Universitätspublikationen
Refine
Year of publication
Document Type
- Article (13385)
- Part of Periodical (3455)
- Doctoral Thesis (3276)
- Contribution to a Periodical (2132)
- Book (2063)
- Working Paper (1884)
- Preprint (1469)
- Review (1063)
- Report (909)
- Conference Proceeding (692)
Language
- English (16849)
- German (13801)
- Portuguese (231)
- Spanish (123)
- Italian (66)
- French (64)
- Multiple languages (59)
- Turkish (12)
- Ukrainian (10)
- slo (7)
Keywords
- Deutschland (132)
- COVID-19 (95)
- inflammation (92)
- Financial Institutions (90)
- ECB (67)
- Capital Markets Union (65)
- SARS-CoV-2 (63)
- Financial Markets (61)
- Adorno (58)
- Banking Union (50)
Institute
- Medizin (6550)
- Präsidium (5053)
- Physik (3316)
- Wirtschaftswissenschaften (2269)
- Gesellschaftswissenschaften (2015)
- Biowissenschaften (1739)
- Frankfurt Institute for Advanced Studies (FIAS) (1504)
- Biochemie und Chemie (1391)
- Sustainable Architecture for Finance in Europe (SAFE) (1389)
- Informatik (1376)
This thesis explores a variety of methods of text quantification applicable in the field of educational text technology. Besides the cohort of existing linguistic, lexical, syntactic, and semantic text quantification methods, additional methods based on Bidirectional Encoder Representations from Transformers (BERT) are introduced and analysed. The model, developed in this thesis, is tested on a multilingual data composed of task descriptions used in Test of Understanding in College Economics (TUCE). Quantitative features extracted from raw textual data are analysed using an array of evaluation methods with the goal of finding the best predictors of the target variable - the rate of correct student responses in TUCE.
Event-by-event multiplicity fluctuations in nucleus-nucleus collisions from low SPS up to RHIC energies have been studied within the HSD transport approach. Fluctuations of baryonic number and electric charge also have been explored for Pb+Pb collisions at SPS energies in comparison to the experimental data from NA49. We find a dominant role of the fluctuations in the nucleon participant number for the final hadron multiplicity fluctuations and a strong influence of the experimental acceptance on the final results. Critical Point and Onset of Deconfinement - 4th International Workshop July 9 - 13, 2007 Darmstadt, Germany
The multiplicity fluctuations in A+A collisions at SPS and RHIC energies are studied within the HSD transport approach. We find a dominant role of the fluctuations in the nucleon participant number for the final fluctuations. In order to extract physical fluctuations one should decrease the fluctuations in the participants number. This can be done considering very central collisions. The system size dependence of the multiplicity fluctuations in central A+A collisions at the SPS energy range – obtained in the HSD and UrQMD transport models – is presented. The results can be used as a ‘background’ for experimental measurements of fluctuations as a signal of the critical point. Event-by-event fluctuations of the K/p , K/p and p/p ratios in A+A collisions are also studied. Event-by-event fluctuations of the kaon to pion number ratio in nucleus-nucleus collisions are studied for SPS and RHIC energies. We find that the HSD model can qualitatively reproduce the measured excitation function for the K/p ratio fluctuations in central Au+Au (or Pb+Pb) collisions from low SPS up to top RHIC energies. The forward-backward correlation coefficient measured by the STAR Collaboration in Au+Au collisions at RHIC is also studied. We discuss the effects of initial collision geometry and centrality bin definition on correlations in nucleus-nucleus collisions. We argue that a study of the dependence of correlations on the centrality bin definition as well as the bin size may distinguish between these ‘trivial’ correlations and correlations arising from ‘new physics’. 5th International Workshop on Critical Point and Onset of Deconfinement - CPOD 2009, June 08 - 12 2009 Brookhaven National Laboratory, Long Island, New York, USA
Background: Subarachnoid hemorrhage (SAH) is mainly caused by ruptured cerebral aneurysms but in up to 15% of patients with SAH no bleeding source could be identified. Our objective was to analyze patient characteristics, clinical outcome and prognostic factors in patients suffering from non-aneurysmal SAH.
Methods: From 1999 to 2009, data of 125 patients with non-aneurysmal SAH were prospectively entered into a database. All patients underwent repetitive cerebral angiography. Outcome was assessed according to the modified Rankin Scale (mRS) (mRS 0-2 favorable vs. 3-6 unfavorable). Also, patients were divided in two groups according to the distribution of blood in the CT scan (perimesencephalic and non-perimesencephalic SAH).
Results: 106 of the 125 patients were in good WFNS grade (I-III) at admission (85%). Overall, favorable outcome was achieved in 104 of 125 patients (83%). Favorable outcome was associated with younger age (P < 0.001), good admission status (P < 0.0001), and absence of hydrocephalus (P = 0.001).73 of the 125 patients suffered from perimesencephalic SAH, most patients (90%) were in good grade at admission, and 64 achieved favorable outcome.52 of the 125 patients suffered from non-perimesencephalic SAH and 40 were in good grade at admission. Also 40 patients achieved favorable outcome.
Conclusions: Patients suffering from non-aneurysmal SAH have better prognosis compared to aneurysm related SAH and poor admission status was the only independent predictor of unfavorable outcome in the multivariate analysis. Patients with a non-perimesencephalic SAH have an increased risk of a worse neurological outcome. These patients should be monitored attentively.
This thesis contributes to the field of machine learning with a specific focus on the methods for learning relations between the inputs. Learning relationships between images is the most common primitive in vision. There are many vision tasks in which relationships across images play an important role. Some of them are motion estimation, activity recognition, stereo vision, multi-view geometry and visual odometry. Many of such tasks mainly depend on motion and disparity cues, which are inferred based on the relations across multiple image pairs. The approaches presented in this thesis mainly deal with, but are not limited to, learning of the representations for motion and depth. This thesis by articles consists of five articles which present relational feature learning models along with their applications in computer vision. In the first article, we present an approach for encoding motion in videos. To this end, we show that the detection of spatial transformations can be viewed as detection of coincidence or synchrony between the given sequence of frames and a sequence of features which are related by the transformation we wish to detect. Learning to detect synchrony is possible by introducing "multiplicative interactions'' into the hidden units of single layered sparse coding models.
We show that the learned motion representations employed for the task of activity recognition achieve competitive performance on multiple benchmarks. Stereo vision is an important challenge in computer vision and useful for many applications in that field. In the second article, we extend the energy based learning models, which were previously used for motion encoding, to the context of depth perception. Given the common architecture of the models for encoding motion and depth, we show that it is possible to define a single model for learning a unified representation for both the cues. Our experimental results show that learning a combined representation for depth and motion makes it possible to achieve state-of-the-art performance at the task of 3-D activity analysis, and to perform better than the existing hand-engineered 3-D motion features. Autoencoder is a popular unsupervised learning method for learning efficient encoding for a given set of data samples. Typically, regularized autoencoders which are used to learn over-complete and sparse representations for the input data, were shown to fail on intrinsically high dimensional data like videos. In the third article, we investigate the reason for such a behavior. It can be observed that the regularized autoencoders typically learn negative hidden unit biases. We show that the learning of negative biases is the result of hidden units being responsible for both the sparsity and the representation of the input data. It is shown that, as a result, the behavior of the model resembles clustering methods which would require exponentially large number of features to model intrinsically high dimensional data. Based on this understanding, we propose a new activation function which decouples the roles of hidden layer and uses linear encoding. This allows to learn representations on data with very high intrinsic dimensionality. We also show that gating connections in the bi-linear models and the single layer models from articles one and two of this thesis can be thought of as a way to attain a linear encoding scheme which allows them to learn good representations on videos. Visual odometry is the task of inferring egomotion of a moving object from visual information such as images and videos. It can primarily be used for the task of localization and has many applications in the fields of robotics and navigation. The work in article four was motivated by the idea of using deep learning techniques, which are successful methods for many vision tasks, for visual odometry. The visual odometry task mainly requires inference of motion and depth information from visual input which can then be mapped to velocity and change in direction. We use relational feature models presented in the articles one and two for inferring a combined motion and depth representation from stereo video sequences. The combined representation is then mapped to discrete velocity and change in direction labels using convolutional neural networks. Our approach is an end-to-end deep learning-based architecture which uses a single type of computational model and learning rule. Preliminary results show that the architecture is capable of learning the mapping from input video to egomotion. Activity recognition is a challenging computer vision task with many real world applications. It is well know that it is a hard task to use computer vision research for real-time applications. In the fifth article of this thesis, we present a real-time activity recognition system based on deep learning based methods. Our approach uses energy based relational feature learning models for the computation of local motion features directly from videos. A bag-of-words over the local motion features is used for the analysis of activity in a given video sequence. We implement this system on a distributed computational platform and demonstrate its performance on the iCub robot. Using GPUs we demonstrate real time performance which makes the deployment of activity recognition systems in real world scenarios possible.
The prevention of tau protein aggregations is a therapeutic goal for the treatment of Alzheimer's disease (AD), and hydromethylthionine (HMT) (also known as leucomethylthioninium-mesylate [LMTM]), is a potent inhibitor of tau aggregation in vitro and in vivo. In two Phase 3 clinical trials in AD, HMT had greater pharmacological activity on clinical endpoints in patients not receiving approved symptomatic treatments for AD (acetylcholinesterase (AChE) inhibitors and/or memantine) despite different mechanisms of action. To investigate this drug interaction in an animal model, we used tau-transgenic L1 and wild-type NMRI mice treated with rivastigmine or memantine prior to adding HMT, and measured changes in hippocampal acetylcholine (ACh) by microdialysis. HMT given alone doubled hippocampal ACh levels in both mouse lines and increased stimulated ACh release induced by exploration of the open field or by infusion of scopolamine. Rivastigmine increased ACh release in both mouse lines, whereas memantine was more active in tau-transgenic L1 mice. Importantly, our study revealed a negative interaction between HMT and symptomatic AD drugs: the HMT effect was completely eliminated in mice that had been pre-treated with either rivastigmine or memantine. Rivastigmine was found to inhibit AChE, whereas HMT and memantine had no effects on AChE or on choline acetyltransferase (ChAT). The interactions observed in this study demonstrate that HMT enhances cholinergic activity in mouse brain by a mechanism of action unrelated to AChE inhibition. Our findings establish that the drug interaction that was first observed clinically has a neuropharmacological basis and is not restricted to animals with tau aggregation pathology. Given the importance of the cholinergic system for memory function, the potential for commonly used AD drugs to interfere with the treatment effects of disease-modifying drugs needs to be taken into account in the design of clinical trials.
Schätzungen der WHO zufolge waren 2015 weltweit rund 71 Millionen Menschen von einer chronischen Hepatitis C-Infektion betroffen. Die chronische Hepatitis C ist mit einem erhöhten Risiko für die Entstehung einer Leberzirrhose und eines hepatozellulären Karzinoms assoziiert. Die NS3/4A-Protease als zentraler Bestandteil der Replikationsmaschinerie des Virus spaltet das HCV-Polyprotein und ist in die Inaktivierung antiviraler Proteine involviert. Durch ihren maßgeblichen Einfluss auf die virale Fitness stellt sie einen entscheidenden Faktor für die chronische Persistenz des Virus im Wirtsorganismus dar. Die Protease ist auch eine wichtige Zielstruktur für spezifische antivirale Medikamente in der Behandlung der chronischen Hepatitis C. Der natürlich vorkommende Polymorphismus Q80K in der NS3/4A-Protease ist bei bis zu 47 % der Patienten schon vor Therapiebeginn feststellbar, insbesondere beim Genotyp 1a. Q80K führt zum Therapieversagen bei makrozyklischen Proteaseinhibitoren, insbesondere Simeprevir. Phylogenetische Analysen konnten zeigen, dass 96 % aller HCV-Gensequenzen mit Q80K von einem gemeinsamen, genetischen Vorfahren abstammen und sich die Mutation seit Mitte des 20. Jahrhunderts scheinbar stabil ausgehend vom nordamerikanischen Kontinent etabliert hat. Daneben wurden mit A91S/T und S174N sogenannte second site-Austausche identifiziert, die assoziiert mit Q80K vorkommen. Ziel dieser Arbeit war es herauszufinden, welchen Einfluss diese second site-Austausche auf die Enzymaktivität und Proteinfaltung der Protease haben und ob sie mögliche Veränderungen durch den Q80K-Polymorphismus kompensieren. Nach Expression und Aufreinigung der NS3/4A-Protease wurden die Effekte von Q80K, A91S/T und S174N auf die Enzymaktivität und Thermostabilität mittels fluoreszenzbasierter Verfahren untersucht und im Zusammenhang mit einer in silico-3D-Strukturanalyse der Protease interpretiert. Es zeigte sich, dass A91S/T und S174N jeweils zu einer Angleichung der Thermostabilität des Proteins an den Wildtyp führen und somit Defizite in der Faltung der Protease durch Q80K kompensiert werden. Aufgrund der experimentellen Daten und der Topografie dieser Austausche innerhalb der NS3-Protease-Helikase-Struktur ist von indirekten Effekten der second site-Austausche auf die replikative Fitness der Virusvarianten auszugehen. Die hier charakterisierten Austausche in der NS3/4A-Protease tragen durch eine Stabilisierung der Proteinfaltung kritisch zur Stabilität des Q80K-Polymorphismus im Proteasegen des HCV Genotyp 1a bei.
Mapping cortical brain asymmetry in 17,141 healthy individuals worldwide via the ENIGMA Consortium
(2017)
Motivated by the on-going discussion on the nature of magnetism in the quantum Ising chain CoNb2O6, we present a first-principles-based analysis of its exchange interactions by applying an \textit{ab initio} approach with additional modelling that accounts for various drawbacks of a purely density functional theory ansatz. With this method we are able to extract and understand the origin of the magnetic couplings under inclusion of all symmetry-allowed terms, and to resolve the conflicting model descriptions in CoNb2O6. We find that the twisted Kitaev chain and the transverse-field ferromagnetic Ising chain views are mutually compatible, although additional off-diagonal exchanges are necessary to provide a complete picture. We show that the dominant exchange interaction is a ligand-centered exchange process - involving the eg electrons -, which is rendered anisotropic by the low-symmetry crystal fields environments in CoNb2O6, giving rise to the dominant Ising exchange, while the smaller bond-dependent anisotropies are found to originate from d−d kinetic exchange processes involving the t2g electrons. We demonstrate the validity of our approach by comparing the predictions of the obtained low-energy model to measured THz and inelastic neutron scattering spectra.
The marine hydrocarbonoclastic bacterium Alcanivorax borkumensis is well known for its ability to successfully degrade various mixtures of n-alkanes occurring in marine oil spills. For effective growth on these compounds, the bacteria possess the unique capability not only to incorporate but also to modify fatty intermediates derived from the alkane degradation pathway. High efficiency of both these processes provides better competitiveness for a single bacteria species among hydrocarbon degraders. To examine the efficiency of A. borkumensis to cope with different sources of fatty acid intermediates, we studied the growth rates and membrane fatty acid patterns of this bacterium cultivated on diesel, biodiesel and rapeseed oil as carbon and energy source. Obtained results revealed significant differences in both parameters depending on growth substrate. Highest growth rates were observed with biodiesel, while growth rates on rapeseed oil and diesel were lower than on the standard reference compound (hexadecane). The most remarkable observation is that cells grown on rapeseed oil, biodiesel, and diesel showed significant amounts of the two polyunsaturated fatty acids linoleic acid and linolenic acid in their membrane. By direct incorporation of these external fatty acids, the bacteria save energy allowing them to degrade those pollutants in a more efficient way. Such fast adaptation may increase resilience of A. borkumensis and allow them to strive and maintain populations in more complex hydrocarbon degrading microbial communities.