical VHDL or Verilog languages but dedicated PLC-like software environment producing both execution code and hardware configuration for the for FPGA chip. As a result the project has created PLC language defined control system implemented in an FPGA structure.

Verification tests were performed on Xilinx Spartan II platform and finally were converted into Spartan III chip family. Although verification tests were run using development boards the real life embedded application is now developed.

Multimode Mobile Terminals require flexible signal processing architectures to support different mobile standards like GSM/EDGE and UMTS/HSDPA in one platform with minimum hardware resources. In this paper, an efficient architecture is presented using same system clock frequency for UMTS and GSM signal processing and an all-digital asynchronous sampling rate converter. In addition, it is shown how this architecture can be efficiently used to solve the problem of timing adjustments in time continuous signals of the UMTS/HSDPA receive and transmit path. The traditional and the proposed approaches are illustrated.