Christian Ullrich – författare

The ESPRIT Project 1072, DIAMOND (Development and Integration of Accurate Mathematical Operations in Numerical Data-Processing) was carried out from January 1986 through April 1989 by the five partners Siemens Miinchen (prime contractor), CWI Amsterdam, University of Karlsruhe (Institut fiir Ange­ wandte Mathematik), NAG Oxford and University of Bath (subcontractor to NAG). The technical work was divided into three main work packages with one additional work package for miscellaneous topics. The major goals of this project, according to its title, were to develop a set of accurate numerical algorithms (work package 3) and to provide tools to support their implementation by means of an embedding of accurate arithmetic into programming languages (work package 1) and by transform­ ation techniques which either improve the accuracy of expression evaluation or detect and eliminate presumable deficiencies in accuracy in existing programs (work package 2). A great variety of working papers describing and discussing the results of these work packages have been written during the collaboration of the project. This book (Accurate Numericlll A 19oritltms) mainly summarizes the results of work package 3, carried through by the two partners Karlsruhe University and NAG Oxford under the leadership of the editors and Dr. G. S. Hodgson. Another book (Improving Floating-Point Programming, J. Tliley, to appear 1989), edited by P.J.L. Wallis, which is one of the DIAMOND project's final deliverables, concentrates more on the fundamental tools.

PASCAL-XSC (PASCAL-eXtension for Scientific Computation) ist eine Erweiterung der weitverbreiteten Programmiersprache PASCAL. Sie verfolgt das Ziel, das Programmieren, insbesondere im Bereich technisch-wissenschaftlicher Anwendungen, durch zusätzliche Sprachkonzepte wie Module, dynamische Felder und Operatoren erheblich zu vereinfachen. Eine optimale Arithmetik mit hochgenauen Standardfunktionen und exakter Ausdrucksauswertung ermöglicht eine automatisierte, zuverlässige Kontrolle der berechneten Ergebnisse. Damit unterstützt PASCAL-XSC die Entwicklung von Routinen mit automatischer Ergebnisverifikation. Mittels der Compiler-Implementierung in C kann PASCAL-XSC auf PC's, Workstations, Großrechnern und Supercomputern gleichermaßen eingesetzt werden. In dem vorliegenden Lehr- und Handbuch wird PASCAL-XSC vollständig beschrieben. Zur praktischen Verwendung und zum leichteren Kennenlernen und Vertrautwerden mit den über PASCAL hinausgehenden neuen Sprachelementen ist ein ausführlicher Abschnitt mit Übungsaufgaben und Lösungen aufgenommen worden. Ein kompletter Satz von Syntaxdiagrammen, sowie ausführliche Register und Verzeichnisse schließen dieses Buch ab.

This manual describes a PASCAL extension for scientific computation with the short title PASCAL-XSC (PASCAL eXtension for Scientific Computation). The language is the result of a long term effort of members of the Institute for Applied Mathematics of Karlsruhe University and several associated scientists. PASCAL­ XSC is intended to make the computer more powerful arithmetically than usual. It makes the computer look like a vector processor to the programmer by providing the vector/matrix operations in a natural form with array data types and the usual operator symbols. Programming of algorithms is thus brought considerably closer to the usual mathematical notation. As an additional feature in PASCAL-XSC, all predefined operators for real and complex numbers and intervals, vectors, matrices, and so on, deliver an answer that differs from the exact result by at most one rounding. Numerical mathematics has devised algorithms that deliver highly accurate and automatically verified results by applying mathematical fixed point theorems. That is, these computations carry their own accuracy control. However, their imple­ mentation requires arithmetic and programming tools that have not been available previously. The development of PASCAL-XSC has been aimed at providing these tools within the PASCAL setting. Work on the subject began during the 1960's with the development of a general theory of computer arithmetic. At first, new algorithms for the realization of the arithmetic operations had to be developed and implemented.

Historical and recent developments at international ?nancial markets show that it is easy to loose money, while it is dif?cult to predict future developments and op- mize decision-making towards maximizing returns and minimizing risk. One of the reasons of our inability to make reliable predictions and to make optimal decisions is the growing complexity of the global economy. This is especially true for the f- eign exchange market (FX market) which is considered as one of the largest and most liquid ?nancial markets. Its grade of ef?ciencyand its complexityis one of the starting points of this volume. From the high complexity of the FX market, Christian Ullrich deduces the - cessity to use tools from machine learning and arti?cial intelligence, e.g., support vector machines, and to combine such methods with sophisticated ?nancial mod- ing techniques. The suitability of this combination of ideas is demonstrated by an empirical study and by simulation. I am pleased to introduce this book to its - dience, hoping that it will provide the reader with interesting ideas to support the understanding of FX markets and to help to improve risk management in dif?cult times. Moreover, I hope that its publication will stimulate further research to contribute to the solution of the many open questions in this area.