Generalization ability of folding networks

The information theoretical learnability of folding networks, a very successful approach capable of dealing with tree structured inputs, is examined. We find bounds on the VC, pseudo-, and fat shattering dimension of folding networks with various activation functions. As a consequence, valid generalization of folding networks can be guaranteed. However, distribution independent bounds on the generalization error cannot exist in principle. We propose two approaches which take the specific distribution into account and allow us to derive explicit bounds on the deviation of the empirical error from the real error of a learning algorithm. The first approach requires the probability of large trees to be limited a priori and the second approach deals with situations where the maximum input height in a concrete learning example is restricted     

Sulfur bonding in MoS2 and Co-Mo-S structures

The structure and bonding in small MoS₂ structures with and without Co is studied theoretically using self-consistent density functional theory with a non-local exchange-correlation energy. The structures model the catalysts used extensively in hydrotreating. We study in detail the structure and binding energies as a function of the amount of sulfur. The calculations show that extensive reconstructions occur at the two types of MoS₂ edges where the sulfur dimerizes and occupies non-lattice positions. These structures are shown to be in good agreement with available experimental data. We also study the energy required to form sulfur vacancies, which are believed to be the active sites for many hydrotreating reactions. The presence of Co atoms at the edges is shown to lead to a significant lowering of the metal-sulfur binding energy. This imposes an increase in the concentration of active sites for the reactions and may thus explain the promoting effect of Co. This revised version was published online in July 2006 with corrections to the Cover Date.     

Cui bono? — Ziele und Inhalte eines Datenschutz-Zertifikats

Um ein Datenschutz-Zertifikat zur Best?tigung eines hohen Datenschutz-Niveaus erfolgreich zu etablieren, müssen die Randbedingungen stimmen. Ausgehend vom Informationsbedarf m?glicher Zielgruppen an ein solches Zertifikat begründet der Beitrag die Datenschutzorganisation als Prüfgegenstand und stellt die aus der Sicht der Autoren wichtigsten Inhalte der Prüfung vor.     

High resolution MR imaging of joint degeneration in the knee of the STR/ORT mouse

MR imaging of the knee joints of the mouse at 2.35 T produces an in-plane resolution of 35 x 70 microns and a slice thickness of 600 microns from the sagittal and coronal planes, in less than 60 min. In normal, live DBA mice, which have no known pathology, the images clearly resolve many joint structures, such as the cruciate ligaments, menisci, and articulating surfaces of the femur and tibia. Gross pathological changes were identified first in excised knees from cadaver mice of the STR/ORT strain. Males of this strain spontaneously exhibit rapid joint degeneration with increasing age compared to females, in which much milder symptoms develop more slowly. Thickening of the patellar tendon, displacement of the patella, deformity, and sclerosis were clearly distinguishable in males compared with females older than 7 mo of age. The same degenerative features were evident in MR images of the knees of the live male but not female STR/ORT mice of this age, indicating that it would be possible to monitor longitudinally by MR imaging progressive development of joint changes in this osteoarthritic model.