Programming really complex and flexible parallel process systems in the Helios(TM) environment forces the programmer to work on the error-prone and complicated low level of socket communication. We have developed a library which provides the main features of the well-known CSP paradigm for usage in combination with standard C programming under Helios(TM). The overall functionality and the essentials of the implementation concepts for the MCSP library are described in this paper.

Die vorliegende Arbeit beschreibt den Funktionsumfang einer Programmbibliothek, die das Programmiermodell der kommunizierenden sequentiellen Prozesse (CSP) in allen wichtigen Aspekten inclusive bedingter Kommunikation und verteilter Terminierung für die Programmierung von Transputern in ANSI-C unter Helios realisiert. Das grundlegende CSP Modell wird dabei um dynamische Aspekte wie die Kreierung von Prozessen und Kommunikationsverbindungen zur Laufzeit erweitert.

3D applications are built on top of procedural low-level graphics packages which are difficult to learn and to use because of their inherent complexity and their renderer oriented design. We present a fine-grained object oriented model which views 3D graphics from the developer’s perspective. Our approach is based on a logical decomposition of the elements of 3D graphics into three major classes: Geometric primitives define shapes and their geometry. Rendering attributes specify quality and appearance of primitives and of the rendering process. Virtual rendering devices process attributes and primitives through a set of generic rendering commands for different types of rendering techniques and packages. Virtual rendering devices encapsulate the functionality of most of today’s graphics packages making them exchangeable even at runtime without the need to recode the application. We have implemented our concepts in VRS, the Virtual Rendering System, as a portable C++ toolkit. Currently we have integrated the standard graphics packages OpenGL, PEX, XGL, and Radiance.

The $\theta$-subsumption problem is crucial to the efficiency of ILP learning systems. We discuss two $\theta$-subsumption algorithms based on strategies for preselecting suitable matching literals. The class of clauses, for which subsumption becomes polynomial, is a superset of the deterministic clauses. We further map the general problem of $\theta$-subsumption to a certain problem of finding a clique of fixed size in a graph, and in return show that a specialization of the pruning strategy of the Carraghan and Pardalos clique algorithm provides a dramatic reduction of the subsumption search space. We also present empirical results for the mesh design data set.

A new approach to support the development of distributed software systems is outlined. As far as static aspects are concerned, standard object-oriented design methods are used. Dynamic aspects like status of services or resource handling in objects and methods, however, play the key role in distributed systems. The main contribution of the approach is a method which puts the focus on these aspects right from the beginning of a design. A new formalism based on extended Petri-Nets -- so called Object Coordination Nets (OCoNs) -- is used to describe the behavior of an object on a per method basis and for the usage of object resources. The formalism is integrated into higher levels of object-oriented design as an additional means to describe dynamic behavior. Integration with a standard sequential object-oriented language like C++ is used to organize method nets in a class context. This provides a type system for the nets as well. The sequential language can be interpreted as filling the computational gaps in the coordination skeleton provided by the OCoN for a method.

In order to rank the performance of machine learning algorithms, many researchs conduct experiments on benchmark datasets. Since most learning algorithms have domain-specific parameters, it is a popular custom to adapt these parameters to obtain a minimal error rate on the test set. The same rate is used to rank the algorithm which causes an optimistic bias. We quantify this bias, showing in particular that an algorithm with more parameters will probably be ranked higher than an equally good algorithm with fewer parameters. We demonstrate this result, showing the number of parameters and trials required in order to pretend to outperform C4.5 or FOIL, respectively, for various benchmark problems. We then describe how unbiased ranking experiments should be conducted.