Improving generalization capabilities of dynamic neural networks

This work addresses the problem of improving the generalization capabilities of continuous recurrent neural networks. The learning task is transformed into an optimal control framework in which the weights and the initial network state are treated as unknown controls. A new learning algorithm based on a variational formulation of Pontrayagin's maximum principle is proposed. Under reasonable assumptions, its convergence is discussed. Numerical examples are given that demonstrate an essential improvement of generalization capabilities after the learning process of a dynamic network.     

Holzinger, A.: Basiswissen Multimedia Bd. 1: Technik. Zahlir Abb., 284 S. ISBN 3-8023-1856-0; Bd. 2: Lernen. Zahlr. Abb., 290 S. ISBN 3-8023-1857-0; Bd. 3: Design. Zahlr. Abb., 280 S. ISBN 3-8023-1858-0. Vogl, Würzburg, 2000. DM 59,-, ATS 431,- (je Band)

Buchanzeige

Holzinger, A.: Basiswissen Multimedia Bd. 1: Technik. Zahlir Abb., 284 S. ISBN 3-8023-1856-0; Bd. 2: Lernen. Zahlr. Abb., 290 S. ISBN 3-8023-1857-0; Bd. 3: Design. Zahlr. Abb., 280 S. ISBN 3-8023-1858-0. Vogl, Würzburg, 2000. DM 59,-, ATS 431,- (je Band)     

Multisensory recognition of actively explored objects.

Shape recognition can be achieved through vision or touch, raising the issue of how this information is shared across modalities. Here we provide a short review of previous findings on cross-modal object recognition and we provide new empirical data on multisensory recognition of actively explored objects. It was previously shown that, similar to vision, haptic recognition of objects fixed in space is orientation specific and that cross-modal object recognition performance was relatively efficient when these views of the objects were matched across the sensory modalities (Newell, Ernst, Tjan, & Bülthoff, 2001). For actively explored (i.e., spatially unconstrained) objects, we now found a cost in cross-modal relative to within-modal recognition performance. At first, this may seem to be in contrast to findings by Newell et al. (2001). However, a detailed video analysis of the visual and haptic exploration behaviour during learning and recognition revealed that one view of the objects was predominantly explored relative to all others. Thus, active visual and haptic exploration is not balanced across object views. The cost in recognition performance across modalities for actively explored objects could be attributed to the fact that the predominantly learned object view was not appropriately matched between learning and recognition test in the cross-modal conditions. Thus, it seems that participants naturally adopt an exploration strategy during visual and haptic object learning that involves constraining the orientation of the objects. Although this strategy ensures good within-modal performance, it is not optimal for achieving the best recognition performance across modalities. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Auftraglöten zum Verschleißschutz von Titanwerkstoffen

Wear Protection of Titanium using Surface Brazing Titanium and titanium alloys possess high specific strengths up to a temperature of about 600 °C in addition to an extraordinary corrosion resistance [1]. The low wear resistance constitutes a crucial impediment for a much broader use. Titanium materials are especially susceptible to friction fatigue and erosion. Coating techniques have to be developed in order to counteract this technical constraint. Surface brazing presents a promising approach. Hard metals mixed with brazing filler metals on a silver and titanium basis were brazed in a vacuum furnace and subsequently characterized. Wear resistance was quantified and optimized using ball on disc measurements.     

Organometallic Bonding in an Ullmann‐Type On‐Surface Chemical Reaction Studied by High‐Resolution Atomic Force Microscopy

The on‐surface Ullmann‐type chemical reaction synthesizes polymers by linking carbons of adjacent molecules on solid surfaces. Although an organometallic compound is recently identified as the reaction intermediate, little is known about the detailed structure of the bonded organometallic species and its influence on the molecule and the reaction. Herein atomic force microscopy at low temperature is used to study the reaction with 3,9‐diiododinaphtho[2,3‐b:2′,3′‐d]thiophene (I‐DNT‐VW), which is polymerized on Ag(111) in vacuum. Thermally sublimated I‐DNT‐VW picks up a Ag surface atom, forming a C? Ag bond at one end after removing an iodine. The C? Ag bond is usually short‐lived, and a C? Ag? C organometallic bond immediately forms with an adjacent molecule. The existence of the bonded Ag atoms strongly affects the bending angle and adsorption height of the molecular unit. Density functional theory calculations reveal the bending mechanism, which reveals that charge from the terminus of the molecule is transferred via the Ag atom into the organometallic bond and strengths the local adsorption to the substrate. Such deformations vanish when the Ag atoms are removed by annealing and C? C bonds are established.     

A FE/BE coupling for the 3D time-dependent eddy current problem. Part I: a priori error estimates

In this paper the discontinuous Galerkin method in time for the coupling of conforming finite element and boundary element methods is established. We derive quasi-optimal a priori error estimates. Numerical examples prove the new scheme to be useful in practice. A posteriori error control and an adaptive algorithm are studied in Part II of this paper.     

Complementary/alternative medicine for asthma: we do not know what we need to know

In a recent report, a series of studies is described showing that individual differences in rat maternal licking and grooming are correlated with their offsprings' later adrenal response to a stressor (Liu et al., 1997). Pups that received more maternal stimulation in infancy had lower ACTH, corticosterone, and CRH mRNA, while they had greater amounts of GR mRNA. Liu et al. also compared maternal behavior in litters where rat pups were handled daily to maternal behavior of nonhandled litters. They found that mothers of handled pups licked and groomed their young significantly more often than did mothers of control litters. In their discussion, Liu et al. proposed that their data support Levine's (1975) thesis that handling effects in infancy are mediated via the mother. Levine had proposed that handling of pups will modify the mother--pup interaction, thereby causing a change in maternal behavior. The purpose of this commentary is to summarize data obtained with mice and rabbits which, in part, support and, in part, challenge the conclusions of Liu et al.