TY - JOUR A1 - Derras, Boumediene A1 - Bard, Pierre-Yves A1 - Cotton, Fabrice Pierre T1 - V-S30, slope, H-800 and f(0): performance of various site-condition proxies in reducing ground-motion aleatory variability and predicting nonlinear site response JF - Earth, planets and space N2 - The aim of this paper is to investigate the ability of various site-condition proxies (SCPs) to reduce ground-motion aleatory variability and evaluate how SCPs capture nonlinearity site effects. The SCPs used here are time-averaged shear-wave velocity in the top 30 m (V-S30), the topographical slope (slope), the fundamental resonance frequency (f(0)) and the depth beyond which V-s exceeds 800 m/s (H800). We considered first the performance of each SCP taken alone and then the combined performance of the 6 SCP pairs [V-S30-f(0)], [V-S30-H-800], [f(0)-slope], [H-800-slope], [V-S30-slope] and [f(0)-H-800]. This analysis is performed using a neural network approach including a random effect applied on a KiK-net subset for derivation of ground-motion prediction equations setting the relationship between various ground-motion parameters such as peak ground acceleration, peak ground velocity and pseudo-spectral acceleration PSA (T), and Mw, RJB, focal depth and SCPs. While the choice of SCP is found to have almost no impact on the median groundmotion prediction, it does impact the level of aleatory uncertainty. VS30 is found to perform the best of single proxies at short periods (T < 0.6 s), while f(0) and H-800 perform better at longer periods; considering SCP pairs leads to significant improvements, with particular emphasis on [V-S30-H-800] and [f(0)-slope] pairs. The results also indicate significant nonlinearity on the site terms for soft sites and that the most relevant loading parameter for characterising nonlinear site response is the "stiff" spectral ordinate at the considered period. KW - Aleatory variability KW - Site-condition proxies KW - KiK-net KW - Neural networks KW - GMPE KW - Nonlinear site response Y1 - 2017 U6 - https://doi.org/10.1186/s40623-017-0718-z SN - 1880-5981 VL - 69 SP - 1623 EP - 1629 PB - Springer CY - Heidelberg ER - TY - THES A1 - Zhelavskaya, Irina T1 - Modeling of the Plasmasphere Dynamics T1 - Modellierung der Plasmasphärendynamik N2 - The plasmasphere is a dynamic region of cold, dense plasma surrounding the Earth. Its shape and size are highly susceptible to variations in solar and geomagnetic conditions. Having an accurate model of plasma density in the plasmasphere is important for GNSS navigation and for predicting hazardous effects of radiation in space on spacecraft. The distribution of cold plasma and its dynamic dependence on solar wind and geomagnetic conditions remain, however, poorly quantified. Existing empirical models of plasma density tend to be oversimplified as they are based on statistical averages over static parameters. Understanding the global dynamics of the plasmasphere using observations from space remains a challenge, as existing density measurements are sparse and limited to locations where satellites can provide in-situ observations. In this dissertation, we demonstrate how such sparse electron density measurements can be used to reconstruct the global electron density distribution in the plasmasphere and capture its dynamic dependence on solar wind and geomagnetic conditions. First, we develop an automated algorithm to determine the electron density from in-situ measurements of the electric field on the Van Allen Probes spacecraft. In particular, we design a neural network to infer the upper hybrid resonance frequency from the dynamic spectrograms obtained with the Electric and Magnetic Field Instrument Suite and Integrated Science (EMFISIS) instrumentation suite, which is then used to calculate the electron number density. The developed Neural-network-based Upper hybrid Resonance Determination (NURD) algorithm is applied to more than four years of EMFISIS measurements to produce the publicly available electron density data set. We utilize the obtained electron density data set to develop a new global model of plasma density by employing a neural network-based modeling approach. In addition to the location, the model takes the time history of geomagnetic indices and location as inputs, and produces electron density in the equatorial plane as an output. It is extensively validated using in-situ density measurements from the Van Allen Probes mission, and also by comparing the predicted global evolution of the plasmasphere with the global IMAGE EUV images of He+ distribution. The model successfully reproduces erosion of the plasmasphere on the night side as well as plume formation and evolution, and agrees well with data. The performance of neural networks strongly depends on the availability of training data, which is limited during intervals of high geomagnetic activity. In order to provide reliable density predictions during such intervals, we can employ physics-based modeling. We develop a new approach for optimally combining the neural network- and physics-based models of the plasmasphere by means of data assimilation. The developed approach utilizes advantages of both neural network- and physics-based modeling and produces reliable global plasma density reconstructions for quiet, disturbed, and extreme geomagnetic conditions. Finally, we extend the developed machine learning-based tools and apply them to another important problem in the field of space weather, the prediction of the geomagnetic index Kp. The Kp index is one of the most widely used indicators for space weather alerts and serves as input to various models, such as for the thermosphere, the radiation belts and the plasmasphere. It is therefore crucial to predict the Kp index accurately. Previous work in this area has mostly employed artificial neural networks to nowcast and make short-term predictions of Kp, basing their inferences on the recent history of Kp and solar wind measurements at L1. We analyze how the performance of neural networks compares to other machine learning algorithms for nowcasting and forecasting Kp for up to 12 hours ahead. Additionally, we investigate several machine learning and information theory methods for selecting the optimal inputs to a predictive model of Kp. The developed tools for feature selection can also be applied to other problems in space physics in order to reduce the input dimensionality and identify the most important drivers. Research outlined in this dissertation clearly demonstrates that machine learning tools can be used to develop empirical models from sparse data and also can be used to understand the underlying physical processes. Combining machine learning, physics-based modeling and data assimilation allows us to develop novel methods benefiting from these different approaches. N2 - Die Plasmasphäre ist eine die Erde umgebende dynamische Region aus kaltem, dichtem Plasma. Ihre Form und Größe sind sehr anfällig für Schwankungen der solaren und geomagnetischen Bedingungen. Ein präzises Modell der Plasmadichte in der Plasmasphäre ist wichtig für die GNSS-Navigation und für die Vorhersage gefährlicher Auswirkungen der kosmischen Strahlung auf Raumfahrzeuge. Die Verteilung des kalten Plasmas und seine dynamische Abhängigkeit vom Sonnenwind und den geomagnetischen Bedingungen sind jedoch nach wie vor nur unzureichend quantifiziert. Bestehende empirische Modelle der Plasmadichte sind in der Regel zu stark vereinfacht, da sie auf statistischen Durchschnittswerten statischer Parameter basieren. Das Verständnis der globalen Dynamik der Plasmasphäre anhand von Beobachtungen aus dem Weltraum bleibt eine Herausforderung, da vorhandene Dichtemessungen spärlich sind und sich auf Orte beschränken, an denen Satelliten In-situ-Beobachtungen liefern können. In dieser Dissertation zeigen wir, wie solche spärlichen Elektronendichtemessungen verwendet werden können, um die globale Elektronendichteverteilung in der Plasmasphäre zu rekonstruieren und ihre dynamische Abhängigkeit vom Sonnenwind und den geomagnetischen Bedingungen zu erfassen. Zunächst entwickeln wir einen automatisierten Algorithmus zur Bestimmung der Elektronendichte aus In-situ-Messungen des elektrischen Feldes der Van Allen Probes Raumsonden. Insbesondere entwerfen wir ein neuronales Netzwerk, um die obere Hybridresonanzfrequenz aus den dynamischen Spektrogrammen abzuleiten, die wir durch die Instrumentensuite „Electric and Magnetic Field Instrument Suite“ (EMFISIS) erhielten, welche dann zur Berechnung der Elektronenzahldichte verwendet wird. Der entwickelte „Neural-network-based Upper Hybrid Resonance Determination“ (NURD)-Algorithmus wird auf mehr als vier Jahre der EMFISIS-Messungen angewendet, um den öffentlich verfügbaren Elektronendichte-Datensatz zu erstellen. Wir verwenden den erhaltenen Elektronendichte-Datensatz, um ein neues globales Modell der Plasmadichte zu entwickeln, indem wir einen auf einem neuronalen Netzwerk basierenden Modellierungsansatz verwenden. Zusätzlich zum Ort nimmt das Modell den zeitlichen Verlauf der geomagnetischen Indizes und des Ortes als Eingabe und erzeugt als Ausgabe die Elektronendichte in der äquatorialebene. Dies wird ausführlich anhand von In-situ-Dichtemessungen der Van Allen Probes-Mission und durch den Vergleich der vom Modell vorhergesagten globalen Entwicklung der Plasmasphäre mit den globalen IMAGE EUV-Bildern der He+ -Verteilung validiert. Das Modell reproduziert erfolgreich die Erosion der Plasmasphäre auf der Nachtseite sowie die Bildung und Entwicklung von Fahnen und stimmt gut mit den Daten überein. Die Leistung neuronaler Netze hängt stark von der Verfügbarkeit von Trainingsdaten ab, die für Intervalle hoher geomagnetischer Aktivität nur spärlich vorhanden sind. Um zuverlässige Dichtevorhersagen während solcher Intervalle zu liefern, können wir eine physikalische Modellierung verwenden. Wir entwickeln einen neuen Ansatz zur optimalen Kombination der neuronalen Netzwerk- und physikbasierenden Modelle der Plasmasphäre mittels Datenassimilation. Der entwickelte Ansatz nutzt sowohl die Vorteile neuronaler Netze als auch die physikalischen Modellierung und liefert zuverlässige Rekonstruktionen der globalen Plasmadichte für ruhige, gestörte und extreme geomagnetische Bedingungen. Schließlich erweitern wir die entwickelten auf maschinellem Lernen basierten Werkzeuge und wenden sie auf ein weiteres wichtiges Problem im Bereich des Weltraumwetters an, die Vorhersage des geomagnetischen Index Kp. Der Kp-Index ist einer der am häufigsten verwendeten Indikatoren für Weltraumwetterwarnungen und dient als Eingabe für verschiedene Modelle, z.B. für die Thermosphäre, die Strahlungsgürtel und die Plasmasphäre. Es ist daher wichtig, den Kp-Index genau vorherzusagen. Frühere Arbeiten in diesem Bereich verwendeten hauptsächlich künstliche neuronale Netze, um Kurzzeit-Kp-Vorhersagen zu treffen, wobei deren Schlussfolgerungen auf der jüngsten Vergangenheit von Kp- und Sonnenwindmessungen am L1-Punkt beruhten. Wir analysieren, wie sich die Leistung neuronaler Netze im Vergleich zu anderen Algorithmen für maschinelles Lernen verhält, um kurz- und längerfristige Kp-Voraussagen von bis zu 12 Stunden treffen zu können. Zusätzlich untersuchen wir verschiedene Methoden des maschinellen Lernens und der Informationstheorie zur Auswahl der optimalen Eingaben für ein Vorhersagemodell von Kp. Die entwickelten Werkzeuge zur Merkmalsauswahl können auch auf andere Probleme in der Weltraumphysik angewendet werden, um die Eingabedimensionalität zu reduzieren und die wichtigsten Treiber zu identifizieren. Die in dieser Dissertation skizzierten Untersuchungen zeigen deutlich, dass Werkzeuge für maschinelles Lernen sowohl zur Entwicklung empirischer Modelle aus spärlichen Daten als auch zum Verstehen zugrunde liegender physikalischer Prozesse genutzt werden können. Die Kombination von maschinellem Lernen, physikbasierter Modellierung und Datenassimilation ermöglicht es uns, kombinierte Methoden zu entwickeln, die von unterschiedlichen Ansätzen profitieren. KW - Plasmasphere KW - Inner magnetosphere KW - Neural networks KW - Machine learning KW - Modeling KW - Kp index KW - Geomagnetic activity KW - Data assimilation KW - Validation KW - IMAGE EUV KW - Kalman filter KW - Plasmasphäre KW - Innere Magnetosphäre KW - Neuronale Netze KW - Maschinelles Lernen KW - Modellieren KW - Forecasting KW - Kp-Index KW - Geomagnetische Aktivität KW - Datenassimilation KW - Validierung KW - Kalman Filter KW - Prognose Y1 - 2020 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-482433 ER - TY - JOUR A1 - Reimann, Max A1 - Klingbeil, Mandy A1 - Pasewaldt, Sebastian A1 - Semmo, Amir A1 - Trapp, Matthias A1 - Döllner, Jürgen Roland Friedrich T1 - Locally controllable neural style transfer on mobile devices JF - The Visual Computer N2 - Mobile expressive rendering gained increasing popularity among users seeking casual creativity by image stylization and supports the development of mobile artists as a new user group. In particular, neural style transfer has advanced as a core technology to emulate characteristics of manifold artistic styles. However, when it comes to creative expression, the technology still faces inherent limitations in providing low-level controls for localized image stylization. In this work, we first propose a problem characterization of interactive style transfer representing a trade-off between visual quality, run-time performance, and user control. We then present MaeSTrO, a mobile app for orchestration of neural style transfer techniques using iterative, multi-style generative and adaptive neural networks that can be locally controlled by on-screen painting metaphors. At this, we enhance state-of-the-art neural style transfer techniques by mask-based loss terms that can be interactively parameterized by a generalized user interface to facilitate a creative and localized editing process. We report on a usability study and an online survey that demonstrate the ability of our app to transfer styles at improved semantic plausibility. KW - Non-photorealistic rendering KW - Style transfer KW - Neural networks KW - Mobile devices KW - Interactive control KW - Expressive rendering Y1 - 2019 U6 - https://doi.org/10.1007/s00371-019-01654-1 SN - 0178-2789 SN - 1432-2315 VL - 35 IS - 11 SP - 1531 EP - 1547 PB - Springer CY - New York ER - TY - JOUR A1 - Prasse, Paul A1 - Knaebel, Rene A1 - Machlica, Lukas A1 - Pevny, Tomas A1 - Scheffer, Tobias T1 - Joint detection of malicious domains and infected clients JF - Machine learning N2 - Detection of malware-infected computers and detection of malicious web domains based on their encrypted HTTPS traffic are challenging problems, because only addresses, timestamps, and data volumes are observable. The detection problems are coupled, because infected clients tend to interact with malicious domains. Traffic data can be collected at a large scale, and antivirus tools can be used to identify infected clients in retrospect. Domains, by contrast, have to be labeled individually after forensic analysis. We explore transfer learning based on sluice networks; this allows the detection models to bootstrap each other. In a large-scale experimental study, we find that the model outperforms known reference models and detects previously unknown malware, previously unknown malware families, and previously unknown malicious domains. KW - Machine learning KW - Neural networks KW - Computer security KW - Traffic data KW - Https traffic Y1 - 2019 U6 - https://doi.org/10.1007/s10994-019-05789-z SN - 0885-6125 SN - 1573-0565 VL - 108 IS - 8-9 SP - 1353 EP - 1368 PB - Springer CY - Dordrecht ER - TY - JOUR A1 - Beyreuther, Moritz A1 - Hammer, Conny A1 - Wassermann, Joachim A1 - Ohrnberger, Matthias A1 - Megies, Tobias T1 - Constructing a hidden Markov Model based earthquake detector: application to induced seismicity JF - Geophysical journal international N2 - The triggering or detection of seismic events out of a continuous seismic data stream is one of the key issues of an automatic or semi-automatic seismic monitoring system. In the case of dense networks, either local or global, most of the implemented trigger algorithms are based on a large number of active stations. However, in the case of only few available stations or small events, for example, like in monitoring volcanoes or hydrothermal power plants, common triggers often show high false alarms. In such cases detection algorithms are of interest, which show reasonable performance when operating even on a single station. In this context, we apply Hidden Markov Models (HMM) which are algorithms borrowed from speech recognition. However, many pitfalls need to be avoided to apply speech recognition technology directly to earthquake detection. We show the fit of the model parameters in an innovative way. State clustering is introduced to refine the intrinsically assumed time dependency of the HMMs and we explain the effect coda has on the recognition results. The methodology is then used for the detection of anthropogenicly induced earthquakes for which we demonstrate for a period of 3.9 months of continuous data that the single station HMM earthquake detector can achieve similar detection rates as a common trigger in combination with coincidence sums over two stations. To show the general applicability of state clustering we apply the proposed method also to earthquake classification at Mt. Merapi volcano, Indonesia. KW - Time-series analysis KW - Neural networks KW - fuzzy logic KW - Seismic monitoring and test-ban treaty verification KW - Volcano seismology Y1 - 2012 U6 - https://doi.org/10.1111/j.1365-246X.2012.05361.x SN - 0956-540X VL - 189 IS - 1 SP - 602 EP - 610 PB - Wiley-Blackwell CY - Malden ER -