Technische Universität Darmstadt

About: Technische Universität Darmstadt is a education organization based out in Darmstadt, Germany. It is known for research contribution in the topics: Computer science & Context (language use). The organization has 17316 authors who have published 40619 publications receiving 937916 citations. The organization is also known as: Darmstadt University of Technology & University of Darmstadt.

MoverScore: Text Generation Evaluating with Contextualized Embeddings and Earth Mover Distance

Abstract: A robust evaluation metric has a profound impact on the development of text generation systems. A desirable metric compares system output against references based on their semantics rather than surface forms. In this paper we investigate strategies to encode system and reference texts to devise a metric that shows a high correlation with human judgment of text quality. We validate our new metric, namely MoverScore, on a number of text generation tasks including summarization, machine translation, image captioning, and data-to-text generation, where the outputs are produced by a variety of neural and non-neural systems. Our findings suggest that metrics combining contextualized representations with a distance measure perform the best. Such metrics also demonstrate strong generalization capability across tasks. For ease-of-use we make our metrics available as web service.

Learning to select and generalize striking movements in robot table tennis

Abstract: Learning new motor tasks from physical interactions is an important goal for both robotics and machine learning. However, when moving beyond basic skills, most monolithic machine learning approaches fail to scale. For more complex skills, methods that are tailored for the domain of skill learning are needed. In this paper, we take the task of learning table tennis as an example and present a new framework that allows a robot to learn cooperative table tennis from physical interaction with a human. The robot first learns a set of elementary table tennis hitting movements from a human table tennis teacher by kinesthetic teach-in, which is compiled into a set of motor primitives represented by dynamical systems. The robot subsequently generalizes these movements to a wider range of situations using our mixture of motor primitives approach. The resulting policy enables the robot to select appropriate motor primitives as well as to generalize between them. Finally, the robot plays with a human table tennis partner and learns online to improve its behavior. We show that the resulting setup is capable of playing table tennis using an anthropomorphic robot arm.

High-performance fuel cell cathodes exclusively containing atomically dispersed iron active sites

Abstract: Platinum group metal-free (PGM-free) catalysts for the oxygen reduction reaction (ORR) with atomically dispersed FeN4 sites have emerged as a potential replacement for low-PGM catalysts in acidic polymer electrolyte fuel cells (PEFCs). In this work, we carefully tuned the doped Fe content in zeolitic imidazolate framework (ZIF)-8 precursors and achieved complete atomic dispersion of FeN4 sites, the sole Fe species in the catalyst based on Mosbauer spectroscopy data. The Fe–N–C catalyst with the highest density of active sites achieved respectable ORR activity in rotating disk electrode (RDE) testing with a half-wave potential (E1/2) of 0.88 ± 0.01 V vs. the reversible hydrogen electrode (RHE) in 0.5 M H2SO4 electrolyte. The activity degradation was found to be more significant when holding the potential at 0.85 V relative to standard potential cycling (0.6–1.0 V) in O2 saturated acid electrolyte. The post-mortem electron microscopy analysis provides insights into possible catalyst degradation mechanisms associated with Fe–N coordination cleavage and carbon corrosion. High ORR activity was confirmed in fuel cell testing, which also divulged the promising performance of the catalysts at practical PEFC voltages. We conclude that the key factor behind the high ORR activity of the Fe–N–C catalyst is the optimum Fe content in the ZIF-8 precursor. While too little Fe in the precursors results in an insufficient density of FeN4 sites, too much Fe leads to the formation of clusters and an ensuing significant loss in catalytic activity due to the loss of atomically dispersed Fe to inactive clusters or even nanoparticles. Advanced electron microscopy was used to obtain insights into the clustering of Fe atoms as a function of the doped Fe content. The Fe content in the precursor also affects other key catalyst properties such as the particle size, porosity, nitrogen-doping level, and carbon microstructure. Thanks to using model catalysts exclusively containing FeN4 sites, it was possible to directly correlate the ORR activity with the density of FeN4 species in the catalyst.

Normal impact of a liquid drop on a dry surface: model for spreading and receding

Abstract: The normal impact of a liquid drop on a dry solid surface is studied experimentally and theoretically. In this paper a strictly theoretical model is introduced, which predicts the evolution of the drop diameter. The spreading and receding phases of the impact are described by the motion of a rim appearing at the edge of the liquid film (lamella) due to the surface-tension forces. The mass and the momentum equations of the rim are considered, taking into account the effects of inertial, viscous and surface forces, and wettability. Also, simplified approximations for the maximum spreading diameter of the drop and for the velocity of the merging of the rim in the receding phase are obtained. The theoretical predictions agree well with available experimental data.