On-line learning of linear functions

We present an algorithm for the on-line learning of linear functions which is optimal to within a constant factor with respect to bounds on the sum of squared errors for a worst case sequence of trials. The bounds are logarithmic in the number of variables. Furthermore, the algorithm is shown to be optimally robust with respect to noise in the data (again to within a constant factor).     

Phase I trial of the thioether phospholipid analogue BM 41.440 in cancer patients

The online version of the original article can be found at     

Geographic information systems and spatial analysis

Geographical Information Systems (GIS) are capable of acquiring spatially indexed data from a variety of sources, changing the data into useful formats, storing the data, retrieving and manupulating the data for analysis, and then generating the output required by a given user. Their great strength is based on the ability to handle large, multilayered, heterogenous databases and to query about the existence, location and properties of a wide range of spatial objects in an interactive way. The lack of analytical and modelling functionality is, however, widely recognised as a major deficiency of current systems. There is a wide agreement in both the GIS community and the modelling community that the future success of GIS technology will depend to a large extent on incorporating more powerful analytical and modelling capabilities. This paper discusses some major directions and strategies to increase both the analytical and modelling capabilities and the level of intelligence of geographic information systems.     

Comparison and evaluation of creep crack growth rate and fracture time for Alloy 800 by the Q* parameter

Creep crack growth tests were carried out under several temperature and stress conditions for Fe-based superalloy Alloy 800 and the Q* parameter was applied to an evaluation of the crack growth rate at elevated temperatures for DEN and CT specimens. Creep fracture times were evaluated with the parameter derived by integrating the creep crack growth rate formula expressed in terms of the Q* parameter for DEN, CT and smooth specimens of Alloy 800 under several temperature and stress conditions. It was found that the creep crack growth rate for CT specimens is represented as the different data bands by the Q* parameter for DEN specimens. However, it can be evaluated by parallel movement along the Q* axis in the Q* parameter-da/dt diagram for Alloy 800. The parameter {T(log₁₀t_f+20)×10⁻³} derived from the Q* parameter can compare and evaluate the difference in specimen shape and loading mechanism on creep fracture time using the identical parameter for Alloy 800.     

Abnormal Grain Growth in Alumina: Synergistic Effects of Yttria and Silica

Abnormal grain growth without strong anisotropy or faceting of the grains has been observed in high-purity yttria-doped alumina specimens, often starting at the surface and spreading right through the bulk at higher sintering temperatures. This appears to occur because of an interaction between Si contamination from sintering and the yttria doping; no such effect is seen for undoped samples. Similar microstructures were observed after deliberate Y/Si codoping. Analytical STEM showed that some grain boundaries bordering on large grains contained more Si than Y. HRTEM and diffuse dark-field imaging revealed thin (0.5–0.9 nm) disordered layers at some boundaries bordering large grains. It appears that Si impurities are accumulating at some boundaries and together with the Y inducing a grain boundary structural transformation that accounts for the dramatically increased mobility of these boundaries.     

Entwicklung einer Dekompositionsstrategie für die Strukturoptimierung komplexer Bauteile

Zusammenfassung Der Einsatz von Finite-Elemente-Optimierungsprogrammen für die konstruktive Auslegung von Bauteilen ist mit gro?em Rechenaufwand und Speicherplatzbedarf verbunden. Die Dekompositionsstrategie stellt ein effizientes Verfahren dar, das zu einer drastischen Reduzierung der Rechenzeiten bei der L?sung von Problemen der Gewichtsminimierung führt. Darüber hinaus verringert sich der immens hohe Bedarf an Plattenplatz und nicht zuletzt wird ein schnelleres Konvergenzverhalten bei der Optimierung erreicht. Die in diesem Beitrag vorgestellten Arbeiten sind Gegenstand des von der Deutschen Forschungsgemeinschaft (DFG) gef?rderten Forschungsprojektes “Entwicklung von Dekompositionsstrategien für die Strukturoptimierung unter Einsatz von Parallelrechnern?