Search

Mediation analysis for survival outcomes is challenging. Most existing methods quantify the treatment effect using the hazard ratio (HR) and attempt to decompose the HR into the direct effect of treatment plus an indirect, or mediated, effect. However, the HR is not expressible as an expectation, which complicates this decomposition, both in terms of estimation and interpretation. Here, we present an alternative approach which leverages pseudo-values to simplify estimation and inference. Pseudo-values take censoring into account during their construction, and once derived, can be modeled in the same way as any continuous outcome. Thus, pseudo-values enable mediation analysis for a survival outcome to fit seamlessly into standard mediation software (e.g. CMAverse in R). Pseudo-values are easy to calculate via a leave-one-observation-out procedure (i.e. jackknifing) and the calculation can be accelerated when the influence function of the estimator is known. Mediation analysis for causal effects defined by survival probabilities, restricted mean survival time, and cumulative incidence functions - in the presence of competing risks - can all be performed within this framework. Extensive simulation studies demonstrate that the method is unbiased across 324 scenarios/estimands and controls the type-I error at the nominal level under the null of no mediation. We illustrate the approach using data from the PARADIGMS clinical trial for the treatment of pediatric multiple sclerosis using fingolimod. In particular, we evaluate whether an imaging biomarker lies on the causal path between treatment and time-to-relapse, which aids in justifying this biomarker as a surrogate outcome. Our approach greatly simplifies mediation analysis for survival data and provides a decomposition of the total effect that is both intuitive and interpretable., Comment: Mediation, Pseudo-Values, Time-to-event, Survival Analysis, Restricted Mean Survival Time, Competing Risks

We propose a novel strategy to fit experimental data using a UV complete amplitude ansatz satisfying the constraints of Analyticity, Crossing, and Unitarity. We focus on $\pi\pi$ scattering combining both experimental and lattice data. The fit strategy requires using S-matrix Bootstrap methods and non-convex Particle Swarm Optimization techniques. Using this procedure, we numerically construct a full-fledged scattering amplitude that fits the data and contains the known QCD spectrum that couples to $\pi \pi$ states below $1.4$ GeV. The amplitude constructed agrees below the two-particle threshold with the two-loop $\chi$PT prediction. Moreover, we correctly predict the $D_2$ phase shift, the appearance of a spin three state, and the behavior of the high-energy total cross-section. Finally, we find a genuine tetraquark resonance around 2 GeV, which we argue might be detected by looking into the decays of B mesons., Comment: 11 pages + appendices, 14 figures, fixed some typos, improved the text, added more references

We explore the physical mechanisms responsible for generating the graviton pole in twice-subtracted dispersion relations, both in flat space and in AdS. To characterize these mechanisms, we analyze the energy scale at which the graviton pole is generated in scattering experiments at various impact parameters. At large impact parameters, we identify the eikonal model of high-energy gravitational scattering as a universal mechanism that generates the graviton pole in dispersion relations. At smaller impact parameters, the graviton pole can arise from stringy higher-spin resonances. The length scale at which the graviton pole generation scale departs from its semiclassical eikonal value indicates the breakdown of gravitational EFT. In flat space, we derive a Tauberian theorem for the graviton pole, which must be satisfied by any UV completion of gravity that admits twice-subtracted dispersion relations. In AdS, free or non-holographic CFTs offer an alternative mechanism to generate the graviton pole. More broadly, we find that the existing picture of high-energy gravitational scattering, including phenomena such as black hole formation and various stringy effects, is compatible with the twice-subtracted dispersion relations., Comment: 36 pages + appendices, 6 figures

We explore the space of meromorphic amplitudes with extra constraints coming from the shape of the leading Regge trajectory. This information comes in two guises: it bounds the maximal spin of exchanged particles of a given mass; it leads to sum rules obeyed by the discontinuity of the amplitude, which express the softness of scattering at high energies. We assume that the leading Regge trajectory is linear, and we derive bounds on the low-energy Wilson coefficients using the dual and primal approaches. For the graviton-graviton scattering in four dimensions, the maximal spin constraint leads to slightly more stringent bounds than those that follow from general constraints of analyticity, crossing, and unitarity. The exponential softness at high energies is manifest in our primal approach and is not used in our implementation of the dual approach. Nevertheless, we observe the agreement between the bounds obtained from both. We conclude that high-energy superpolynomial softness does not leave an obvious imprint on the low-energy observables. We exhibit a unitary three-parameter deformation of the Veneziano amplitude for the open string case. It has a novel, exponentially soft behavior at high energies and fixed angles. We generalize the previous analysis of this regime and present a stringy version of the lower bound on high-energy, fixed-angle scattering by Cerulus and Martin., Comment: 33 pages + appendices, 18 figures

In this work, we develop a method based on robust control techniques to synthesize robust time-varying state-feedback policies for finite, infinite, and receding horizon control problems subject to convex quadratic state and input constraints. To ensure constraint satisfaction of our policy, we employ (initial state)-to-peak gain techniques. Based on this idea, we formulate linear matrix inequality conditions, which are simultaneously convex in the parameters of an affine control policy, a Lyapunov function along the trajectory and multiplier variables for the uncertainties in a time-varying linear fractional transformation model. In our experiments this approach is less conservative than standard tube-based robust model predictive control methods., Comment: Extended version of a contribution to be submitted to the European Control Conference 2024

We study the space of $2\to 2$ scattering amplitudes of neutral Goldstone bosons in four space-time dimensions. We establish universal bounds on the first two non-universal Wilson coefficients of the low energy Effective Field Theory (EFT) for such particles. We reconstruct the analytic, crossing-symmetric, and unitary amplitudes saturating our bounds, and we study their physical content. We uncover non-perturbative Regge trajectories by continuing our numerical amplitudes to complex spins. We then explore the consequence of additional constraints arising when we impose the knowledge about the EFT up to the cut-off scale. In the process, we improve on some aspects of the numerical $S$-matrix bootstrap technology for massless particles., Comment: 28 pages + appendices, 16 figures

Finite time convergence to functionally important target states is a key component of many biological processes. We previously found that the terminal approach phase of such dynamics exhibits universal types of stochastic dynamics that differ qualitatively between noise-dominated and force-dominated regimes of the approach dynamics. While for the noise-dominated regime the approach dynamics is uninformative about the underlying force law, in the force-dominated regime it enables the accurate inference of the underlying dynamics. Biological systems often exhibit substantial parameter heterogeneity, for instance through copy number fluctuations of key molecules or variability in modulating factors. Here, we extend our theory of target state aligned (TSA) stochastic dynamics to investigate the impact of parameter heterogeneity in the underlying stochastic dynamics. We examine the approach to target states for a wide range of dynamical laws and additive as well as multiplicative noise. We find that the distinct regimes of noise-dominated and force-dominated dynamics strongly differ in their sensitivity to parameter heterogeneity. In the noise-dominated regime, TSA ensembles are insensitive to parameter heterogeneity in the force law, but sensitive to sample to sample heterogeneity in the diffusion constant. For force-dominated dynamics, both parameter heterogeneity in the force law and diffusion constant change the behaviour of the non-stationary statistics and in particular the two-time-covariance functions. In this regime, TSA ensembles provide a sensitive readout of parameter heterogeneity. Under natural conditions, parameter heterogeneity in many biological systems cannot be experimentally controlled or eliminated. Our results provide a systematic theoretical foundation for the analysis of target state directed dynamics in a large class of systems with substantial heterogeneity.

The dynamics of living systems often serves the purpose of reaching functionally important target states. We previously proposed a theory to analyze stochastic biological dynamics evolving towards target states in reverse time. However, a large class of systems in biology can only be adequately described using state-dependent noise, which had not been discussed. For example, in gene regulatory networks, biochemical signaling networks or neuronal circuits, count fluctuations are the dominant noise component. We characterize such dynamics as an ensemble of target state aligned (TSA) trajectories and characterize its temporal evolution in reverse-time by generalized Fokker-Planck and stochastic differential equations with multiplicative noise. We establish the classification of boundary conditions for target state modeling for a wide range of power law dynamics, and derive a universal low-noise approximation of the final phase of target state convergence. Our work expands the range of theoretically tractable systems in biology and enables novel experimental design strategies for systems that involve target states.

We study 2-to-2 scattering amplitudes of massless spin one particles in $d=4$ space-time dimensions, like real world photons. We define a set of non-perturbative observables (Wilson coefficients) which describe these amplitudes at low energies. We use full non-linear unitarity to construct various novel numerical bounds on these observables. For completeness, we also rederive some bounds using positivity only. We discover and explain why some of these Wilson coefficients cannot be bounded., Comment: 39 pages + appendices, 20 figures

We review the basic assumptions and spell out the detailed arguments that lead to the bound on the Regge growth of gravitational scattering amplitudes. The minimal extra ingredient compared to the gapped case - in addition to unitarity, analyticity, subexponentiality, and crossing - is the assumption that scattering at large impact parameters is controlled by known semi-classical physics. We bound the Regge growth of amplitudes both with the fixed transferred momentum and smeared over it. Our basic conclusion is that gravitational scattering amplitudes admit dispersion relations with two subtractions. For a sub-class of smeared amplitudes, black hole formation reduces the number of subtractions to one. Finally, using dispersion relations with two subtractions we derive bounds on the local growth of relativistic scattering amplitudes. Schematically, the local bound states that the amplitude cannot grow faster than $s^2$., Comment: 17 pages + appendices, 9 figures ; v2: minor typos and corrections

In complex crystals close to melting or at finite temperatures, different types of defects are ubiquitous and their role becomes relevant in the mechanical response of these solids. Conventional elasticity theory fails to provide a microscopic basis to include and account for the motion of point-defects in an otherwise ordered crystalline structure. We study the elastic properties of a point-defect rich crystal within a first-principles theoretical framework derived from microscopic equations of motion. This framework allows us to make specific predictions pertaining to the mechanical properties which we can validate through deformation experiments performed in Molecular Dynamics simulations.

App stores allow users to give valuable feedback on apps, and developers to find this feedback and use it for the software evolution. However, finding user feedback that matches existing bug reports in issue trackers is challenging as users and developers often use a different language. In this work, we introduce DeepMatcher, an automatic approach using state-of-the-art deep learning methods to match problem reports in app reviews to bug reports in issue trackers. We evaluated DeepMatcher with four open-source apps quantitatively and qualitatively. On average, DeepMatcher achieved a hit ratio of 0.71 and a Mean Average Precision of 0.55. For 91 problem reports, DeepMatcher did not find any matching bug report. When manually analyzing these 91 problem reports and the issue trackers of the studied apps, we found that in 47 cases, users actually described a problem before developers discovered and documented it in the issue tracker. We discuss our findings and different use cases for DeepMatcher., Comment: Accepted for publication to the 43rd International Conference on Software Engineering (ICSE21)

One-shot action recognition allows the recognition of human-performed actions with only a single training example. This can influence human-robot-interaction positively by enabling the robot to react to previously unseen behaviour. We formulate the one-shot action recognition problem as a deep metric learning problem and propose a novel image-based skeleton representation that performs well in a metric learning setting. Therefore, we train a model that projects the image representations into an embedding space. In embedding space the similar actions have a low euclidean distance while dissimilar actions have a higher distance. The one-shot action recognition problem becomes a nearest-neighbor search in a set of activity reference samples. We evaluate the performance of our proposed representation against a variety of other skeleton-based image representations. In addition, we present an ablation study that shows the influence of different embedding vector sizes, losses and augmentation. Our approach lifts the state-of-the-art by 3.3% for the one-shot action recognition protocol on the NTU RGB+D 120 dataset under a comparable training setup. With additional augmentation our result improved over 7.7%., Comment: 8 pages, 8 figures, 4 tables

Purpose: To enable fast and reliable assessment of subcutaneous and visceral adipose tissue compartments derived from whole-body MRI. Methods: Quantification and localization of different adipose tissue compartments from whole-body MR images is of high interest to examine metabolic conditions. For correct identification and phenotyping of individuals at increased risk for metabolic diseases, a reliable automatic segmentation of adipose tissue into subcutaneous and visceral adipose tissue is required. In this work we propose a 3D convolutional neural network (DCNet) to provide a robust and objective segmentation. In this retrospective study, we collected 1000 cases (66$\pm$ 13 years; 523 women) from the Tuebingen Family Study and from the German Center for Diabetes research (TUEF/DZD), as well as 300 cases (53$\pm$ 11 years; 152 women) from the German National Cohort (NAKO) database for model training, validation, and testing with a transfer learning between the cohorts. These datasets had variable imaging sequences, imaging contrasts, receiver coil arrangements, scanners and imaging field strengths. The proposed DCNet was compared against a comparable 3D UNet segmentation in terms of sensitivity, specificity, precision, accuracy, and Dice overlap. Results: Fast (5-7seconds) and reliable adipose tissue segmentation can be obtained with high Dice overlap (0.94), sensitivity (96.6%), specificity (95.1%), precision (92.1%) and accuracy (98.4%) from 3D whole-body MR datasets (field of view coverage 450x450x2000mm${}^3$). Segmentation masks and adipose tissue profiles are automatically reported back to the referring physician. Conclusion: Automatic adipose tissue segmentation is feasible in 3D whole-body MR data sets and is generalizable to different epidemiological cohort studies with the proposed DCNet., Comment: This manuscript has been accepted for publication in Radiology: Artificial Intelligence (https://pubs.rsna.org/journal/ai), which is published by the Radiological Society of North America (RSNA)

Experts agree that keeping systems up to date is a powerful security measure. Previous work found that users sometimes explicitly refrain from performing timely updates, e.g., due to bad experiences which has a negative impact on end-user security. Another important user group has been investigated less extensively: system administrators, who are responsible for keeping complex and heterogeneous system landscapes available and secure. In this paper, we sought to understand administrators' behavior, experiences, and attitudes regarding updates in a corporate environment. Based on the results of an interview study, we developed an online survey and quantified common practices and obstacles (e.g., downtime or lack of information about updates). The findings indicate that even experienced administrators struggle with update processes as the consequences of an update are sometimes hard to assess. Therefore, we argue that more usable monitoring and update processes are essential to guarantee IT security at scale.

Reliably predictive simulation of complex flows requires a level of model sophistication and robustness exceeding the capabilities of current Reynolds-averaged Navier-Stokes (RANS) models. The necessary capability can often be provided by well-resolved large eddy simulation (LES), but, for many flows of interest, such simulations are too computationally intensive to be performed routinely. In principle, hybrid RANS/LES (HRL) models capable of transitioning through arbitrary levels of modeled and resolved turbulence would ameliorate both RANS deficiencies and LES expense. However, these HRL approaches have led to a host of unique complications, in addition to those already present in RANS and LES. This work proposes a modeling approach aimed at overcoming such challenges. The approach presented here relies on splitting the turbulence model into three distinct components: two responsible for the standard subgrid model roles of either providing the unresolved stress or dissipation and a third which reduces the model length scale by creating resolved turbulence. This formulation renders blending functions unnecessary in HRL. Further, the split-model approach both reduces the physics-approximation burden on simple eddy-viscosity-based models and provides convenient flexibility in model selection. In regions where the resolution is adequate to support additional turbulence, fluctuations are generated at the smallest locally resolved scales of motion. This active forcing drives the system towards a balance between RANS and grid-resolved LES for any combination of resolution and flow while the split-model formulation prevents local disruption to the total stress. The model is demonstrated on fully-developed, incompressible channel flow [1] and the periodic hill [2], in which it is shown to produce accurate results and avoid common HRL shortcomings, such as model stress depletion., Comment: 16 Figures

Like natural seismicity, induced seismicity caused by fluid injection also shows a power law size distribution, and its gradient b-value (ratio of small to large earthquakes) is often used for seismic hazard analysis. Despite well-known relationship that b-value is negatively correlated with differential stress for natural earthquakes, there is no understanding of the physical causes for b-value variations in injection-induced seismicity in the scale where the differential is nearly constant. We investigate a b-value dependence on the relative shear stress acting on existing fractures and show that the seismicity occurring along existing fractures with high shear stress have significantly lower b-values than does that associated with lower shear stress fractures. The b-value for injection induced seismicity is dependent on relative shear stress on faults. Our results provide a novel physical explanation for the b-value variations of induced seismicity., Comment: 15 pages, 8 figures, Supplementary information

Large Eddy Simulation (LES) of turbulence in complex geometries is often conducted using strongly inhomogeneous resolution. The issues associated with resolution inhomogeneity are related to the noncommutativity of the filtering and differentiation operators, which introduces a commutation term into the governing equations. Neglect of this commutation term gives rise to commutation error. While the commutation error is well recognized, it is often ignored in practice. Moreover, the commutation error arising from the implicit part of the filter (i.e., projection onto the underlying discretization) has not been well investigated. Modeling the commutator between numerical projection and differentiation is crucial for correcting errors induced by resolution inhomogeneity in practical LES settings, which typically rely solely on implicit filtering. Here, we employ a multiscale asymptotic analysis to investigate the characteristics of the commutator. This provides a statistical description of the commutator, which can serve as a target for the statistical characteristics of a commutator model. Further, we investigate how commutation error manifests in simulation and demonstrate its impact on the convection of a packet of homogeneous isotropic turbulence through an inhomogeneous grid. A connection is made between the commutation error and the propagation properties of the underlying numerics. A modeling approach for the commutator is proposed that is applicable to LES with filters that include projections to the discrete solution space and that respects the numerical properties of the LES evolution equation. It may also be useful in addressing other LES modeling issues such as discretization error., Comment: 27 pages, 9 figures

Smart speakers are gaining popularity. However, such devices can put the user's privacy at risk whenever hot-words are misinterpreted and voice data is recorded without the user's consent. To mitigate such risks, smart speakers provide privacy control mechanisms like the build-in mute button. Unfortunately, previous work indicated that such mute buttons are rarely used. In this paper, we present the Privacy Hat, a tangible device which can be placed on the smart speaker to prevent the device from listening. We designed the Privacy Hat based on the results of a focus group and developed a working prototype. We hypothesize that the specific user experience of this physical and tangible token makes the use of privacy-enhancing technology more graspable for the user. As a consequence, we expect that the Privacy Hat nudges users to more actively use privacy-enhancing features like the mute button. In addition, we propose the Privacy Hat as a study tool as we hypothesize that the artifact supports participants in reflecting their behaviour. We report on the concept, the prototype and our preliminary results., Comment: SOUPS 2019 Poster Session

Nowadays, most mobile devices support biometric authentication schemes like fingerprint or face unlock. However, these probabilistic mechanisms can only be activated in combination with a second alternative factor, usually knowledge-based authentication. In this paper, we show that this aspect can be exploited in a bypass attack. In this bypass attack, the attacker forces the user to "bypass" the biometric authentication by, for example, resetting the phone. This forces the user to enter an easy-to-observe passcode instead. We present the threat model and provide preliminary results of an online survey. Based on our results, we discuss potential countermeasures. We conclude that better feedback design and security-optimized fallback mechanisms can help further improve the overall security of mobile unlock mechanisms while preserving usability., Comment: Poster Session SOUPS 2019

With the rise of social media like Twitter and of software distribution platforms like app stores, users got various ways to express their opinion about software products. Popular software vendors get user feedback thousandfold per day. Research has shown that such feedback contains valuable information for software development teams such as problem reports or feature and support inquires. Since the manual analysis of user feedback is cumbersome and hard to manage many researchers and tool vendors suggested to use automated analyses based on traditional supervised machine learning approaches. In this work, we compare the results of traditional machine learning and deep learning in classifying user feedback in English and Italian into problem reports, inquiries, and irrelevant. Our results show that using traditional machine learning, we can still achieve comparable results to deep learning, although we collected thousands of labels.

Large eddy simulation (LES) of turbulence in complex geometries and domains is often conducted with high aspect ratio resolution cells of varying shapes and orientations. The effects of such anisotropic resolution are often simplified or neglected in subgrid model formulation. Here, we examine resolution induced anisotropy and demonstrate that, even for isotropic turbulence, anisotropic resolution induces mild resolved Reynolds stress anisotropy and significant anisotropy in second-order resolved velocity gradient statistics. In large eddy simulations of homogeneous isotropic turbulence with anisotropic resolution, it is shown that commonly used subgrid models, including those that consider resolution anisotropy in their formulation, perform poorly. The one exception is the anisotropic minimum dissipation model proposed by Rozema et al. (Phys. of Fluids 27, 085107, 2015). A simple new model is presented here that is formulated with an anisotropic eddy diffusivity that depends explicitly on the anisotropy of the resolution. It also performs well, and is remarkable because unlike other LES subgrid models, the eddy diffusivity only depends on statistical characteristics of the turbulence (in this case the dissipation rate), not on fluctuating quantities. In other subgrid modeling formulations, such as the dynamic procedure, limiting flow dependence to statistical quantities in this way could have advantages., Comment: 19 pages, 11 figures

Predictive simulation of many complex flows requires moving beyond Reynolds-averaged Navier-Stokes (RANS) based models to representations resolving at least some scales of turbulence in at least some regions of the flow. To resolve turbulence where necessary while avoiding the cost of performing large eddy simulation (LES) everywhere, a broad range of hybrid RANS/LES methods have been developed. While successful in some situations, existing methods exhibit a number of deficiencies which limit their predictive capability in many cases of interest, for instance in many flows involving smooth wall separation. These deficiencies result from inappropriate blending approaches and resulting inconsistencies between the resolved and modeled turbulence as well as errors inherited from the underlying RANS and LES models. This work details these problems and their effects in hybrid simulations, and develops a modeling paradigm aimed at overcoming these challenges. The new approach discards typical blending approaches in favor of a hybridization strategy in which the RANS and LES model components act through separate models formulated using the mean and fluctuating velocity, respectively. Further, a forcing approach in which fluctuating content is actively transferred from the modeled to the resolved scales is introduced. Finally, the model makes use of an anisotropic LES model that is intended to represent the effects of grid anisotropy. The model is demonstrated on fully-developed, incompressible channel flow and shown to be very promising., Comment: 21 pages, 8 figures, AIAA SciTech 2019 conference paper

User comments have become an essential part of online journalism. However, newsrooms are often overwhelmed by the vast number of diverse comments, for which a manual analysis is barely feasible. Identifying meta-comments that address or mention newsrooms, individual journalists, or moderators and that may call for reactions is particularly critical. In this paper, we present an automated approach to identify and classify meta-comments. We compare comment classification based on manually extracted features with an end-to-end learning approach. We develop, optimize, and evaluate multiple classifiers on a comment dataset of the large German online newsroom SPIEGEL Online and the 'One Million Posts' corpus of DER STANDARD, an Austrian newspaper. Both optimized classification approaches achieved encouraging $F_{0.5}$ values between 76% and 91%. We report on the most significant classification features with the results of a qualitative analysis and discuss how our work contributes to making participation in online journalism more constructive., Comment: Accepted for publication to the 21st ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW18)

In this contribution, a direct comparison of the Offset-QAM-OFDM (OQAM-OFDM) and the Cyclic Prefix OFDM (CP-OFDM) scheme is given for an 802.11a based system. Therefore, the chosen algorithms and choices of design are described and evaluated as a whole system in terms of bit and frame error rate (BER/FER) performance as well as spectral efficiency and complexity in the presence of multipath propagation for different modulation orders. The results show that the OQAM-OFDM scheme exhibits similar BER and FER performance at a 24% higher spectral efficiency and achievable throughput at the cost of an up to five times increased computational complexity., Comment: in Proceedings of the 16th International OFDM-Workshop, 2011 Hamburg, Germany

The isothermal compressibility of a general crystal is analyzed within classical density functional theory. Our approach can be used for homogeneous and unstrained crystals containing an arbitrarily high density of local defects. We start by coarse-graining the microscopic particle density and then obtain the long wavelength limits of the correlation functions of elasticity theory and the thermodynamic derivatives. We explicitly show that the long wavelength limit of the microscopic density correlation function differs from the isothermal compressibility. It also cannot be obtained from the static structure factor measured in a scattering experiment. We apply our theory to crystals consisting of soft particles which can multiply occupy lattice sites ('cluster crystals'). The multiple occupancy results in a strong local disorder over an extended range of temperatures. We determine the cluster crystals' isothermal compressibility, the fluctuations of the lattice occupation numbers and their correlation functions, and the dispersion relations. We also discuss their low-temperature phase diagram., Comment: 20 pages, 6 figures. arXiv admin note: substantial text overlap with arXiv:1206.6183