Showing papers by "University of Paderborn published in 2018"

ESPNet: End-to-end speech processing toolkit

Abstract: This paper introduces a new open source platform for end-to-end speech processing named ESPnet. ESPnet mainly focuses on end-to-end automatic speech recognition (ASR), and adopts widely-used dynamic neural network toolkits, Chainer and PyTorch, as a main deep learning engine. ESPnet also follows the Kaldi ASR toolkit style for data processing, feature extraction/format, and recipes to provide a complete setup for speech recognition and other speech processing experiments. This paper explains a major architecture of this software platform, several important functionalities, which differentiate ESPnet from other open source ASR toolkits, and experimental results with major ASR benchmarks.

ESPnet: End-to-End Speech Processing Toolkit

Abstract: This paper introduces a new open source platform for end-to-end speech processing named ESPnet. ESPnet mainly focuses on end-to-end automatic speech recognition (ASR), and adopts widely-used dynamic neural network toolkits, Chainer and PyTorch, as a main deep learning engine. ESPnet also follows the Kaldi ASR toolkit style for data processing, feature extraction/format, and recipes to provide a complete setup for speech recognition and other speech processing experiments. This paper explains a major architecture of this software platform, several important functionalities, which differentiate ESPnet from other open source ASR toolkits, and experimental results with major ASR benchmarks.

Metasurface holography: from fundamentals to applications

Abstract: Holography has emerged as a vital approach to fully engineer the wavefronts of light since its invention dating back to the last century. However, the typically large pixel size, small field of view and limited space-bandwidth impose limitations in the on-demand high-performance applications, especially for three-dimensional displays and large-capacity data storage. Meanwhile, metasurfaces have shown great potential in controlling the propagation of light through the well-tailored scattering behavior of the constituent ultrathin planar elements with a high spatial resolution, making them suitable for holographic beam-shaping elements. Here, we review recent developments in the field of metasurface holography, from the classification of metasurfaces to the design strategies for both free-space and surface waves. By employing the concepts of holographic multiplexing, multiple information channels, such as wavelength, polarization state, spatial position and nonlinear frequency conversion, can be employed using metasurfaces. Meanwhile, the switchable metasurface holography by the integration of functional materials stimulates a gradual transition from passive to active elements. Importantly, the holography principle has become a universal and simple approach to solving inverse engineering problems for electromagnetic waves, thus allowing various related techniques to be achieved.

Multichannel vectorial holographic display and encryption.

Abstract: Since its invention, holography has emerged as a powerful tool to fully reconstruct the wavefronts of light including all the fundamental properties (amplitude, phase, polarization, wave vector, and frequency). For exploring the full capability for information storage/display and enhancing the encryption security of metasurface holograms, smart multiplexing techniques together with suitable metasurface designs are highly demanded. Here, we integrate multiple polarization manipulation channels for various spatial phase profiles into a single birefringent vectorial hologram by completely avoiding unwanted cross-talk. Multiple independent target phase profiles with quantified phase relations that can process significantly different information in different polarization states are realized within a single metasurface. For our metasurface holograms, we demonstrate high fidelity, large efficiency, broadband operation, and a total of twelve polarization channels. Such multichannel polarization multiplexing can be used for dynamic vectorial holographic display and can provide triple protection for optical security. The concept is appealing for applications of arbitrary spin to angular momentum conversion and various phase modulation/beam shaping elements.

On the Stability of DNA Origami Nanostructures in Low‐Magnesium Buffers

Abstract: DNA origami structures have great potential as functional platforms in various biomedical applications. Many applications, however, are incompatible with the high Mg2+ concentrations commonly believed to be a prerequisite for maintaining DNA origami integrity. Herein, we investigate DNA origami stability in low-Mg2+ buffers. DNA origami stability is found to crucially depend on the availability of residual Mg2+ ions for screening electrostatic repulsion. The presence of EDTA and phosphate ions may thus facilitate DNA origami denaturation by displacing Mg2+ ions from the DNA backbone and reducing the strength of the Mg2+ -DNA interaction, respectively. Most remarkably, these buffer dependencies are affected by DNA origami superstructure. However, by rationally selecting buffer components and considering superstructure-dependent effects, the structural integrity of a given DNA origami nanostructure can be maintained in conventional buffers even at Mg2+ concentrations in the low-micromolar range.

Finite-time blow-up in low-dimensional Keller–Segel systems with logistic-type superlinear degradation

Abstract: We consider radially symmetric solutions of the Keller–Segel system with generalized logistic source given by $$\begin{aligned} \left\{ \begin{array}{l} u_t = \Delta u - abla \cdot (u abla v) + \lambda u - \mu u^\kappa , \\ 0 = \Delta v - v + u, \end{array} \right. \qquad \qquad (\star ) \end{aligned}$$ under homogeneous Neumann boundary conditions in the ball $$\Omega =B_R(0) \subset \mathbb {R}^n$$ for $$n\ge 3$$ and $$R>0$$ , where $$\lambda \in \mathbb {R}, \mu >0$$ and $$\kappa >1$$ . Under the assumption that $$\begin{aligned} \kappa < \left\{ \begin{array}{ll} \frac{7}{6} &{}\quad \text {if } n\in \{3,4\}, \\ 1+ \frac{1}{2(n-1)} &{}\quad \text {if } n \ge 5, \end{array} \right. \end{aligned}$$ a condition on the initial data is derived which is seen to be sufficient to ensure the occurrence of finite-time blow-up for the corresponding solution of ( $$\star $$ ). Moreover, this criterion is shown to be mild enough so as to allow for the conclusion that in fact any positive continuous radial function on $$\overline{\Omega }$$ is the limit in $$L^1(\Omega )$$ of a sequence $$(u_{0k})_{k\in \mathbb {N}}$$ of continuous radial initial data which are such that for each $$k\in \mathbb {N}$$ the associated initial-boundary value problem for ( $$\star $$ ) exhibits a finite-time explosion phenomenon in the above sense. In particular, this apparently provides the first rigorous detection of blow-up in a superlinearly dampened but otherwise essentially original Keller–Segel system in the physically relevant three-dimensional case.

ML-Plan: Automated machine learning via hierarchical planning

Abstract: Automated machine learning (AutoML) seeks to automatically select, compose, and parametrize machine learning algorithms, so as to achieve optimal performance on a given task (dataset). Although current approaches to AutoML have already produced impressive results, the field is still far from mature, and new techniques are still being developed. In this paper, we present ML-Plan, a new approach to AutoML based on hierarchical planning. To highlight the potential of this approach, we compare ML-Plan to the state-of-the-art frameworks Auto-WEKA, auto-sklearn, and TPOT. In an extensive series of experiments, we show that ML-Plan is highly competitive and often outperforms existing approaches.

Structural stability of DNA origami nanostructures under application-specific conditions.

Abstract: With the introduction of the DNA origami technique, it became possible to rapidly synthesize almost arbitrarily shaped molecular nanostructures at nearly stoichiometric yields The technique furthermore provides absolute addressability in the sub-nm range, rendering DNA origami nanostructures highly attractive substrates for the controlled arrangement of functional species such as proteins, dyes, and nanoparticles Consequently, DNAorigami nanostructures have found applications in numerous areas of fundamental and applied research, ranging from drug delivery to biosensing to plasmonics to inorganic materials synthesis Since many of those applications rely on structurally intact, well-definedDNA origami shapes, the issue of DNA origami stability under numerous application-relevant environmental conditions has received increasing interest in the past few years In this mini-review we discuss the structural stability, denaturation, and degradation of DNA origami nanostructures under different conditions relevant to the fields of biophysics and biochemistry, biomedicine, and materials science, and the methods to improve their stability for desired applications

The Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants

Abstract: Reasoning is a crucial part of natural language argumentation. To comprehend an argument, one must analyze its warrant, which explains why its claim follows from its premises. As arguments are highly contextualized, warrants are usually presupposed and left implicit. Thus, the comprehension does not only require language understanding and logic skills, but also depends on common sense. In this paper we develop a methodology for reconstructing warrants systematically. We operationalize it in a scalable crowdsourcing process, resulting in a freely licensed dataset with warrants for 2k authentic arguments from news comments. On this basis, we present a new challenging task, the argument reasoning comprehension task. Given an argument with a claim and a premise, the goal is to choose the correct implicit warrant from two options. Both warrants are plausible and lexically close, but lead to contradicting claims. A solution to this task will define a substantial step towards automatic warrant reconstruction. However, experiments with several neural attention and language models reveal that current approaches do not suffice.

Why Reinvent the Wheel: Let's Build Question Answering Systems Together

Abstract: Modern question answering (QA) systems need to flexibly integrate a number of components specialised to fulfil specific tasks in a QA pipeline. Key QA tasks include Named Entity Recognition and Disambiguation, Relation Extraction, and Query Building. Since a number of different software components exist that implement different strategies for each of these tasks, it is a major challenge to select and combine the most suitable components into a QA system, given the characteristics of a question. We study this optimisation problem and train classifiers, which take features of a question as input and have the goal of optimising the selection of QA components based on those features. We then devise a greedy algorithm to identify the pipelines that include the suitable components and can effectively answer the given question. We implement this model within Frankenstein, a QA framework able to select QA components and compose QA pipelines. We evaluate the effectiveness of the pipelines generated by Frankenstein using the QALD and LC-QuAD benchmarks. These results not only suggest that Frankenstein precisely solves the QA optimisation problem but also enables the automatic composition of optimised QA pipelines, which outperform the static Baseline QA pipeline. Thanks to this flexible and fully automated pipeline generation process, new QA components can be easily included in Frankenstein, thus improving the performance of the generated pipelines.

Tailoring nonlinear processes for quantum optics with pulsed temporal-mode encodings

Abstract: The time-frequency degree of freedom is a powerful resource for implementing high-dimensional quantum information processing. In particular, field-orthogonal pulsed temporal modes offer a flexible framework compatible with both long-distance fiber networks and integrated waveguide devices. In order for this architecture to be fully utilized, techniques to reliably generate diverse quantum states of light and accurately measure complex temporal waveforms must be developed. To this end, nonlinear processes mediated by spectrally shaped pump pulses in group-velocity engineered waveguides and crystals provide a capable toolbox. In this review, we examine how tailoring the phase-matching conditions of parametric downconversion and sum-frequency generation allows for highly pure single-photon generation, flexible temporal-mode entanglement, and accurate measurement of time-frequency photon states. We provide an overview of experimental progress towards these goals and summarize challenges that remain in the field.

Direct Characterization of Ultrafast Energy-Time Entangled Photon Pairs.

Abstract: Energy-time entangled photons are critical in many quantum optical phenomena and have emerged as important elements in quantum information protocols. Entanglement in this degree of freedom often manifests itself on ultrafast time scales, making it very difficult to detect, whether one employs direct or interferometric techniques, as photon-counting detectors have insufficient time resolution. Here, we implement ultrafast photon counters based on nonlinear interactions and strong femtosecond laser pulses to probe energy-time entanglement in this important regime. Using this technique and single-photon spectrometers, we characterize all the spectral and temporal correlations of two entangled photons with femtosecond resolution. This enables the witnessing of energy-time entanglement using uncertainty relations and the direct observation of nonlocal dispersion cancellation on ultrafast time scales. These techniques are essential to understand and control the energy-time degree of freedom of light for ultrafast quantum optics.

NARA-WPE: A Python package for weighted prediction error dereverberation in Numpy and Tensorflow for online and offline processing

Abstract: NARA-WPE is a Python software package providing implementations of the weighted prediction error (WPE) dereverberation algorithm. WPE has been shown to be a highly effective tool for speech dereverberation, thus improving the perceptual quality of the signal and improving the recognition performance of downstream automatic speech recognition (ASR). It is suitable both for single-channel and multi-channel applications. The package consist of (1) a Numpy implementation which can easily be integrated into a custom Python toolchain, and (2) a TensorFlow implementation which allows integration into larger computational graphs and enables backpropagation through WPE to train more advanced front-ends. This package comprises of an iterative offline (batch) version, a block-online version, and a frame-online version which can be used in moderately low latency applications, e.g. digital speech assistants.

International Entrepreneurship: A Meta-Analysis on the Internationalization and Performance Relationship

Abstract: The article conducts a meta–analysis on the relationship between internationalization and firm performance in international entrepreneurship. Empirical evidence from 15,648 internationalizing entre...

Towards Encrypted MPC for Linear Constrained Systems

Abstract: We present an encrypted model predictive control (MPC) scheme for linear constrained systems. More precisely, we show that homomorphic encryption can be used to realize a secure and private cloud-based evaluation of an MPC if the corresponding piecewise affine control law is explicitly given.

Performance Assessment of IEEE 802.11p with an Open Source SDR-Based Prototype

Abstract: We present a complete simulation and experimentation framework for IEEE 802.11p. The core of the framework is a Software Defined Radio (SDR)-based Orthogonal Frequency Division Multiplexing (OFDM) transceiver that we validated extensively by means of simulations, interoperability tests, and, ultimately, by conducting a field test. Being SDR-based, the transceiver offers important benefits: It provides access to all data down to and including the physical layer, allowing for a better understanding of the system. Based on open and programmable hardware and software, the transceiver is completely transparent and all implementation details can be studied and, if needed, modified. Finally, it enables a seamless switch between simulations and experiments and, thus, helps to bridge the gap between theory and practice. Comparing the transceiver’s performance with independent results from simulations and experiments, we underline its potential to be used as a tool for further studies of IEEE 802.11p networks both in field operational tests as well as for simulation-based development of novel physical layer solutions. To make the framework accessible to fellow researchers and to allow reproduction of the results, we released it under an Open Source license.

Beyond the molecular movie: Dynamics of bands and bonds during a photoinduced phase transition

Abstract: Ultrafast nonequilibrium dynamics offer a route to study the microscopic interactions that govern macroscopic behavior. In particular, photoinduced phase transitions (PIPTs) in solids provide a test case for how forces, and the resulting atomic motion along a reaction coordinate, originate from a nonequilibrium population of excited electronic states. Using femtosecond photoemission, we obtain access to the transient electronic structure during an ultrafast PIPT in a model system: indium nanowires on a silicon(111) surface. We uncover a detailed reaction pathway, allowing a direct comparison with the dynamics predicted by ab initio simulations. This further reveals the crucial role played by localized photoholes in shaping the potential energy landscape and enables a combined momentum- and real-space description of PIPTs, including the ultrafast formation of chemical bonds.

Imaging through Nonlinear Metalens Using Second Harmonic Generation

Abstract: The abrupt phase change of light at metasurfaces provides high flexibility in wave manipulation without the need for accumulation of propagating phase through dispersive materials. In the linear optical regime, one important application field of metasurfaces is imaging by planar metalenses, which enables device miniaturization and aberration correction compared to conventional optical microlens systems. With the incorporation of nonlinear responses into passive metasurfaces, optical functionalities of metalenses are anticipated to be further enriched, leading to completely new application areas. Here, imaging with nonlinear metalenses that combine the function of an ultrathin planar lens with simultaneous frequency conversion is demonstrated. With such nonlinear metalenses, imaging of objects with near infrared light while the image appears in the second harmonic signal of visible frequency range is experimentally demonstrated. Furthermore, the functionality of these nonlinear metalenses can be modified by switching the handedness of the circularly polarized fundamental wave, leading to either real or virtual nonlinear image formation. Nonlinear metalenses not only enable infrared light imaging through a visible detector but also have the ability to modulate nonlinear optical responses through an ultrathin metasurface device while the fundamental wave remains unaffected, which offers the capability of nonlinear information processing with novel optoelectronic devices.

Retrieval of the Best Counterargument without Prior Topic Knowledge

Abstract: Given any argument on any controversial topic, how to counter it? This question implies the challenging retrieval task of finding the best counterargument. Since prior knowledge of a topic cannot be expected in general, we hypothesize the best counterargument to invoke the same aspects as the argument while having the opposite stance. To operationalize our hypothesis, we simultaneously model the similarity and dissimilarity of pairs of arguments, based on the words and embeddings of the arguments’ premises and conclusions. A salient property of our model is its independence from the topic at hand, i.e., it applies to arbitrary arguments. We evaluate different model variations on millions of argument pairs derived from the web portal idebate.org. Systematic ranking experiments suggest that our hypothesis is true for many arguments: For 7.6 candidates with opposing stance on average, we rank the best counterargument highest with 60% accuracy. Even among all 2801 test set pairs as candidates, we still find the best one about every third time.

Do Android taint analysis tools keep their promises

Abstract: In recent years, researchers have developed a number of tools to conduct taint analysis of Android applications. While all the respective papers aim at providing a thorough empirical evaluation, comparability is hindered by varying or unclear evaluation targets. Sometimes, the apps used for evaluation are not precisely described. In other cases, authors use an established benchmark but cover it only partially. In yet other cases, the evaluations differ in terms of the data leaks searched for, or lack a ground truth to compare against. All those limitations make it impossible to truly compare the tools based on those published evaluations. We thus present ReproDroid, a framework allowing the accurate comparison of Android taint analysis tools. ReproDroid supports researchers in inferring the ground truth for data leaks in apps, in automatically applying tools to benchmarks, and in evaluating the obtained results. We use ReproDroid to comparatively evaluate on equal grounds the six prominent taint analysis tools Amandroid, DIALDroid, DidFail, DroidSafe, FlowDroid and IccTA. The results are largely positive although four tools violate some promises concerning features and accuracy. Finally, we contribute to the area of unbiased benchmarking with a new and improved version of the open test suite DroidBench.

Event-Triggered Quantized Communication-Based Distributed Convex Optimization

Abstract: A NOVEL distributed algorithm based on multiple agents with continuous-time dynamics is proposed for a convex optimization problem where the objective function is the summation of local objective functions and the state of each agent is subject to a convex constraint set. Considering the limited bandwidth of the communication channels, we introduce a dynamic quantizer for each agent. To further save on communication costs, we develop an event-based broadcasting scheme for each agent. In comparison with algorithms that rely on continuous communication, the proposed algorithm serves to save communication expenditure by exploiting temporal and spatial aspects. Though a joint design of dynamic quantizers and event-trigger functions are under mild conditions, the states of the agents asymptotically approach the global optimal point with an adjustable error bound without incurring Zeno behavior.

Effect of Dietary Sugar Intake on Biomarkers of Subclinical Inflammation: A Systematic Review and Meta-Analysis of Intervention Studies

Abstract: It has been postulated that dietary sugar consumption contributes to increased inflammatory processes in humans, and that this may be specific to fructose (alone, in sucrose or in high-fructose corn syrup (HFCS)). Therefore, we conducted a meta-analysis and systematic literature review to evaluate the relevance of fructose, sucrose, HFCS, and glucose consumption for systemic levels of biomarkers of subclinical inflammation. MEDLINE, EMBASE, and Cochrane libraries were searched for controlled intervention studies that report the effects of dietary sugar intake on (hs)CRP, IL-6, IL-18, IL-1RA, TNF-α, MCP-1, sICAM-1, sE-selectin, or adiponectin. Included studies were conducted on adults or adolescents with ≥20 participants and ≥2 weeks duration. Thirteen studies investigating 1141 participants were included in the meta-analysis. Sufficient studies (≥3) to pool were only available for (hs)CRP. Using a random effects model, pooled effects of the interventions (investigated as mean difference (MD)) revealed no differences in (hs)CRP between fructose intervention and glucose control groups (MD: −0.03 mg/L (95% CI: −0.52, 0.46), I2 = 44%). Similarly, no differences were observed between HFCS and sucrose interventions (MD: 0.21 mg/L (−0.11, 0.53), I2 = 0%). The quality of evidence was evaluated using Nutrigrade, and was rated low for these two comparisons. The limited evidence available to date does not support the hypothesis that dietary fructose, as found alone or in HFCS, contributes more to subclinical inflammation than other dietary sugars.

Quantum-Limited Time-Frequency Estimation through Mode-Selective Photon Measurement.

Abstract: By projecting onto complex optical mode profiles, it is possible to estimate arbitrarily small separations between objects with quantum-limited precision, free of uncertainty arising from overlapping intensity profiles. Here we extend these techniques to the time-frequency domain using mode-selective sum-frequency generation with shaped ultrafast pulses. We experimentally resolve temporal and spectral separations between incoherent mixtures of single-photon level signals ten times smaller than their optical bandwidths with a tenfold improvement in precision over the intensity-only Cramer-Rao bound.

The Connection between NHC Ligand Count and Photophysical Properties in Fe(II) Photosensitizers: An Experimental Study

Abstract: Four homo- and heteroleptic complexes bearing both polypyridyl units and N-heterocyclic carbene (NHC) donor functions are studied as potential noble metal-free photosensitizers. The complexes [FeII(L1)(terpy)][PF6]2, [FeII(L2)2][PF6]2, [FeII(L1)(L3)][PF6]2, and [FeII(L3)2][PF6]2 (terpy = 2,2':6',2″ terpyridine, L1 = 2,6-bis[3-(2,6-diisopropylphenyl)imidazol-2-ylidene]pyridine, L2 = 2,6-bis[3-isopropylimidazol-2-ylidene]pyridine, L3 = 1-(2,2'-bipyridyl)-3-methylimidazol-2-ylidene) contain tridentate ligands of the C^N^C and N^N^C type, respectively, resulting in a Fe-NHC number between two and four. Thorough ground state characterization by single crystal diffraction, electrochemistry, valence-to-core X-ray emission spectroscopy (VtC-XES), and high energy resolution fluorescence detected X-ray absorption near edge structure (HERFD-XANES) in combination with ab initio calculations show a correlation between the geometric and electronic structure of these new compounds and the number of the NHC donor functions. These results serve as a basis for the investigation of the excited states by ultrafast transient absorption spectroscopy, where the lifetime of the 3MLCT states is found to increase with the NHC donor count. The results demonstrate for the first time the close interplay between the number of NHC functionalities in Fe(II) complexes and their photochemical properties, as revealed in a comparison of the activity as photosensitizers in photocatalytic proton reduction.