Search

Humans excel at building generalizations of new concepts from just one single example. Contrary to this, current computer vision models typically require large amount of training samples to achieve a comparable accuracy. In this work we present a Bayesian model of perception that learns using only minimal data, a prototypical probabilistic program of an object. Specifically, we propose a generative inverse graphics model of primitive shapes, to infer posterior distributions over physically consistent parameters from one or several images. We show how this representation can be used for downstream tasks such as few-shot classification and pose estimation. Our model outperforms existing few-shot neural-only classification algorithms and demonstrates generalization across varying lighting conditions, backgrounds, and out-of-distribution shapes. By design, our model is uncertainty-aware and uses our new differentiable renderer for optimizing global scene parameters through gradient descent, sampling posterior distributions over object parameters with Markov Chain Monte Carlo (MCMC), and using a neural based likelihood function.

Assembly state recognition facilitates the execution of assembly procedures, offering feedback to enhance efficiency and minimize errors. However, recognizing assembly states poses challenges in scalability, since parts are frequently updated, and the robustness to execution errors remains underexplored. To address these challenges, this paper proposes an approach based on representation learning and the novel intermediate-state informed loss function modification (ISIL). ISIL leverages unlabeled transitions between states and demonstrates significant improvements in clustering and classification performance for all tested architectures and losses. Despite being trained exclusively on images without execution errors, thorough analysis on error states demonstrates that our approach accurately distinguishes between correct states and states with various types of execution errors. The integration of the proposed algorithm can offer meaningful assistance to workers and mitigate unexpected losses due to procedural mishaps in industrial settings. The code is available at: https://timschoonbeek.github.io/state_rec, Comment: 8 pages, 8 figures

Nanomaterials bring to expression unique electronic properties that promote advanced functionality and technologies. Albeit, nanoscale growth presents paramount challenges for synthesis limiting the diversity in structures and compositions. Here, we demonstrate solid-state topotactic exchange that converts Wurtzite InAs nanowires into Zintl phase Eu$_3$In$_2$As$_4$ nanowires. In situ evaporation of Eu and As over InAs nanowire cores in molecular beam epitaxy results in mutual exchange of Eu from the shell and In from the core. A continuous Eu$_3$In$_2$As$_4$ shell thereby grows that gradually consumes the InAs core and converts it into a single phase Eu$_3$In$_2$As$_4$ nanowire. Topotaxy, which facilitates the mutual exchange, is supported by the substructure of the As matrix which is similar across the Wurtzite InAs and Zintl Eu$_3$In$_2$As$_4$. We provide initial evidence of an antiferromagnetic transition at T$_N$ $\sim$ 6.5 K in the Zintl phase Eu$_3$In$_2$As$_4$ nanowires. Ab initio calculation confirms the antiferromagnetic state and classifies Eu$_3$In$_2$As$_4$ as a $C_2 T$ axion insulator hosting both chiral hinge modes and unpinned Dirac surface states. The topotactic mutual-exchange growth of Zintl Eu$_3$In$_2$As$_4$ nanowires thus enables the exploration of intricate magneto-topological states of nanomaterials. Moreover, it may open the path for topotactic mutual-exchange synthesis of nanowires made of other exotic compounds., Comment: 45 pages, 23 figures

The practical use of Bayesian Optimization (BO) in engineering applications imposes special requirements: high sampling efficiency on the one hand and finding a robust solution on the other hand. We address the case of adversarial robustness, where all parameters are controllable during the optimization process, but a subset of them is uncontrollable or even adversely perturbed at the time of application. To this end, we develop an efficient information-based acquisition function that we call Robust Entropy Search (RES). We empirically demonstrate its benefits in experiments on synthetic and real-life data. The results showthat RES reliably finds robust optima, outperforming state-of-the-art algorithms.

Renewable hydrogen storage in salt caverns requires fast injection and production rates to cope with the imbalance between energy production and consumption. Such operational conditions raise concerns about the mechanical stability of salt caverns. Choosing an appropriate constitutive model for salt mechanics is an important step in investigating this issue, and many constitutive models with several parameters have been presented in the literature. However, a robust calibration strategy to reliably determine which model and which parameter set represent the given rock, based on stress-strain data, remains an unsolved challenge. For the first time in the community, we present a multi-step strategy to determine a single parameter set based on many deformation datasets for salt rocks. Towards this end, we first develop a comprehensive constitutive model able to capture all relevant nonlinear deformation physics of transient, reverse, and steady-state creep. The determination of the single set of representative material parameters is achieved by framing the calibration process as an optimization problem, for which the global PSO algorithm is employed. Dynamic data integration is achieved by a multi-step calibration strategy for a situation where experiments are included one at a time, as they become available. Additionally, our calibration strategy is made flexible to account for mild heterogeneity between rock samples, resulting in a single set of parameters that is representative of the deformation datasets. As a rigorous mathematical analysis and the lack of relevant experimental datasets, we consider a wide range of synthetic experimental data, inspired by the existing sparse relevant data in the literature. The results of our performance analyses show that the proposed calibration strategy is robust and accuracy is improved as more experiments are included for calibration., Comment: 27 pages, 23 figures

In computer vision, a larger effective receptive field (ERF) is associated with better performance. While attention natively supports global context, its quadratic complexity limits its applicability to tasks that benefit from high-resolution input. In this work, we extend Hyena, a convolution-based attention replacement, from causal sequences to bidirectional data and two-dimensional image space. We scale Hyena's convolution kernels beyond the feature map size, up to 191$\times$191, to maximize ERF while maintaining sub-quadratic complexity in the number of pixels. We integrate our two-dimensional Hyena, HyenaPixel, and bidirectional Hyena into the MetaFormer framework. For image categorization, HyenaPixel and bidirectional Hyena achieve a competitive ImageNet-1k top-1 accuracy of 84.9% and 85.2%, respectively, with no additional training data, while outperforming other convolutional and large-kernel networks. Combining HyenaPixel with attention further improves accuracy. We attribute the success of bidirectional Hyena to learning the data-dependent geometric arrangement of pixels without a fixed neighborhood definition. Experimental results on downstream tasks suggest that HyenaPixel with large filters and a fixed neighborhood leads to better localization performance.

Understanding the relation between cortical neuronal network structure and neuronal activity is a fundamental unresolved question in neuroscience, with implications to our understanding of the mechanism by which neuronal networks evolve over time, spontaneously or under stimulation. It requires a method for inferring the structure and composition of a network from neuronal activities. Tracking the evolution of networks and their changing functionality will provide invaluable insight into the occurrence of plasticity and the underlying learning process. We devise a probabilistic method for inferring the effective network structure by integrating techniques from Bayesian statistics, statistical physics and principled machine learning. The method and resulting algorithm allow one to infer the effective network structure, identify the excitatory and inhibitory nature of its constituents, and predict neuronal spiking activities by employing the inferred structure. We validate the method and algorithm's performance using synthetic data, spontaneous activity of an in silico emulator and realistic in vitro neuronal networks of modular and homogeneous connectivity, demonstrating excellent structure inference and activity prediction. We also show that our method outperforms commonly used existing methods for inferring neuronal network structure. Inferring the evolving effective structure of neuronal networks will provide new insight into the learning process due to stimulation in general and will facilitate the development of neuron-based circuits with computing capabilities.

The properties of complex networked systems arise from the interplay between the dynamics of their elements and the underlying topology. Thus, to understand their behaviour, it is crucial to convene as much information as possible about their topological organization. However, in a large systems such as neuronal networks, the reconstruction of such topology is usually carried out from the information encoded in the dynamics on the network, such as spike train time series, and by measuring the Transfer Entropy between system elements. The topological information recovered by these methods does not necessarily capture the connectivity layout, but rather the causal flow of information between elements. New theoretical frameworks, such as Integrated Information Decomposition ($\Phi$-ID), allow to explore the modes in which information can flow between parts of a system, opening a rich landscape of interactions between network topology, dynamics and information. Here, we apply $\Phi$-ID on $in$ $silico$ and $in$ $vitro$ data to decompose the usual Transfer Entropy measure into different modes of information transfer, namely synergistic, redundant or unique. We demonstrate that the unique information transfer is the most relevant measure to uncover structural topological details from network activity data, while redundant information only introduces residual information for this application. Although the retrieved network connectivity is still functional, it captures more details of the underlying structural topology by avoiding to take into account emergent high-order interactions and information redundancy between elements, which are important for the functional behavior, but mask the detection of direct simple interactions between elements constituted by the structural network topology., Comment: 13 pages, 5 figures

Although action recognition for procedural tasks has received notable attention, it has a fundamental flaw in that no measure of success for actions is provided. This limits the applicability of such systems especially within the industrial domain, since the outcome of procedural actions is often significantly more important than the mere execution. To address this limitation, we define the novel task of procedure step recognition (PSR), focusing on recognizing the correct completion and order of procedural steps. Alongside the new task, we also present the multi-modal IndustReal dataset. Unlike currently available datasets, IndustReal contains procedural errors (such as omissions) as well as execution errors. A significant part of these errors are exclusively present in the validation and test sets, making IndustReal suitable to evaluate robustness of algorithms to new, unseen mistakes. Additionally, to encourage reproducibility and allow for scalable approaches trained on synthetic data, the 3D models of all parts are publicly available. Annotations and benchmark performance are provided for action recognition and assembly state detection, as well as the new PSR task. IndustReal, along with the code and model weights, is available at: https://github.com/TimSchoonbeek/IndustReal ., Comment: Accepted for WACV 2024. 15 pages, 9 figures, including supplementary materials

Artificial Intelligence (AI) has made impressive progress in recent years and represents a key technology that has a crucial impact on the economy and society. However, it is clear that AI and business models based on it can only reach their full potential if AI applications are developed according to high quality standards and are effectively protected against new AI risks. For instance, AI bears the risk of unfair treatment of individuals when processing personal data e.g., to support credit lending or staff recruitment decisions. The emergence of these new risks is closely linked to the fact that the behavior of AI applications, particularly those based on Machine Learning (ML), is essentially learned from large volumes of data and is not predetermined by fixed programmed rules. Thus, the issue of the trustworthiness of AI applications is crucial and is the subject of numerous major publications by stakeholders in politics, business and society. In addition, there is mutual agreement that the requirements for trustworthy AI, which are often described in an abstract way, must now be made clear and tangible. One challenge to overcome here relates to the fact that the specific quality criteria for an AI application depend heavily on the application context and possible measures to fulfill them in turn depend heavily on the AI technology used. Lastly, practical assessment procedures are needed to evaluate whether specific AI applications have been developed according to adequate quality standards. This AI assessment catalog addresses exactly this point and is intended for two target groups: Firstly, it provides developers with a guideline for systematically making their AI applications trustworthy. Secondly, it guides assessors and auditors on how to examine AI applications for trustworthiness in a structured way.

Fatigue strength estimation is a costly manual material characterization process in which state-of-the-art approaches follow a standardized experiment and analysis procedure. In this paper, we examine a modular, Machine Learning-based approach for fatigue strength estimation that is likely to reduce the number of experiments and, thus, the overall experimental costs. Despite its high potential, deployment of a new approach in a real-life lab requires more than the theoretical definition and simulation. Therefore, we study the robustness of the approach against misspecification of the prior and discretization of the specified loads. We identify its applicability and its advantageous behavior over the state-of-the-art methods, potentially reducing the number of costly experiments., Comment: 2nd Annual AAAI Workshop on AI to Accelerate Science and Engineering (AI2ASE)

We show how the Weil pairing can be used to evaluate the assigned characters of an imaginary quadratic order $\mathcal{O}$ in an unknown ideal class $[\mathfrak{a}] \in \mathrm{Cl}(\mathcal{O})$ that connects two given $\mathcal{O}$-oriented elliptic curves $(E, \iota)$ and $(E', \iota') = [\mathfrak{a}](E, \iota)$. When specialized to ordinary elliptic curves over finite fields, our method is conceptually simpler and often somewhat faster than a recent approach due to Castryck, Sot\'akov\'a and Vercauteren, who rely on the Tate pairing instead. The main implication of our work is that it breaks the decisional Diffie-Hellman problem for practically all oriented elliptic curves that are acted upon by an even-order class group. It can also be used to better handle the worst cases in Wesolowski's recent reduction from the vectorization problem for oriented elliptic curves to the endomorphism ring problem, leading to a method that always works in sub-exponential time., Comment: 18 pp

Let $\sigma$ denote an endomorphism of a smooth algebraic group $G$ over the algebraic closure of a finite field, and assume all iterates of $\sigma$ have finitely many fixed points. Steinberg gave a formula for the number of fixed points of $\sigma$ (and hence of all of its iterates $\sigma^n$) in the semisimple case, leading to a representation of its Artin-Mazur zeta function as a rational function. We generalise this to an arbitrary (smooth) algebraic group $G$, where the number of fixed points $\sigma_n$ of $\sigma^n$ can depend on $p$-adic properties of $n$. We axiomatise the structure of the sequence $(\sigma_n)$ via the concept of a `finite-adelically distorted' (FAD-)sequence. Such sequences also occur in topological dynamics, and our subsequent results about zeta functions and asymptotic counting of orbits apply equally well in that situation; for example, to $S$-integer dynamical systems, additive cellular automata and other compact abelian groups. We prove dichotomies for the associated Artin-Mazur zeta function, and study the analogue of the Prime Number Theorem for the function counting periodic orbits of length $\leq N$. For an algebraic group $G$ we express the error term via the $\ell$-adic cohomological zeta function of $G$., Comment: 176 pages, 6 figures, v2: reorganised, extended some of the basic material

The Hilbert class polynomial has as roots the j-invariants of elliptic curves whose endomorphism ring is a given imaginary quadratic order. It can be used to compute elliptic curves over finite fields with a prescribed number of points. Since its coefficients are typically rather large, there has been continued interest in finding alternative modular functions whose corresponding class polynomials are smaller. Best known are Weber's functions, that reduce the size by a factor of 72 for a positive density subset of imaginary quadratic discriminants. On the other hand, Br\"oker and Stevenhagen showed that no modular function will ever do better than a factor of 100.83. We introduce a generalization of class polynomials, with reduction factors that are not limited by the Br\"oker-Stevenhagen bound. We provide examples matching Weber's reduction factor. For an infinite family of discriminants, their reduction factors surpass those of all previously known modular functions by a factor at least 2., Comment: 28 pp. 5 fig

When training data is scarce, the incorporation of additional prior knowledge can assist the learning process. While it is common to initialize neural networks with weights that have been pre-trained on other large data sets, pre-training on more concise forms of knowledge has rather been overlooked. In this paper, we propose a novel informed machine learning approach and suggest to pre-train on prior knowledge. Formal knowledge representations, e.g. graphs or equations, are first transformed into a small and condensed data set of knowledge prototypes. We show that informed pre-training on such knowledge prototypes (i) speeds up the learning processes, (ii) improves generalization capabilities in the regime where not enough training data is available, and (iii) increases model robustness. Analyzing which parts of the model are affected most by the prototypes reveals that improvements come from deeper layers that typically represent high-level features. This confirms that informed pre-training can indeed transfer semantic knowledge. This is a novel effect, which shows that knowledge-based pre-training has additional and complementary strengths to existing approaches.

The availability of representative datasets is an essential prerequisite for many successful artificial intelligence and machine learning models. However, in real life applications these models often encounter scenarios that are inadequately represented in the data used for training. There are various reasons for the absence of sufficient data, ranging from time and cost constraints to ethical considerations. As a consequence, the reliable usage of these models, especially in safety-critical applications, is still a tremendous challenge. Leveraging additional, already existing sources of knowledge is key to overcome the limitations of purely data-driven approaches. Knowledge augmented machine learning approaches offer the possibility of compensating for deficiencies, errors, or ambiguities in the data, thus increasing the generalization capability of the applied models. Even more, predictions that conform with knowledge are crucial for making trustworthy and safe decisions even in underrepresented scenarios. This work provides an overview of existing techniques and methods in the literature that combine data-driven models with existing knowledge. The identified approaches are structured according to the categories knowledge integration, extraction and conformity. In particular, we address the application of the presented methods in the field of autonomous driving., Comment: 111 pages, Added section on Run-time Network Verification

How can humans remain in control of artificial intelligence (AI)-based systems designed to perform tasks autonomously? Such systems are increasingly ubiquitous, creating benefits - but also undesirable situations where moral responsibility for their actions cannot be properly attributed to any particular person or group. The concept of meaningful human control has been proposed to address responsibility gaps and mitigate them by establishing conditions that enable a proper attribution of responsibility for humans; however, clear requirements for researchers, designers, and engineers are yet inexistent, making the development of AI-based systems that remain under meaningful human control challenging. In this paper, we address the gap between philosophical theory and engineering practice by identifying, through an iterative process of abductive thinking, four actionable properties for AI-based systems under meaningful human control, which we discuss making use of two applications scenarios: automated vehicles and AI-based hiring. First, a system in which humans and AI algorithms interact should have an explicitly defined domain of morally loaded situations within which the system ought to operate. Second, humans and AI agents within the system should have appropriate and mutually compatible representations. Third, responsibility attributed to a human should be commensurate with that human's ability and authority to control the system. Fourth, there should be explicit links between the actions of the AI agents and actions of humans who are aware of their moral responsibility. We argue that these four properties will support practically-minded professionals to take concrete steps toward designing and engineering for AI systems that facilitate meaningful human control., Comment: Preprint. Published AI and Ethics (2022): https://doi.org/10.1007/s43681-022-00167-3

The high-intensity time-of-flight neutron diffractometer POWTEX for powder and texture analysis is currently being built to be operated in the eastern guide hall of the research reactor FRM II near Munich. Having heavily faced the world-wide 3He crisis in 2009, we promptly initiated the development of 3He-free detector alternatives that are tailor-made for the requirements of large-area diffractometers. Herein, we report about the 2017 enterprise to operate one building unit of the final POWTEX detector at the neutron powder diffractometer POWGEN at the Spallation Neutron Source of the Oak Ridge National Lab, USA. As a result, we present the first angular- and wavelength-dependent data of the POWTEX detector, current data-reduction steps as implemented in Mantid and also multi-dimensional refinement results of two samples (diamond and BaZn(NCN)2) by use of the GSAS-II software suite.

Many machine learning applications can benefit from simulated data for systematic validation - in particular if real-life data is difficult to obtain or annotate. However, since simulations are prone to domain shift w.r.t. real-life data, it is crucial to verify the transferability of the obtained results. We propose a novel framework consisting of a generative label-to-image synthesis model together with different transferability measures to inspect to what extent we can transfer testing results of semantic segmentation models from synthetic data to equivalent real-life data. With slight modifications, our approach is extendable to, e.g., general multi-class classification tasks. Grounded on the transferability analysis, our approach additionally allows for extensive testing by incorporating controlled simulations. We validate our approach empirically on a semantic segmentation task on driving scenes. Transferability is tested using correlation analysis of IoU and a learned discriminator. Although the latter can distinguish between real-life and synthetic tests, in the former we observe surprisingly strong correlations of 0.7 for both cars and pedestrians., Comment: The first two authors contributed equally. Accepted at the 4th Workshop on "Ensuring and Validating Safety for Automated Vehicles" (WS13), IV2021. Under IEEE Copyright

This survey presents an overview of integrating prior knowledge into machine learning systems in order to improve explainability. The complexity of machine learning models has elicited research to make them more explainable. However, most explainability methods cannot provide insight beyond the given data, requiring additional information about the context. We propose to harness prior knowledge to improve upon the explanation capabilities of machine learning models. In this paper, we present a categorization of current research into three main categories which either integrate knowledge into the machine learning pipeline, into the explainability method or derive knowledge from explanations. To classify the papers, we build upon the existing taxonomy of informed machine learning and extend it from the perspective of explainability. We conclude with open challenges and research directions.

The use of deep neural networks (DNNs) in safety-critical applications like mobile health and autonomous driving is challenging due to numerous model-inherent shortcomings. These shortcomings are diverse and range from a lack of generalization over insufficient interpretability to problems with malicious inputs. Cyber-physical systems employing DNNs are therefore likely to suffer from safety concerns. In recent years, a zoo of state-of-the-art techniques aiming to address these safety concerns has emerged. This work provides a structured and broad overview of them. We first identify categories of insufficiencies to then describe research activities aiming at their detection, quantification, or mitigation. Our paper addresses both machine learning experts and safety engineers: The former ones might profit from the broad range of machine learning topics covered and discussions on limitations of recent methods. The latter ones might gain insights into the specifics of modern ML methods. We moreover hope that our contribution fuels discussions on desiderata for ML systems and strategies on how to propel existing approaches accordingly., Comment: 94 pages

An important pillar for safe machine learning (ML) is the systematic mitigation of weaknesses in neural networks to afford their deployment in critical applications. An ubiquitous class of safety risks are learned shortcuts, i.e. spurious correlations a network exploits for its decisions that have no semantic connection to the actual task. Networks relying on such shortcuts bear the risk of not generalizing well to unseen inputs. Explainability methods help to uncover such network vulnerabilities. However, many of these techniques are not directly applicable if access to the network is constrained, in so-called black-box setups. These setups are prevalent when using third-party ML components. To address this constraint, we present an approach to detect learned shortcuts using an interpretable-by-design network as a proxy to the black-box model of interest. Leveraging the proxy's guarantees on introspection we automatically extract candidates for learned shortcuts. Their transferability to the black box is validated in a systematic fashion. Concretely, as proxy model we choose a BagNet, which bases its decisions purely on local image patches. We demonstrate on the autonomous driving dataset A2D2 that extracted patch shortcuts significantly influence the black box model. By efficiently identifying such patch-based vulnerabilities, we contribute to safer ML models., Comment: Under IEEE Copyright; accepted at the SAIAD (Safe Artificial Intelligence for Automated Driving) Workshop at CVPR 2021

Data-driven sensor interpretation in autonomous driving can lead to highly implausible predictions as can most of the time be verified with common-sense knowledge. However, learning common knowledge only from data is hard and approaches for knowledge integration are an active research area. We propose to use a partly human-designed, partly learned set of rules to describe relations between objects of a traffic scene on a high level of abstraction. In doing so, we improve and robustify existing deep neural networks consuming low-level sensor information. We present an initial study adapting the well-established Probabilistic Soft Logic (PSL) framework to validate and improve on the problem of semantic segmentation. We describe in detail how we integrate common knowledge into the segmentation pipeline using PSL and verify our approach in a set of experiments demonstrating the increase in robustness against several severe image distortions applied to the A2D2 autonomous driving data set., Comment: Published at SAIAD (Safe Artificial Intelligence for Automated Driving) workshop at CVPR2021

The first-passage-time distribution of a leaky integrate-and-fire neuron driven by a characteristically coloured noise is approximated by matching a transient and a steady-state solution of the membrane voltage distribution. These approximations follow from a simple manipulation, made possible by the specific `eigen' colouring of the noise, which allows to express the membrane potential as a Gaussian diffusion process on top of a deterministic exponential movement. Following, the presented method is extended to the case of an arbitrarily coloured noise driving by factoring out the `eigen' noise and replacing the residue with an equivalent Gaussian process. It is shown that the obtained expressions agree well with numerical simulations for different values of the neuron parameters and noise colouring., Comment: 11 pages, 3 figures

An expression for the group delay of the FitzHugh-Nagumo model in response to low amplitude input is obtained by linearisation of the cubic term of the voltage equation around its stable fixed-point. It is found that a negative group delay exists for low frequencies, indicating that the evolution of slowly fluctuating signals are anticipated by the voltage dynamics. The effects of the group delay for different types of signals are shown numerically for the non-linearised FitzHugh-Nagumo model, and some observations on the signal aspects that are anticipated are stated., Comment: 9 pages, 6 figures

Neurons are connected to other neurons by axons and dendrites that conduct signals with finite velocities, resulting in delays between the firing of a neuron and the arrival of the resultant impulse at other neurons. Since delays greatly complicate the analytical treatment and interpretation of models, they are usually neglected or taken to be uniform, leading to a lack in the comprehension of the effects of delays in neural systems. This paper shows that heterogeneous transmission delays make small groups of neurons respond selectively to inputs with differing frequency spectra. By studying a single integrate-and-fire neuron receiving correlated time-shifted inputs, it is shown how the frequency response is linked to both the strengths and delay times of the afferent connections. The results show that incorporating delays alters the functioning of neural networks, and changes the effect that neural connections and synaptic strengths have., Comment: 11 pages, 6 figures

Metal organic chemical vapor deposition (MOCVD) is one of the main methodologies used for thin film fabrication in the semiconductor industry today and is considered one of the most promising routes to achieve large-scale and high-quality 2D transition metal dichalcogenides (TMDCs). However, if not taken special measures, MOCVD suffers from some serious drawbacks, such as small domain size and carbon contamination, resulting in poor optical and crystal quality, which may inhibit its implementation for the large-scale fabrication of atomic-thin semiconductors. Here we present a Growth-Etch MOCVD (GE-MOCVD) methodology, in which a small amount of water vapor is introduced during the growth, while the precursors are delivered in pulses. The evolution of the growth as a function of the amount of water vapor, the number and type of cycles and the gas composition is described. We show a significant domain size increase is achieved relative to our conventional process. The improved crystal quality of WS2 (and WSe2) domains was demonstrated by means of Raman spectroscopy, photoluminescence (PL) spectroscopy and HRTEM studies. Moreover, time-resolved PL studies show very long exciton lifetimes, comparable to those observed in mechanically exfoliated flakes. Thus, the GE-MOCVD approach presented here may facilitate their integration into a wide range of applications., Comment: 29 pages, 6 figures

Monte Carlo (MC) dropout is one of the state-of-the-art approaches for uncertainty estimation in neural networks (NNs). It has been interpreted as approximately performing Bayesian inference. Based on previous work on the approximation of Gaussian processes by wide and deep neural networks with random weights, we study the limiting distribution of wide untrained NNs under dropout more rigorously and prove that they as well converge to Gaussian processes for fixed sets of weights and biases. We sketch an argument that this property might also hold for infinitely wide feed-forward networks that are trained with (full-batch) gradient descent. The theory is contrasted by an empirical analysis in which we find correlations and non-Gaussian behaviour for the pre-activations of finite width NNs. We therefore investigate how (strongly) correlated pre-activations can induce non-Gaussian behavior in NNs with strongly correlated weights., Comment: Accepted at the ICML 2020 workshop for Uncertainty and Robustness in Deep Learning

Type 2 Diabetes (T2D) is a chronic metabolic disorder that can lead to blindness and cardiovascular disease. Information about early stage T2D might be present in retinal fundus images, but to what extent these images can be used for a screening setting is still unknown. In this study, deep neural networks were employed to differentiate between fundus images from individuals with and without T2D. We investigated three methods to achieve high classification performance, measured by the area under the receiver operating curve (ROC-AUC). A multi-target learning approach to simultaneously output retinal biomarkers as well as T2D works best (AUC = 0.746 [$\pm$0.001]). Furthermore, the classification performance can be improved when images with high prediction uncertainty are referred to a specialist. We also show that the combination of images of the left and right eye per individual can further improve the classification performance (AUC = 0.758 [$\pm$0.003]), using a simple averaging approach. The results are promising, suggesting the feasibility of screening for T2D from retinal fundus images., Comment: to be published in the proceeding of SPIE - Medical Imaging 2020, 6 pages, 1 figure

Recent years have seen a growing interest in strong coupling between plasmons and excitons, as a way to generate new quantum optical testbeds and influence chemical dynamics and reactivity. Strong coupling to bright plasmonic modes has been achieved even with single quantum emitters. Dark plasmonic modes fare better in some applications due to longer lifetimes, but are difficult to probe as they are subradiant. Here, we apply electron energy loss (EEL) spectroscopy to demonstrate that a dark mode of an individual plasmonic bowtie can interact with a small number of quantum emitters, as evidenced by Rabi-split spectra. Coupling strengths of up to 85 meV place the bowtie-emitter devices at the onset of the strong coupling regime. Remarkably, the coupling occurs at the bowtie gap periphery, even while the electron beam probes their center. Our findings pave the way for using EEL spectroscopy to study exciton-plasmon interactions involving non-emissive photonic modes.

We study fixed points of iterates of dynamically affine maps (a generalisation of Latt\`es maps) over algebraically closed fields of positive characteristic $p$. We present and study certain hypotheses that imply a dichotomy for the Artin-Mazur zeta function of the dynamical system: it is either rational or non-holonomic, depending on specific characteristics of the map. We also study the algebraicity of the so-called tame zeta function, the generating function for periodic points of order coprime to $p$. We then verify these hypotheses for dynamically affine maps on the projective line, generalising previous work of Bridy, and, in arbitrary dimension, for maps on Kummer varieties arising from multiplication by integers on abelian varieties., Comment: Lois van der Meijden co-authored Appendix B. 31 pp

While repeating fast radio bursts (FRBs) remain scarce in number, they provide a unique opportunity for follow-up observations that enhance our knowledge of their sources and potentially of the FRB population as a whole. Attaining more burst spectra could lead to a better understanding of the origin of these bright, millisecond-duration radio pulses. We therefore performed $\sim$20 hr of simultaneous observations on FRB 121102 with the Effelsberg 100-m radio telescope and the Low Frequency Array (LOFAR) to constrain the spectral behaviour of bursts from FRB 121102 at 1.4 GHz and 150 MHz. This campaign resulted in the detection of nine new bursts at 1.4 GHz but no simultaneous detections with LOFAR. Assuming that the ratio of the fluence at two frequencies scales as a power law, we placed a lower limit of $\alpha$ > -1.2 $\pm$ 0.4 on the spectral index for the fluence of the instantaneous broad band emission of FRB 121102. For the derivation of this limit, a realistic fluence detection threshold for LOFAR was determined empirically assuming a burst would be scattered as predicted by the NE2001 model. A significant variation was observed in the burst repeat rate R at L-band. During observations in September 2016, nine bursts were detected, giving R = 1.1 $\pm$ 0.4 hr$^{-1}$, while in November no bursts were detected, yielding R < 0.3 hr$^{-1}$ (95% confidence limit). This variation is consistent with earlier seen episodic emission of FRB 121102. In a blind and targeted search, no bursts were found with LOFAR at 150 MHz, resulting in a repeat rate limit of R < 0.16 hr$^{-1}$ (95% confidence limit). Burst repeat rate ratios of FRB 121102 at 3, 2, 1.4, and 0.15 GHz are consistent within the uncertainties with a flattening of its spectrum below 1 GHz., Comment: 8 pages, 2 tables, 1 figure

The Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017 has defined ambitious new benchmarks to advance the state-of-the-art in autonomous operation of ground-based and flying robots. This article covers our approaches to solve the two challenges that involved micro aerial vehicles (MAV). Challenge 1 required reliable target perception, fast trajectory planning, and stable control of an MAV in order to land on a moving vehicle. Challenge 3 demanded a team of MAVs to perform a search and transportation task, coined "Treasure Hunt", which required mission planning and multi-robot coordination as well as adaptive control to account for the additional object weight. We describe our base MAV setup and the challenge-specific extensions, cover the camera-based perception, explain control and trajectory-planning in detail, and elaborate on mission planning and team coordination. We evaluated our systems in simulation as well as with real-robot experiments during the competition in Abu Dhabi. With our system, we-as part of the larger team NimbRo-won the MBZIRC Grand Challenge and achieved a third place in both subchallenges involving flying robots.

Robust and fast motion estimation and mapping is a key prerequisite for autonomous operation of mobile robots. The goal of performing this task solely on a stereo pair of video cameras is highly demanding and bears conflicting objectives: on one hand, the motion has to be tracked fast and reliably, on the other hand, high-level functions like navigation and obstacle avoidance depend crucially on a complete and accurate environment representation. In this work, we propose a two-layer approach for visual odometry and SLAM with stereo cameras that runs in real-time and combines feature-based matching with semi-dense direct image alignment. Our method initializes semi-dense depth estimation, which is computationally expensive, from motion that is tracked by a fast but robust keypoint-based method. Experiments on public benchmark and proprietary datasets show that our approach is faster than state-of-the-art methods without losing accuracy and yields comparable map building capabilities. Moreover, our approach is shown to handle large inter-frame motion and illumination changes much more robustly than its direct counterparts.

Most window management systems support multitasking by allowing users to open, resize, position, and switch between application windows. Although multitasking has become a way of life for most knowledge workers, our current understanding of how users use window management features to switch between multiple tasks---which may comprise multiple application windows---is limited. In this paper, we present a study providing an in-depth analysis of how task switching is supported in Windows 7. As part of analysis, we developed an interface-agnostic classification of common task switching operations supported by window managers which can be used to quantify the time spent on each constituting action. Our study shows that task switching is a time intensive activity and highlights the dominant actions that contribute to task switch time. Furthermore, our classification highlights the specific operations that are optimized by more recent and experimental window managers and allows identifying opportunities for design that could further reduce the overhead of switching between tasks., Comment: Includes an ancillary video figure detailing a task switch during the experiment: 'Example_task_switch.mp4'

The past years have shown a remarkable growth in use-cases for micro aerial vehicles (MAVs). Conceivable indoor applications require highly robust environment perception, fast reaction to changing situations, and stable navigation, but reliable sources of absolute positioning like GNSS or compass measurements are unavailable during indoor flights. We present a high-performance autonomous inventory MAV for operation inside warehouses. The MAV navigates along warehouse aisles and detects the placed stock in the shelves alongside its path with a multimodal sensor setup containing an RFID reader and two high-resolution cameras. We describe in detail the SLAM pipeline based on a 3D lidar, the setup for stock recognition, the mission planning and trajectory generation, as well as a low-level routine for avoidance of dynamical or previously unobserved obstacles. Experiments were performed in an operative warehouse of a logistics provider, in which an external warehouse management system provided the MAV with high-level inspection missions that are executed fully autonomously., Comment: IEEE Robotics and Automation Letters (RA-L), 2018. Accepted for IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, Spain, October 2018. For supplementary material and videos, see http://ais.uni-bonn.de/videos/IROS_2018_InventAIRy/

Fast radio bursts are a new class of transient radio phenomena currently detected as millisecond radio pulses with very high dispersion measures. As new radio surveys begin searching for FRBs a large population is expected to be detected in real-time, triggering a range of multi-wavelength and multi-messenger telescopes to search for repeating bursts and/or associated emission. Here we propose a method for disseminating FRB triggers using Virtual Observatory Events (VOEvents). This format was developed and is used successfully for transient alerts across the electromagnetic spectrum and for multi-messenger signals such as gravitational waves. In this paper we outline a proposed VOEvent standard for FRBs that includes the essential parameters of the event and where these parameters should be specified within the structure of the event. An additional advantage to the use of VOEvents for FRBs is that the events can automatically be ingested into the FRB Catalogue (FRBCAT) enabling real-time updates for public use. We welcome feedback from the community on the proposed standard outlined below and encourage those interested to join the nascent working group forming around this topic., Comment: 11 pages, 2 figures, parameter definition table in appendix

We investigate necessary as well as sufficient conditions under which the Laurent series coefficients $f_{\boldsymbol{n}}$ associated to a multivariate rational function satisfy Gauss congruences, that is $f_{\boldsymbol{m}p^r} \equiv f_{\boldsymbol{m}p^{r-1}}$ modulo $p^r$. For instance, we show that these congruences hold for certain determinants of logarithmic derivatives. As an application, we completely classify rational functions $P/Q$ satisfying the Gauss congruences in the case that $Q$ is linear in each variable., Comment: 20 pages

Endohedral lanthanide ions packed inside carbon nanotubes (CNTs) in a one-dimensional assembly have been studied with a combination of high resolution transmission electron microscopy (HRTEM), scanning transmission X-ray microscopy (STXM), and X-ray magnetic circular dichroism (XMCD). By correlating HRTEM and STXM images we show that structures down to 30 nm are resolved with chemical contrast and record X-ray absorption spectra from endohedral lanthanide ions embedded in individual nanoscale CNT bundles. XMCD measurements of an Er$_3$N@C$_{80}$ bulk sample and a macroscopic assembly of filled CNTs indicates that the magnetic properties of the endohedral Er3+ ions are unchanged when encapsulated in CNTs. This study demonstrates the feasibility of local magnetic X-ray characterization of low concentrations of lanthanide ions embedded in molecular nanostructures.

Active control over the shape, composition and crystalline habit of nanocrystals is a long sought-after goal. Various methods have been shown to enable post-synthesis modification of nanoparticles, including the use of the Kirkendall effect, galvanic replacement and cation or anion exchange, all taking advantage of enhanced solid-state diffusion on the nanoscale. In all these processes, however, alteration of the nanoparticles requires introduction of new precursor materials. Here we show that for cesium lead halide perovskite nanoparticles, a reversible structural and compositional change can be induced at room temperature solely by modification of the ligand composition in solution. Reversible transformation of cubic CsPbX3 nanocrystals to rhombohedral Cs4PbX6 nanocrystals is achieved by controlling the ratio of oleylamine to oleic acid capping molecules. This scheme does not only enable fabrication of high purity monodisperse Cs4PbX6 nanoparticles with controlled sizes. Rather, dependent on the Cs4PbX6 nanoparticles final size afforded by reaction time, the back reaction yields CsPbX3 nanoplatelets with controlled thickness. These results shed some light on the unique properties of this family of materials, provide hints to the origin of their superb photovoltaic performance and point at possible routes for their controlled fabrication and stabilization.

Excitons and plasmons are the two most fundamental types of collective electronic excitations occurring in solids. Traditionally, they have been studied separately using bulk techniques that probe their average energetic structure over large spatial regions. However, as the dimensions of materials and devices continue to shrink, it becomes crucial to understand how these excitations depend on local variations in the crystal- and chemical structure on the atomic scale. Here we use monochromated low-loss scanning-transmission-electron-microscopy electron-energy-loss (LL-STEM-EEL) spectroscopy, providing the best simultaneous energy and spatial resolution achieved to-date to unravel the full set of electronic excitations in few-layer MoS2 nanosheets over a wide energy range. Using first-principles many-body calculations we confirm the excitonic nature of the peaks at ~2eV and ~3eV in the experimental EEL spectrum and the plasmonic nature of higher energy-loss peaks. We also rationalise the non-trivial dependence of the EEL spectrum on beam and sample geometry such as the number of atomic layers and distance to steps and edges. Moreover, we show that the excitonic features are dominated by the long wavelength (q=0) components of the probing field, while the plasmonic features are sensitive to a much broader range of q-vectors, indicating a qualitative difference in the spatial character of the two types of collective excitations. Our work provides a template protocol for mapping the local nature of electronic excitations that open new possibilities for studying photo-absorption and energy transfer processes on a nanometer scale., Comment: Main paper: 19 pages, 5 figures Accepted for publication, Manuscript number: NPJ2DMATERIALS-00006, ISSN 2397-7132

Background: The increasing volume and variety of genotypic and phenotypic data is a major defining characteristic of modern biomedical sciences. At the same time, the limitations in technology for generating data and the inherently stochastic nature of biomolecular events have led to the discrepancy between the volume of data and the amount of knowledge gleaned from it. A major bottleneck in our ability to understand the molecular underpinnings of life is the assignment of function to biological macromolecules, especially proteins. While molecular experiments provide the most reliable annotation of proteins, their relatively low throughput and restricted purview have led to an increasing role for computational function prediction. However, accurately assessing methods for protein function prediction and tracking progress in the field remain challenging. Methodology: We have conducted the second Critical Assessment of Functional Annotation (CAFA), a timed challenge to assess computational methods that automatically assign protein function. One hundred twenty-six methods from 56 research groups were evaluated for their ability to predict biological functions using the Gene Ontology and gene-disease associations using the Human Phenotype Ontology on a set of 3,681 proteins from 18 species. CAFA2 featured significantly expanded analysis compared with CAFA1, with regards to data set size, variety, and assessment metrics. To review progress in the field, the analysis also compared the best methods participating in CAFA1 to those of CAFA2. Conclusions: The top performing methods in CAFA2 outperformed the best methods from CAFA1, demonstrating that computational function prediction is improving. This increased accuracy can be attributed to the combined effect of the growing number of experimental annotations and improved methods for function prediction., Comment: Submitted to Genome Biology

From a liberal perspective, pluralism and viewpoint diversity are seen as a necessary condition for a well-functioning democracy. Recently, there have been claims that viewpoint diversity is diminishing in online social networks, putting users in a "bubble", where they receive political information which they agree with. The contributions from our investigations are fivefold: (1) we introduce different dimensions of the highly complex value viewpoint diversity using political theory; (2) we provide an overview of the metrics used in the literature of viewpoint diversity analysis; (3) we operationalize new metrics using the theory and provide a framework to analyze viewpoint diversity in Twitter for different political cultures; (4) we share our results for a case study on minorities we performed for Turkish and Dutch Twitter users; (5) we show that minority users cannot reach a large percentage of Turkish Twitter users. With the last of these contributions, using theory from communication scholars and philosophers, we show how minority access is missing from the typical dimensions of viewpoint diversity studied by computer scientists and the impact it has on viewpoint diversity analysis., Comment: Computers in Human Behavior, 2014

In this work we present an in-depth analysis of the user behaviors on different Social Sharing systems. We consider three popular platforms, Flickr, Delicious and StumbleUpon, and, by combining techniques from social network analysis with techniques from semantic analysis, we characterize the tagging behavior as well as the tendency to create friendship relationships of the users of these platforms. The aim of our investigation is to see if (and how) the features and goals of a given Social Sharing system reflect on the behavior of its users and, moreover, if there exists a correlation between the social and tagging behavior of the users. We report our findings in terms of the characteristics of user profiles according to three different dimensions: (i) intensity of user activities, (ii) tag-based characteristics of user profiles, and (iii) semantic characteristics of user profiles.

Layered molecular materials and especially MoS2 are already accepted as promising candidates for nanoelectronics. In contrast to the bulk material, the observed electron mobility in single-layer MoS2 is unexpectedly low. Here we reveal the occurrence of intrinsic defects in MoS2 layers, known as inversion domains, where the layer changes its direction through a line defect. The line defects are observed experimentally by atomic resolution TEM. The structures were modeled and the stability and electronic properties of the defects were calculated using quantum-mechanical calculations based on the Density-Functional Tight-Binding method. The results of these calculations indicate the occurrence of new states within the band gap of the semiconducting MoS2. The most stable non-stoichiometric defect structures are observed experimentally, one of which contains metallic Mo-Mo bonds and another one bridging S atoms.

Electron tomography is becoming an increasingly important tool in materials science for studying the three-dimensional morphologies and chemical compositions of nanostructures. The image quality obtained by many current algorithms is seriously affected by the problems of missing wedge artefacts and nonlinear projection intensities due to diffraction effects. The former refers to the fact that data cannot be acquired over the full $180^\circ$ tilt range; the latter implies that for some orientations, crystalline structures can show strong contrast changes. To overcome these problems we introduce and discuss several algorithms from the mathematical fields of geometric and discrete tomography. The algorithms incorporate geometric prior knowledge (mainly convexity and homogeneity), which also in principle considerably reduces the number of tilt angles required. Results are discussed for the reconstruction of an InAs nanowire.

The phenomenon of a partial 2H\rightarrow1T phase transition within multiwalled WS2 nanotubes under substitutional Rhenium doping is discovered by means of high-resolution transmission electron microscopy. Using density-functional calculations for the related MoS2 compound we consider a possible origin of this phase transition, which was known formerly only for WS2 and MoS2 intercalated by alkali metals. An interplay between the stability of layered or nanotubular forms of 2H and 1T allotropes is found to be intimately related with their electronic structures and electro-donating ability of an impurity., Comment: submitted to JPC on 09th August 2011, still under revision!

We present a comprehensive characterization of an individual multiwalled carbon nanotube transport device combining electron microscopy and Raman spectroscopy with electrical measurements. Each method gives complementary information that mutually help to interpret each other. A sample design that allows for combining these investigation methods on individual carbon nanotube devices is introduced. This offers a direct correlation of transport features and shifts of Raman modes with structural properties as e.g. the contact interface and the morphology of the nanotube.

With the growing popularity of Social Web applications, more and more user data is published on the Web everyday. Our research focuses on investigating ways of mining data from such platforms that can be used for modeling users and for semantically augmenting user profiles. This process can enhance adaptation and personalization in various adaptive Web-based systems. In this paper, we present the U-Sem people modeling service, a framework for the semantic enrichment and mining of people's profiles from usage data on the Social Web. We explain the architecture of our people modeling service and describe its application in an adult e-learning context as an example. Versions: Mar 21, 10:10, Mar 25, 09:37, Comment: 1st International Workshop on Usage Analysis and the Web of Data (USEWOD2011) in the 20th International World Wide Web Conference (WWW2011), Hyderabad, India, March 28th, 2011