abstract
stringlengths
42
2.09k
Numerical solutions of the Enskog-Vlasov (EV) equation are used to determine the velocity distribution function of atoms spontaneously evaporating into near-vacuum conditions. It is found that an accurate approximation is provided by a half-Maxwellian including a drift velocity combined with different characteristic temperatures for the velocity components normal and parallel to the liquid-vapor interface. The drift velocity and the temperature anisotropy reduce as the liquid bulk temperature decreases but persist for relatively low temperatures corresponding to a vapor behaviour which is only slightly non-ideal. Deviations from the undrifted isotropic half-Maxwellian are shown to be consequences of collisions in the liquid-vapor interface which preferentially backscatter atoms with lower normal-velocity component.
Brain tissue is a heterogeneous material, constituted by a soft matrix filled with cerebrospinal fluid. The interactions between, and the complexity of each of these components are responsible for the non-linear rate-dependent behaviour that characterizes what is one of the most complex tissue in nature. Here, we investigate the influence of the cutting rate on the fracture properties of brain, through wire cutting experiments. We also present a model for the rate-dependent behaviour of fracture propagation in soft materials, which comprises the effects of fluid interaction through a poro-hyperelastic formulation. The method is developed in the framework of finite strain continuum mechanics, implemented in a commercial finite element code, and applied to the case of an edge-crack remotely loaded by a controlled displacement. Experimental and numerical results both show a toughening effect with increasing rates, which is linked to the energy dissipated by the fluid-solid interactions in the process zone ahead of the crack.
In this paper, we design decentralized control strategies for the two-dimensional movement of autonomous vehicles on lane-free roads. The bicycle kinematic model is used to model the dynamics of the vehicles, and each vehicle determines its control input based only on its own speed and on the distance from other (adjacent) vehicles and the boundary of the road. Potential functions and Barbalat's lemma are employed to prove the following properties, which are ensured by the proposed controller: (i) the vehicles do not collide with each other or with the boundary of the road; (ii) the speeds of all vehicles are always positive, i.e., no vehicle moves backwards at any time; (iii) the speed of all vehicles remain below a given speed limit; (iv) all vehicle speeds converge to a given longitudinal speed set-point; and (v) the accelerations, lateral speeds, and orientations of all vehicles tend to zero. The efficiency of the proposed 2-D cruise controllers is illustrated by means of numerical examples.
A multi-modal framework to generated user intention distributions when operating a mobile vehicle is proposed in this work. The model learns from past observed trajectories and leverages traversability information derived from the visual surroundings to produce a set of future trajectories, suitable to be directly embedded into a perception-action shared control strategy on a mobile agent, or as a safety layer to supervise the prudent operation of the vehicle. We base our solution on a conditional Generative Adversarial Network with Long-Short Term Memory cells to capture trajectory distributions conditioned on past trajectories, further fused with traversability probabilities derived from visual segmentation with a Convolutional Neural Network. The proposed data-driven framework results in a significant reduction in error of the predicted trajectories (versus the ground truth) from comparable strategies in the literature (e.g. Social-GAN) that fail to account for information other than the agent's past history. Experiments were conducted on a dataset collected with a custom wheelchair model built onto the open-source urban driving simulator CARLA, proving also that the proposed framework can be used with a small, un-annotated dataset.
Several physical mechanisms are involved in excavating granular materials beneath a vertical jet of gas. These occur, for example, beneath the exhaust plume of a rocket landing on the soil of the Moon or Mars. A series of experiments and simulations have been performed to provide a detailed view of the complex gas/soil interactions. Measurements have also been taken from the Apollo lunar landing videos and from photographs of the resulting terrain, and these help to demonstrate how the interactions extrapolate into the lunar environment. It is important to understand these processes at a fundamental level to support the on-going design of higher-fidelity numerical simulations and larger-scale experiments. These are needed to enable future lunar exploration wherein multiple hardware assets will be placed on the Moon within short distances of one another. The high-velocity spray of soil from landing spacecraft must be accurately predicted and controlled lest it erosively damage the surrounding hardware.
Terahertz (THz) emission spectroscopy is a powerful method that allows one to measure the ultrafast dynamics of polarization, current, or magnetization in a material based on THz emission from the material. However, the practical implementation of this method can be challenging, and can result in significant errors in the reconstruction of the quantity of interest. Here, we experimentally and theoretically demonstrate a rigorous method of signal reconstruction in THz emission spectroscopy, and describe the main experimental and theoretical sources of reconstruction error. We identify the linear line-of-sight geometry of the THz emission spectrometer as the optimal configuration for accurate, fully calibrated THz signal reconstruction. As an example, we apply our reconstruction method to ultrafast THz magnetometry experiment, where we recover the ultrafast magnetization dynamics in a photoexcited iron film, including both its temporal shape and absolute magnitude.
We demonstrate an external cavity laser formed by combining a silicon nitride photonic integrated circuit with a reflective semiconductor optical amplifier. The laser uses an alignment tolerant edge coupler formed by a multi-mode waveguide splitter right at the edge of the silicon nitride chip that relaxes the required alignment to the III-V gain chip and equally splits the power among its two output waveguides. Both the ground and first order mode are excited in the coupler and reach the quadrature condition at the waveguide junction, ensuring equal power to be coupled to both. Two high-quality-factor ring resonators arranged in Vernier configuration close a Sagnac loop between the two waveguides. In addition to wideband frequency tuning, they result in a longer effective cavity length. The alignment tolerant coupler increases the alignment tolerance in the two directions parallel to the chip surface by a factor 3 relative to conventional edge couplers, making it ideal for gain chip integration via pick-and-place technology. Lasing is maintained in a misalignment range of $\pm$6 $\mu$m in the direction along the edge of the chip. A Lorentzian laser linewidth of 42 kHz is achieved.
Engineering Thermodynamics has been the core course of many science and engineering majors around the world, including energy and power, mechanical engineering, civil engineering, aerospace, cryogenic refrigeration, food engineering, chemical engineering, and environmental engineering, among which gas power cycle is one of the important contents. However, many Engineering Thermodynamics textbooks focus only on evaluating the thermal efficiency of gas power cycle, while the important concept of specific cycle work is ignored. Based on the generalized temperature-entropy diagram for the gas power cycles proposed by the authors, an ideal Otto cycle and an ideal Miller-Diesel cycle are taking as examples for the thermodynamic analyses of gas power cycles. The optimum compression ratio (or the pressure ratio) for the maximum specific cycle work or the maximum mean effective pressure is analyzed and determined. The ideal Otto and the ideal Miller-Diesel cycles, and also other gas power cycles for movable applications, are concluded that the operation under the maximum specific cycle work or the maximum mean effective pressure, instead of under the higher efficiency, is more economic and more reasonable. We concluded that the very important concept, i.e., the optimum compression (or pressure) ratio for the gas power cycles, should be emphasized in the Engineering Thermodynamics teaching process and in the latter revised or the newly edited textbooks, in order to better guide the engineering applications.
We investigate the potential of a recently proposed model for 3D compressible MHD turbulence (Chevillard et al. 2010; Durrive et al. 2021) to be used as a tool to characterize statistically 2D and 3D turbulent data. This model is parametrized by a dozen of free (intuitive, physically motivated) parameters, which control the statistics of the fields (density, velocity and magnetic fields). The present study is a proof of concept study: (i) we restrict ourselves to the incompressible hydrodynamical part of the model, (ii) we consider as data centroid velocity maps, and (iii) we let only three of the free parameters vary (namely the correlation length, the Hurst parameter and the intermittency parameter). Within this framework, we demonstrate that, given a centroid velocity map, we can find in an automated manner (i.e. by a Markov Chain Monte Carlo analysis) values of the parameters such that the model resembles the given map, i.e. which reproduces its statistics fairly well. Hence, thanks to this procedure, one may characterize statistically, and thus compare, various turbulent data. In other words, we show how this model may be used as a metric to compare observational or simulated data sets. In addition, because this model is numerically particularly fast (nearly 500 times faster than the numerical simulation we use to generate our reference data) it may be used as a surrogate model. Finally, by this process we also initiate the first systematic exploration of the parameter space of this model. Doing so, we show how the parameters impact the visual and the statistical properties of centroid velocity maps, and exhibit the correlations between the various parameters, providing new insight into the model.
An essential problem of swarm robotics is how members of the swarm knows the positions of other robots. The main aim of this research is to develop a cost-effective and simple vision-based system to detect the range, bearing, and heading of the robots inside a swarm using a multi-purpose passive landmark. A small Zumo robot equipped with Raspberry Pi, PiCamera is utilized for the implementation of the algorithm, and different kinds of multipurpose passive landmarks with nonsymmetrical patterns, which give reliable information about the range, bearing and heading in a single unit, are designed. By comparing the recorded features obtained from image analysis of the landmark through systematical experimentation and the actual measurements, correlations are obtained, and algorithms converting those features into range, bearing and heading are designed. The reliability and accuracy of algorithms are tested and errors are found within an acceptable range.
We investigate a well defined heterostructure constituted by magnetic Fe layers sandwiched between graphene (Gr) and Ir(111). The challenging task to avoid Fe-C solubility and Fe-Ir intermixing has been achieved with atomic controlled Fe intercalation at moderate temperature below 500 K. Upon intercalation of a single ordered Fe layer in registry with the Ir substrate, an intermixing of the Gr bands and Fe d states breaks the symmetry of the Dirac cone, with a downshift in energy of the apex by about 3 eV, and well-localized Fe intermixed states induced in the energy region just below the Fermi level. First principles electronic structure calculations show a large spin splitting of the Fe states, resulting in a majority spin channel almost fully occupied and strongly hybridized with Gr {\pi} states. X-ray magnetic circular dichroism on the Gr/Fe/Ir heterostructure reveals an ordered spin configuration with a ferromagnetic response of Fe layer(s), with enhanced spin and orbital configurations with respect to the bcc-Fe bulk values. The magnetization switches from a perpendicular easy magnetization axis when the Fe single layer is lattice matched with the Ir(111) surface to a parallel one when the Fe thin film is almost commensurate with graphene.
The dynamics of the coupled electron-nuclear spin system is studied in an ensemble of singly-charged (In,Ga)As/GaAs quantum dots (QDs) using periodic optical excitation at 1 GHz repetition rate. In combination with the electron-nuclei interaction, the highly repetitive excitation allows us to lock the electron spins into magnetic resonance in a transverse external magnetic field. Sweeping the field to higher values, the locking leads to an effective "diamagnetic" response of significant strength due to dynamic nuclear polarization, which shields the QD electrons at least partly from the external field and can even keep the internal magnetic field constant up to 1.3 T field variation. We model the effect through a magnetic field-dependent polarization rate of the nuclei, from which we suggest a strategy for adjusting the nuclear polarization through the detuning between optical excitation and electronic transition, in addition to tuning the magnetic field.
We develop a novel approach to non-relativistic closed bosonic string theory that is based on a string $1/c^2$ expansion of the relativistic string, where $c$ is the speed of light. This approach has the benefit that one does not need to take a limit of a string in a near-critical Kalb-Ramond background. The $1/c^2$-expanded Polyakov action at next-to-leading order reproduces the known action of non-relativistic string theory provided that the target space obeys an appropriate foliation constraint. We compute the spectrum in a flat target space, with one circle direction that is wound by the string, up to next-to-leading order and show that it reproduces the spectrum of the Gomis-Ooguri string.
In this note we continue giving the characterisation of weights for two-weight Hardy inequalities to hold on general metric measure spaces possessing polar decompositions. Since there may be no differentiable structure on such spaces, the inequalities are given in the integral form in the spirit of Hardy's original inequality. This is a continuation of our paper [M. Ruzhansky and D. Verma. Hardy inequalities on metric measure spaces, Proc. R. Soc. A., 475(2223):20180310, 2018] where we treated the case $p\leq q$. Here the remaining range $p>q$ is considered, namely, $0<q<p$, $1<p<\infty.$ We give examples obtaining new weighted Hardy inequalities on $\mathbb R^n$, on homogeneous groups, on hyperbolic spaces, and on Cartan-Hadamard manifolds. We note that doubling conditions are not required for our analysis.
In continual learning, a system learns from non-stationary data streams or batches without catastrophic forgetting. While this problem has been heavily studied in supervised image classification and reinforcement learning, continual learning in neural networks designed for abstract reasoning has not yet been studied. Here, we study continual learning of analogical reasoning. Analogical reasoning tests such as Raven's Progressive Matrices (RPMs) are commonly used to measure non-verbal abstract reasoning in humans, and recently offline neural networks for the RPM problem have been proposed. In this paper, we establish experimental baselines, protocols, and forward and backward transfer metrics to evaluate continual learners on RPMs. We employ experience replay to mitigate catastrophic forgetting. Prior work using replay for image classification tasks has found that selectively choosing the samples to replay offers little, if any, benefit over random selection. In contrast, we find that selective replay can significantly outperform random selection for the RPM task.
A set $D$ of vertices in an isolate-free graph $G$ is a semitotal dominating set of $G$ if $D$ is a dominating set of $G$ and every vertex in $D$ is within distance $2$ from another vertex of $D$.The semitotal domination number of $G$ is the minimum cardinality of a semitotal dominating set of $G$ and is denoted by $\gamma_{t2}(G)$. In this paper after computation of semitotal domination number of specific graphs, we count the number of this kind of dominating sets of arbitrary size in some graphs.
Many techniques have been proposed for image reconstruction in medical imaging that aim to recover high-quality images especially from limited or corrupted measurements. Model-based reconstruction methods have been particularly popular (e.g., in magnetic resonance imaging and tomographic modalities) and exploit models of the imaging system's physics together with statistical models of measurements, noise and often relatively simple object priors or regularizers. For example, sparsity or low-rankness based regularizers have been widely used for image reconstruction from limited data such as in compressed sensing. Learning-based approaches for image reconstruction have garnered much attention in recent years and have shown promise across biomedical imaging applications. These methods include synthesis dictionary learning, sparsifying transform learning, and different forms of deep learning involving complex neural networks. We briefly discuss classical model-based reconstruction methods and then review reconstruction methods at the intersection of model-based and learning-based paradigms in detail. This review includes many recent methods based on unsupervised learning, and supervised learning, as well as a framework to combine multiple types of learned models together.
Common statistical measures of uncertainty such as $p$-values and confidence intervals quantify the uncertainty due to sampling, that is, the uncertainty due to not observing the full population. However, sampling is not the only source of uncertainty. In practice, distributions change between locations and across time. This makes it difficult to gather knowledge that transfers across data sets. We propose a measure of uncertainty or instability that quantifies the distributional instability of a statistical parameter with respect to Kullback-Leibler divergence, that is, the sensitivity of the parameter under general distributional perturbations within a Kullback-Leibler divergence ball. In addition, we propose measures to elucidate the instability of parameters with respect to directional or variable-specific shifts. Measuring instability with respect to directional shifts can be used to detect the type of shifts a parameter is sensitive to. We discuss how such knowledge can inform data collection for improved estimation of statistical parameters under shifted distributions. We evaluate the performance of the proposed measure on real data and show that it can elucidate the distributional (in-)stability of a parameter with respect to certain shifts and can be used to improve the accuracy of estimation under shifted distributions.
We propose the gentle measurement principle (GMP) as one of the principles at the foundation of quantum mechanics. It asserts that if a set of states can be distinguished with high probability, they can be distinguished by a measurement that leaves the states almost invariant, including correlation with a reference system. While GMP is satisfied in both classical and quantum theories, we show, within the framework of general probabilistic theories, that it imposes strong restrictions on the law of physics. First, the measurement uncertainty of a pair of observables cannot be significantly larger than the preparation uncertainty. Consequently, the strength of the CHSH nonlocality cannot be maximal. The parameter in the stretched quantum theory, a family of general probabilistic theories that includes the quantum theory, is also limited. Second, the conditional entropy defined in terms of a data compression theorem satisfies the chain inequality. Not only does it imply information causality and Tsirelson's bound, but it singles out the quantum theory from the stretched one. All these results show that GMP would be one of the principles at the heart of quantum mechanics.
We give some formulas of poly-Cauchy numbers by the $r$-Stirling transform. In the case of the classical or poly-Bernoulli numbers, the formulas are with Stirling numbers of the first kind. In our case of the classical or poly-Cauchy numbers, the formulas are with Stirling numbers of the second kind. We also discuss annihilation formulas for poly-Cauchy number with negative indices.
In this paper, we introduce a complete system for autonomous flight of quadrotors in dynamic environments with onboard sensing. Extended from existing work, we develop an occlusion-aware dynamic perception method based on depth images, which classifies obstacles as dynamic and static. For representing generic dynamic environment, we model dynamic objects with moving ellipsoids and fuse static ones into an occupancy grid map. To achieve dynamic avoidance, we design a planning method composed of modified kinodynamic path searching and gradient-based optimization. The method leverages manually constructed gradients without maintaining a signed distance field (SDF), making the planning procedure finished in milliseconds. We integrate the above methods into a customized quadrotor system and thoroughly test it in realworld experiments, verifying its effective collision avoidance in dynamic environments.
The Banach-Picard iteration is widely used to find fixed points of locally contractive (LC) maps. This paper extends the Banach-Picard iteration to distributed settings; specifically, we assume the map of which the fixed point is sought to be the average of individual (not necessarily LC) maps held by a set of agents linked by a communication network. An additional difficulty is that the LC map is not assumed to come from an underlying optimization problem, which prevents exploiting strong global properties such as convexity or Lipschitzianity. Yet, we propose a distributed algorithm and prove its convergence, in fact showing that it maintains the linear rate of the standard Banach-Picard iteration for the average LC map. As another contribution, our proof imports tools from perturbation theory of linear operators, which, to the best of our knowledge, had not been used before in the theory of distributed computation.
Cosmos has always sparked human curiosity to unveil and speculate its fascinating secrets. This curiosity has ultimately opened a window to other worlds. After years of observation, computation, and data analysis, scientists have revealed the diversity in exoplanets that have been helpful in the characterization and further investigation of biosignatures and technosignatures. This article presents some of the scientific advances made in extraterrestrial planetary science that guarantees to search through the thick clouds of the mysterious planets in the near future.
Let $G$ be a group with identity $e$. Let $R$ be a $G$-graded commutative ring with identity and $M$ a graded $R$-module. In this paper, we introduce the concept of graded $I_{e}$-prime submodule as a generalization of a graded prime submodule for $I=\oplus_{g\in G}I_{g}$ a fixed graded ideal of $R$. We give a number of results concerning of these classes of graded submodules and their homogeneous components. A proper graded submodule $N$ of $M$ is said to be a graded $I_{e}$-prime submodule of $M$ if whenever $% r_{g}\in h(R)$ and $m_{h}\in h(M)$ with $r_{g}m_{h}\in N-I_{e}N,$ then either $r_{g}\in (N:_{R}M)$ or $m_{h}\in N.$
Acquisition of training data for the standard semantic segmentation is expensive if requiring that each pixel is labeled. Yet, current methods significantly deteriorate in weakly supervised settings, e.g. where a fraction of pixels is labeled or when only image-level tags are available. It has been shown that regularized losses - originally developed for unsupervised low-level segmentation and representing geometric priors on pixel labels - can considerably improve the quality of weakly supervised training. However, many common priors require optimization stronger than gradient descent. Thus, such regularizers have limited applicability in deep learning. We propose a new robust trust region approach for regularized losses improving the state-of-the-art results. Our approach can be seen as a higher-order generalization of the classic chain rule. It allows neural network optimization to use strong low-level solvers for the corresponding regularizers, including discrete ones.
We present the confirmation of a new sub-Neptune close to the transition between super-Earths and sub-Neptunes transiting the M2 dwarf TOI- 269 (TIC 220479565, V = 14.4 mag, J = 10.9 mag, Rstar = 0.40 Rsun, Mstar = 0.39 Msun, d = 57 pc). The exoplanet candidate has been identified in multiple TESS sectors, and validated with high-precision spectroscopy from HARPS and ground-based photometric follow-up from ExTrA and LCO-CTIO. We determined mass, radius, and bulk density of the exoplanet by jointly modeling both photometry and radial velocities with juliet. The transiting exoplanet has an orbital period of P = 3.6977104 +- 0.0000037 days, a radius of 2.77 +- 0.12 Rearth, and a mass of 8.8 +- 1.4 Mearth. Since TOI-269 b lies among the best targets of its category for atmospheric characterization, it would be interesting to probe the atmosphere of this exoplanet with transmission spectroscopy in order to compare it to other sub-Neptunes. With an eccentricity e = 0.425+0.082-0.086, TOI-269 b has one of the highest eccentricities of the exoplanets with periods less than 10 days. The star being likely a few Gyr old, this system does not appear to be dynamically young. We surmise TOI-269 b may have acquired its high eccentricity as it migrated inward through planet-planet interactions.
As Artificial Intelligence as a Service gains popularity, protecting well-trained models as intellectual property is becoming increasingly important. Generally speaking, there are two common protection methods: ownership verification and usage authorization. In this paper, we propose Non-Transferable Learning (NTL), a novel approach that captures the exclusive data representation in the learned model and restricts the model generalization ability to certain domains. This approach provides effective solutions to both model verification and authorization. For ownership verification, watermarking techniques are commonly used but are often vulnerable to sophisticated watermark removal methods. Our NTL-based model verification approach instead provides robust resistance to state-of-the-art watermark removal methods, as shown in extensive experiments for four of such methods over the digits, CIFAR10 & STL10, and VisDA datasets. For usage authorization, prior solutions focus on authorizing specific users to use the model, but authorized users can still apply the model to any data without restriction. Our NTL-based authorization approach instead provides data-centric usage protection by significantly degrading the performance of usage on unauthorized data. Its effectiveness is also shown through experiments on a variety of datasets.
Neural networks have emerged as a powerful way to approach many practical problems in quantum physics. In this work, we illustrate the power of deep learning to predict the dynamics of a quantum many-body system, where the training is \textit{based purely on monitoring expectation values of observables under random driving}. The trained recurrent network is able to produce accurate predictions for driving trajectories entirely different than those observed during training. As a proof of principle, here we train the network on numerical data generated from spin models, showing that it can learn the dynamics of observables of interest without needing information about the full quantum state. This allows our approach to be applied eventually to actual experimental data generated from a quantum many-body system that might be open, noisy, or disordered, without any need for a detailed understanding of the system. This scheme provides considerable speedup for rapid explorations and pulse optimization. Remarkably, we show the network is able to extrapolate the dynamics to times longer than those it has been trained on, as well as to the infinite-system-size limit.
We study the dynamics of the quasi-one-dimensional Ising-Heisenberg antiferromagnet BaCo2V2O8 under a transverse magnetic field. Combining inelastic neutron scattering experiments and theoretical analyses by field theories and numerical simulations, we mainly elucidate the structure of the spin excitation spectrum in the high field phase, appearing above the quantum phase transition point mu0Hc ~ 10 T. We find that it is characterized by collective solitonic excitations superimposed on a continuum. These solitons are strongly bound in pairs due to the effective staggered field induced by the nondiagonal g tensor of the compound, and are topologically different from the fractionalized spinons in the weak field region. The dynamical susceptibility numerically calculated with the infinite time-evolving block decimation method shows an excellent agreement with the measured spectra, which enables us to identify the dispersion branches with elementary excitations. The lowest energy dispersion has an incommensurate nature and has a local minimum at an irrational wave number due to the applied transverse field.
In this paper, we explore generalizable, perception-to-action robotic manipulation for precise, contact-rich tasks. In particular, we contribute a framework for closed-loop robotic manipulation that automatically handles a category of objects, despite potentially unseen object instances and significant intra-category variations in shape, size and appearance. Previous approaches typically build a feedback loop on top of a real-time 6-DOF pose estimator. However, representing an object with a parameterized transformation from a fixed geometric template does not capture large intra-category shape variation. Hence we adopt the keypoint-based object representation proposed in kPAM for category-level pick-and-place, and extend it to closed-loop manipulation policies with contact-rich tasks. We first augment keypoints with local orientation information. Using the oriented keypoints, we propose a novel object-centric action representation in terms of regulating the linear/angular velocity or force/torque of these oriented keypoints. This formulation is surprisingly versatile -- we demonstrate that it can accomplish contact-rich manipulation tasks that require precision and dexterity for a category of objects with different shapes, sizes and appearances, such as peg-hole insertion for pegs and holes with significant shape variation and tight clearance. With the proposed object and action representation, our framework is also agnostic to the robot grasp pose and initial object configuration, making it flexible for integration and deployment.
Utilizing intelligent reflecting surface (IRS) was proven to be efficient in improving the energy efficiency for wireless networks. In this paper, we investigate the passive beamforming and channel estimation for IRS assisted wireless communications with low-resolution analog-to-digital converters (ADCs) at the receiver. We derive the approximate achievable rate by using the Bussgang theorem. Based on the derived analytical achievable rate expression, we maximize the achievable rate by using semidefinite programming (SDP), branch-and-bound (BB), and gradient-based approaches. A maximum likelihood (ML) estimator is then proposed for channel estimation by considering the $\mathrm{1}$-bit quantization ADC. Numerical result shows that the proposed beamforming design and channel estimation method significantly outperforms the existing methods.
The impact of ram pressure stripping on galaxy evolution is well known. Recent multi-wavelength data have revealed many examples of galaxies undergoing stripping, often accompanied with multi-phase tails. As energy transfer in the multi-phase medium is an outstanding question in astrophysics, galaxies in stripping are great objects to study. Despite the recent burst of observational evidence, the relationship between gas in different phases in the tails is poorly known. Here we report a strong linear correlation between the X-ray surface brightness and the H$\alpha$ surface brightness of the diffuse gas in the stripped tails at $\sim$ 10 - 40 kpc scales, with a slope of $\sim$ 3.5. This discovery provides evidence for the mixing of the stripped interstellar medium with the hot intracluster medium as the origin of the multi-phase tails. The established relation in stripped tails, also in comparison with the likely related correlations in similar environments like galactic winds and X-ray cool cores, provides an important test for models of energy transfer in the multi-phase gas. It also indicates the importance of the H$\alpha$ data to study clumping and turbulence in the intracluster medium.
Automated metaphor detection is a challenging task to identify metaphorical expressions of words in a sentence. To tackle this problem, we adopt pre-trained contextualized models, e.g., BERT and RoBERTa. To this end, we propose a novel metaphor detection model, namely metaphor-aware late interaction over BERT (MelBERT). Our model not only leverages contextualized word representation but also benefits from linguistic metaphor identification theories to distinguish between the contextual and literal meaning of words. Our empirical results demonstrate that MelBERT outperforms several strong baselines on four benchmark datasets, i.e., VUA-18, VUA-20, MOH-X, and TroFi.
Lending decisions are usually made with proprietary models that provide minimally acceptable explanations to users. In a future world without such secrecy, what decision support tools would one want to use for justified lending decisions? This question is timely, since the economy has dramatically shifted due to a pandemic, and a massive number of new loans will be necessary in the short term. We propose a framework for such decisions, including a globally interpretable machine learning model, an interactive visualization of it, and several types of summaries and explanations for any given decision. The machine learning model is a two-layer additive risk model, which resembles a two-layer neural network, but is decomposable into subscales. In this model, each node in the first (hidden) layer represents a meaningful subscale model, and all of the nonlinearities are transparent. Our online visualization tool allows exploration of this model, showing precisely how it came to its conclusion. We provide three types of explanations that are simpler than, but consistent with, the global model: case-based reasoning explanations that use neighboring past cases, a set of features that were the most important for the model's prediction, and summary-explanations that provide a customized sparse explanation for any particular lending decision made by the model. Our framework earned the FICO recognition award for the Explainable Machine Learning Challenge, which was the first public challenge in the domain of explainable machine learning.
The complex nature of lithium-ion battery degradation has led to many machine learning based approaches to health forecasting being proposed in literature. However, machine learning can be computationally intensive. Linear approaches are faster but have previously been too inflexible for successful prognosis. For both techniques, the choice and quality of the inputs is a limiting factor of performance. Piecewise-linear models, combined with automated feature selection, offer a fast and flexible alternative without being as computationally intensive as machine learning. Here, a piecewise-linear approach to battery health forecasting was compared to a Gaussian process regression tool and found to perform equally well. The input feature selection process demonstrated the benefit of limiting the correlation between inputs. Further trials found that the piecewise-linear approach was robust to changing input size and availability of training data.
Spatial structuring of an optical pulse can lead in some cases upon free propagation to changes in its temporal profile. For example, introducing conventional angular dispersion into the field results in the pulse encountering group-velocity dispersion in free space. However, only limited control is accessible via this strategy. Here we show that precise and versatile control can be exercised in free space over the dispersion profile of so-called `space-time' wave packets: a class of pulsed beams undergirded by non-differentiable angular dispersion. This abstract mathematical feature allows us to tune the magnitude and sign of the different dispersion orders without introducing losses, thereby realizing arbitrary dispersion profiles, and achieving dispersion values unattainable in optical materials away from resonance. Unlike optical materials and photonic structures in which the values of the different dispersion orders are not independent of each other, these orders are addressable separately using our strategy. These results demonstrate the versatility of space-time wave packets as a platform for structured light and point towards their utility in nonlinear and quantum optics.
Collaborative filtering (CF) has achieved great success in the field of recommender systems. In recent years, many novel CF models, particularly those based on deep learning or graph techniques, have been proposed for a variety of recommendation tasks, such as rating prediction and item ranking. These newly published models usually demonstrate their performance in comparison to baselines or existing models in terms of accuracy improvements. However, others have pointed out that many newly proposed models are not as strong as expected and are outperformed by very simple baselines. This paper proposes a simple linear model based on Matrix Factorization (MF), called UserReg, which regularizes users' latent representations with explicit feedback information for rating prediction. We compare the effectiveness of UserReg with three linear CF models that are widely-used as baselines, and with a set of recently proposed complex models that are based on deep learning or graph techniques. Experimental results show that UserReg achieves overall better performance than the fine-tuned baselines considered and is highly competitive when compared with other recently proposed models. We conclude that UserReg can be used as a strong baseline for future CF research.
We establish Arazy-Cwikel type properties for the family of couples $(\ell^{p},\ell^{q})$, $0\le p<q\le\infty$, and show that $(\ell^{p},\ell^{q}) $ is a Calder\'on-Mityagin couple if and only if $q\ge1$. Moreover, we identify interpolation orbits of elements with respect to this couple for all $p$ and $q$ such that $0\le p<q\le\infty$ and obtain a simple positive solution of a Levitina-Sukochev-Zanin problem, clarifying its connections with whether $(\ell^{p},\ell^{q})$ has the Calder\'on-Mityagin property or not.
The online reconstruction of muon tracks in High Energy Physics experiments is a highly demanding task, typically performed with programmable logic boards, such as FPGAs. Complex analytical algorithms are executed in a quasi-real-time environment to identify, select and reconstruct local tracks in often noise-rich environments. A novel approach to the generation of local triggers based on an hybrid combination of Artificial Neural Networks and analytical methods is proposed, targeting the muon reconstruction for drift tube detectors. The proposed algorithm exploits Neural Networks to solve otherwise computationally expensive analytical tasks for the unique identification of coherent signals and the removal of the geometrical ambiguities. The proposed approach is deployed on state-of-the-art FPGA and its performances are evaluated on simulation and on data collected from cosmic rays.
Numerical simulations of monolayer dust crystals in an RF complex plasma were performed to examine the crystal structure and quantify the effects of including the collision enhanced ion current in the charging model. A GEC cell similar to a previous experimental work was modeled for a range of RF voltages, using a continuum description for the plasma and a particle description for dust grains. The time history of each dust grain was monitored. The dust charge was computed using both the OML and the collision enhanced charging (CEC) model applicable to the sheath region. The dust model accounted for the electric force, ion drag force, neutral drag force, gravity, and the ion wake. The CEC model produced a lower charge and lower electric force which agreed better with the experimental data. Then dust crystals composed of 40 to 100 grains were modeled and the levitation height and inter-particle spacing of the resulting crystals was examined. Including the collision enhanced current reduced the inter-particle spacing but only had a minor effect on the levitation height.
We introduce an algebraic model based on the expansion of the determinant of two matrices, one of which is generic, to check the additivity of Z^d-valued set functions. Each individual term of the expansion is deformed through a monomial factor in d indeterminates with exponents defined by the set function. A family of sparse polynomials is derived from Grassmann-Plucker relations, and their compatibility is linked to the factorisation of hyperdeterminants in the ring of Laurent polynomials. It is proved that, in broad generality, this deformation returns a determinantal expansion if and only if it is induced by a diagonal matrix of monomials acting as a kernel into the initial determinant expansion, which guarantees the additivity of the set function. The hypotheses underlying this result are tested through the construction of counterexamples, and their implications are explored in terms of complexity reduction, with special attention to permutations of families of subsets.
The Kahn--Saks inequality is a classical result on the number of linear extensions of finite posets. We give a new proof of this inequality for posets of width two using explicit injections of lattice paths. As a consequence we obtain a $q$-analogue, a multivariate generalization and an equality condition in this case. We also discuss the equality conditions of the Kahn--Saks inequality for general posets and prove several implications between conditions conjectured to be equivalent.
In quantum metrology, nonlinear many-body interactions can enhance the precision of Hamiltonian parameter estimation to surpass the Heisenberg scaling. Here, we consider the estimation of the interaction strength in linear systems with long-range interactions and using the Kitaev chains as a case study, we establish a transition from the Heisenberg to super-Heisenberg scaling in the quantum Fisher information by varying the interaction range. We further show that quantum control can improve the prefactor of the quantum Fisher information. Our results explore the advantage of optimal quantum control and long-range interactions in many-body quantum metrology.
The optical domain is a promising field for physical implementation of neural networks, due to the speed and parallelism of optics. Extreme Learning Machines (ELMs) are feed-forward neural networks in which only output weights are trained, while internal connections are randomly selected and left untrained. Here we report on a photonic ELM based on a frequency-multiplexed fiber setup. Multiplication by output weights can be performed either offline on a computer, or optically by a programmable spectral filter. We present both numerical simulations and experimental results on classification tasks and a nonlinear channel equalization task.
We study the practical consequences of dataset sampling strategies on the performance of recommendation algorithms. Recommender systems are generally trained and evaluated on samples of larger datasets. Samples are often taken in a naive or ad-hoc fashion: e.g. by sampling a dataset randomly or by selecting users or items with many interactions. As we demonstrate, commonly-used data sampling schemes can have significant consequences on algorithm performance -- masking performance deficiencies in algorithms or altering the relative performance of algorithms, as compared to models trained on the complete dataset. Following this observation, this paper makes the following main contributions: (1) characterizing the effect of sampling on algorithm performance, in terms of algorithm and dataset characteristics (e.g. sparsity characteristics, sequential dynamics, etc.); and (2) designing SVP-CF, which is a data-specific sampling strategy, that aims to preserve the relative performance of models after sampling, and is especially suited to long-tail interaction data. Detailed experiments show that SVP-CF is more accurate than commonly used sampling schemes in retaining the relative ranking of different recommendation algorithms.
In this work we leverage commonsense knowledge in form of knowledge paths to establish connections between sentences, as a form of explicitation of implicit knowledge. Such connections can be direct (singlehop paths) or require intermediate concepts (multihop paths). To construct such paths we combine two model types in a joint framework we call Co-nnect: a relation classifier that predicts direct connections between concepts; and a target prediction model that generates target or intermediate concepts given a source concept and a relation, which we use to construct multihop paths. Unlike prior work that relies exclusively on static knowledge sources, we leverage language models finetuned on knowledge stored in ConceptNet, to dynamically generate knowledge paths, as explanations of implicit knowledge that connects sentences in texts. As a central contribution we design manual and automatic evaluation settings for assessing the quality of the generated paths. We conduct evaluations on two argumentative datasets and show that a combination of the two model types generates meaningful, high-quality knowledge paths between sentences that reveal implicit knowledge conveyed in text.
Continuous practices that rely on automation in the software development workflow have been widely adopted by industry for over a decade. Despite this widespread use, software development remains a primarily human-driven activity that is highly creative and collaborative. There has been extensive research on how continuous practices rely on automation and its impact on software quality and development velocity, but relatively little has been done to understand how automation impacts developer behavior and collaboration. In this paper, we introduce a socio-technical theory about continuous practices. The ADEPT theory combines constructs that include humans, processes, documentation, automation and the project environment, and describes propositions that relate these constructs. The theory was derived from phenomena observed in previous empirical studies. We show how the ADEPT theory can explain and describe existing continuous practices in software development, and how it can be used to generate new propositions for future studies to understand continuous practices and their impact on the social and technical aspects of software development.
In this paper, the relativistic quantum dynamics of a scalar particle under the effect of Lorentz symmetry violation determined by a tensor $(K_{F})_{\mu\,\nu\,\alpha\,\beta}$ out of the Standard Model Extension is investigated. We see that the bound-state solution of the modified Klein-Gordon equation can be obtained, and the spectrum of energy and the wave function depends on the Lorentz symmetry breaking parameters
We study the fundamental problem of frequency estimation under both privacy and communication constraints, where the data is distributed among $k$ parties. We consider two application scenarios: (1) one-shot, where the data is static and the aggregator conducts a one-time computation; and (2) streaming, where each party receives a stream of items over time and the aggregator continuously monitors the frequencies. We adopt the model of multiparty differential privacy (MDP), which is more general than local differential privacy (LDP) and (centralized) differential privacy. Our protocols achieve optimality (up to logarithmic factors) permissible by the more stringent of the two constraints. In particular, when specialized to the $\varepsilon$-LDP model, our protocol achieves an error of $\sqrt{k}/(e^{\Theta(\varepsilon)}-1)$ using $O(k\max\{ \varepsilon, \frac{1}{\varepsilon} \})$ bits of communication and $O(k \log u)$ bits of public randomness, where $u$ is the size of the domain.
Focal plane wavefront sensing (FPWFS) is appealing for several reasons. Notably, it offers high sensitivity and does not suffer from non-common path aberrations (NCPA). The price to pay is a high computational burden and the need for diversity to lift any phase ambiguity. If those limitations can be overcome, FPWFS is a great solution for NCPA measurement, a key limitation for high-contrast imaging, and could be used as adaptive optics wavefront sensor. Here, we propose to use deep convolutional neural networks (CNNs) to measure NCPA based on focal plane images. Two CNN architectures are considered: ResNet-50 and U-Net which are used respectively to estimate Zernike coefficients or directly the phase. The models are trained on labelled datasets and evaluated at various flux levels and for two spatial frequency contents (20 and 100 Zernike modes). In these idealized simulations we demonstrate that the CNN-based models reach the photon noise limit in a large range of conditions. We show, for example, that the root mean squared (rms) wavefront error (WFE) can be reduced to < $\lambda$/1500 for $2 \times 10^6$ photons in one iteration when estimating 20 Zernike modes. We also show that CNN-based models are sufficiently robust to varying signal-to-noise ratio, under the presence of higher-order aberrations, and under different amplitudes of aberrations. Additionally, they display similar to superior performance compared to iterative phase retrieval algorithms. CNNs therefore represent a compelling way to implement FPWFS, which can leverage the high sensitivity of FPWFS over a broad range of conditions.
We analyze the behavior of third-order in time linear and nonlinear sound waves in thermally relaxing fluids and gases as the sound diffusivity vanishes. The nonlinear acoustic propagation is modeled by the Jordan--Moore--Gibson--Thompson equation both in its Westervelt and in its Kuznetsov-type forms, that is, including quadratic nonlinearities of the type $(u^2)_{tt}$ and $ (u_t^2 + |\nabla u|^2)_t$. As it turns out, sufficiently smooth solutions of these equations converge in the energy norm to the solutions of the corresponding inviscid models at a linear rate. Numerical experiments illustrate our theoretical findings.
A growing number of people engage in online health forums, making it important to understand the quality of the advice they receive. In this paper, we explore the role of expertise in responses provided to help-seeking posts regarding mental health. We study the differences between (1) interactions with peers; and (2) interactions with self-identified mental health professionals. First, we show that a classifier can distinguish between these two groups, indicating that their language use does in fact differ. To understand this difference, we perform several analyses addressing engagement aspects, including whether their comments engage the support-seeker further as well as linguistic aspects, such as dominant language and linguistic style matching. Our work contributes toward the developing efforts of understanding how health experts engage with health information- and support-seekers in social networks. More broadly, it is a step toward a deeper understanding of the styles of interactions that cultivate supportive engagement in online communities.
The NA62 experiment at CERN reports searches for $K^+\to\mu^+N$ and $K^+\to\mu^+\nu X$ decays, where $N$ and $X$ are massive invisible particles, using the 2016-2018 data set. The $N$ particle is assumed to be a heavy neutral lepton, and the results are expressed as upper limits of ${\cal O}(10^{-8})$ of the neutrino mixing parameter $|U_{\mu4}|^2$ for $N$ masses in the range 200-384 MeV/$c^2$ and lifetime exceeding 50 ns. The $X$ particle is considered a scalar or vector hidden sector mediator decaying to an invisible final state, and upper limits of the decay branching fraction for $X$ masses in the range 10-370 MeV/$c^2$ are reported for the first time, ranging from ${\cal O}(10^{-5})$ to ${\cal O}(10^{-7})$. An improved upper limit of $1.0\times 10^{-6}$ is established at 90% CL on the $K^+\to\mu^+\nu\nu\bar\nu$ branching fraction.
Spoken Language Understanding (SLU) aims to extract the semantics frame of user queries, which is a core component in a task-oriented dialog system. With the burst of deep neural networks and the evolution of pre-trained language models, the research of SLU has obtained significant breakthroughs. However, there remains a lack of a comprehensive survey summarizing existing approaches and recent trends, which motivated the work presented in this article. In this paper, we survey recent advances and new frontiers in SLU. Specifically, we give a thorough review of this research field, covering different aspects including (1) new taxonomy: we provide a new perspective for SLU filed, including single model vs. joint model, implicit joint modeling vs. explicit joint modeling in joint model, non pre-trained paradigm vs. pre-trained paradigm;(2) new frontiers: some emerging areas in complex SLU as well as the corresponding challenges; (3) abundant open-source resources: to help the community, we have collected, organized the related papers, baseline projects and leaderboard on a public website where SLU researchers could directly access to the recent progress. We hope that this survey can shed a light on future research in SLU field.
When detecting anomalies in audio, it can often be necessary to consider concept drift: the distribution of the data may drift over time because of dynamically changing environments, and anomalies may become normal as time elapses. We propose to use adaptive Huffman coding for anomaly detection in audio with concept drift. Compared with the existing method of adaptive Gaussian mixture modeling (AGMM), adaptive Huffman coding does not require a priori information about the clusters and can adjust the number of clusters dynamically depending on the amount of variation in the audio. To control the size of the Huffman tree, we propose to merge clusters that are close to each other instead of replacing rare clusters with new data. This reduces redundancy in the Huffman tree while ensuring that it never forgets past information. On a dataset of audio with concept drift which we have curated ourselves, our proposed method achieves higher area under the curve (AUC) compared with AGMM and fixed-length Huffman trees. The proposed approach is also time-efficient and can be easily extended to other types of time series data (e.g., video).
The Supergiant X-ray binary Vela X-1 represents one of the best astrophysical sources to investigate the wind environment of a O/B star irradiated by an accreting neutron star. Previous studies and hydrodynamic simulations of the system revealed a clumpy environment and the presence of two wakes: an accretion wake surrounding the compact object and a photoionisation wake trailing it along the orbit. Our goal is to conduct, for the first time, high-resolution spectroscopy on Chandra/HETG data at the orbital phase $\varphi_\mathrm{orb} \approx 0.75$, when the line of sight is crossing the photoionisation wake. We aim to conduct plasma diagnostics, inferring the structure and the geometry of the wind. We perform a blind search employing a Bayesian Block algorithm to find discrete spectral features and identify them thanks to the most recent laboratory results or through atomic databases. Plasma properties are inferred both with empirical techniques and with photoionisation models within CLOUDY and SPEX. We detect and identify five narrow radiative recombination continua (Mg XI-XII, Ne IX-X, O VIII) and several emission lines from Fe, S, Si, Mg, Ne, Al, and Na, including four He-like triplets (S XV, Si XIII, Mg XI, and Ne IX). Photoionisation models well reproduce the overall spectrum, except for the near-neutral fluorescence lines of Fe, S, and Si. We conclude that the plasma is mainly photoionised, but more than one component is most likely present, consistent with a multi-phase plasma scenario, where denser and colder clumps of matter are embedded in the hot, photoionised wind of the companion star. Simulations with the future X-ray satellites Athena and XRISM show that a few hundred seconds of exposure will be sufficient to disentangle the lines of the Fe K$\alpha$ doublet and the He-like Fe XXV, improving, in general, the determination of the plasma parameters.
We determine the asymptotic behaviour of certain incomplete Betafunctions.
We generalise a result of Kazarian regarding Kadomtsev-Petviashvili integrability for single Hodge integrals to general cohomological field theories related to Hurwitz-type counting problems or hypergeometric tau-functions. The proof uses recent results on the relations between hypergeometric tau-functions and topological recursion, as well as the Eynard-DOSS correspondence between topological recursion and cohomological field theories. In particular, we recover the result of Alexandrov of KP integrability for triple Hodge integrals with a Calabi-Yau condition.
We compute successfully the launching of two magnetic winds from two circumbinary disks formed after a common envelope event. The launching is produced by the increase of magnetic pressure due to the collapse of the disks. The collapse is due to internal torques produced by a weak poloidal magnetic field. The first wind can be described as a wide jet, with an average mass-loss rate of $\sim 1.3 \times 10^{-7}$ \Moy\ and a maximum radial velocity of $\sim 230$ \kms. The outflow has a half-opening angle of $\sim 20^{\circ}$. Narrow jets are also formed intermittently with velocities up to 3,000 \kms, with mass-loss rates of $\sim 6 \times 10^{-12} $ \Moy\ during short periods of time. The second wind can be described as a wide X-wind, with an average mass-loss rate of $\sim 1.68 \times 10^{-7}$ \Moy\ and a velocity of $\sim 30$ \kms. A narrow jet is also formed with a velocity of 250 \kms, and a mass-loss rates of $\sim 10^{-12} $ \Moy. The computed jets are used to provide inflow boundary conditions for simulations of proto-planetary nebulae. The wide jet evolves into a molecular collimated outflow within a few astronomical units, producing proto-planetary nebulae with bipolar, elongated shapes, whose kinetic energies reach $\sim 4 \times 10^{45}$ erg at 1,000 years. Similarities with observed features in W43A, OH231.8+4.2, and Hen 3-1475 are discussed. The computed wide X-wind produces proto-planetary nebulae with slower expansion velocities, with bipolar and elliptical shapes, and possible starfish type and quadrupolar morphology.
Data from Direct Numerical Simulations of disperse bubbly flows in a vertical channel are used to study the effect of the bubbles on the carrier-phase turbulence. A new method is developed, based on the barycentric map approach, that allows to quantify the anisotropy and componentiality of the flow at any scale. Using this the bubbles are found to significantly enhance flow anisotropy at all scales compared with the unladen case, and for some bubble cases, very strong anisotropy persists down to the smallest flow scales. The strongest anisotropy observed was for the cases involving small bubbles. Concerning the inter-scale energy transfer, our results indicate that for the bubble-laden cases, the energy transfer is from large to small scales, just as for the unladen case. However, there is evidence of an upscale transfer when considering the transfer of energy associated with particular components of the velocity field. Although the direction of the energy transfer is the same with and without the bubbles, the transfer is much stronger for the bubble-laden cases, suggesting that the bubbles play a strong role in enhancing the activity of the nonlinear term in the flow. The normalized forms of the fourth and sixth-order structure functions are also considered, and reveal that the introduction of bubbles into the flow strongly enhances intermittency in the dissipation range, but suppresses it at larger scales. This strong enhancement of the dissipation scale intermittency has significant implications for understanding how the bubbles might modify the mixing properties of turbulent flows.
Preference judgments have been demonstrated as a better alternative to graded judgments to assess the relevance of documents relative to queries. Existing work has verified transitivity among preference judgments when collected from trained judges, which reduced the number of judgments dramatically. Moreover, strict preference judgments and weak preference judgments, where the latter additionally allow judges to state that two documents are equally relevant for a given query, are both widely used in literature. However, whether transitivity still holds when collected from crowdsourcing, i.e., whether the two kinds of preference judgments behave similarly remains unclear. In this work, we collect judgments from multiple judges using a crowdsourcing platform and aggregate them to compare the two kinds of preference judgments in terms of transitivity, time consumption, and quality. That is, we look into whether aggregated judgments are transitive, how long it takes judges to make them, and whether judges agree with each other and with judgments from TREC. Our key findings are that only strict preference judgments are transitive. Meanwhile, weak preference judgments behave differently in terms of transitivity, time consumption, as well as of quality of judgment.
In the past decades, DNA has been intensely studied and exploited in different research areas of nanoscience and nanotechnology. At first glance, DNA-based nanophotonics seems to deviate quite far from the original goal of Nadrian Seeman, the founder of DNA nanotechnology, who hoped to organize biological entities using DNA in high-resolution crystals. As a matter of fact, DNA-based nanophotonics does closely follow his central spirit. That is, apart from being a genetic material for inheritance, DNA is also an ideal material for building molecular devices.
In this paper, we present an autonomous navigation system for goal-driven exploration of unknown environments through deep reinforcement learning (DRL). Points of interest (POI) for possible navigation directions are obtained from the environment and an optimal waypoint is selected, based on the available data. Following the waypoints, the robot is guided towards the global goal and the local optimum problem of reactive navigation is mitigated. Then, a motion policy for local navigation is learned through a DRL framework in a simulation. We develop a navigation system where this learned policy is integrated into a motion planning stack as the local navigation layer to move the robot between waypoints towards a global goal. The fully autonomous navigation is performed without any prior knowledge while a map is recorded as the robot moves through the environment. Experiments show that the proposed method has an advantage over similar exploration methods, without reliance on a map or prior information in complex static as well as dynamic environments.
In nanopore sequencing, electrical signal is measured as DNA molecules pass through the sequencing pores. Translating these signals into DNA bases (base calling) is a highly non-trivial task, and its quality has a large impact on the sequencing accuracy. The most successful nanopore base callers to date use convolutional neural networks (CNN) to accomplish the task. Convolutional layers in CNNs are typically composed of filters with constant window size, performing best in analysis of signals with uniform speed. However, the speed of nanopore sequencing varies greatly both within reads and between sequencing runs. Here, we present dynamic pooling, a novel neural network component, which addresses this problem by adaptively adjusting the pooling ratio. To demonstrate the usefulness of dynamic pooling, we developed two base callers: Heron and Osprey. Heron improves the accuracy beyond the experimental high-accuracy base caller Bonito developed by Oxford Nanopore. Osprey is a fast base caller that can compete in accuracy with Guppy high-accuracy mode, but does not require GPU acceleration and achieves a near real-time speed on common desktop CPUs. Availability: https://github.com/fmfi-compbio/osprey, https://github.com/fmfi-compbio/heron Keywords: nanopore sequencing, base calling, convolutional neural networks, pooling
Deep neural network (DNN)-based speech enhancement ordinarily requires clean speech signals as the training target. However, collecting clean signals is very costly because they must be recorded in a studio. This requirement currently restricts the amount of training data for speech enhancement to less than 1/1000 of that of speech recognition which does not need clean signals. Increasing the amount of training data is important for improving the performance, and hence the requirement of clean signals should be relaxed. In this paper, we propose a training strategy that does not require clean signals. The proposed method only utilizes noisy signals for training, which enables us to use a variety of speech signals in the wild. Our experimental results showed that the proposed method can achieve the performance similar to that of a DNN trained with clean signals.
Deep Metric Learning (DML) is helpful in computer vision tasks. In this paper, we firstly introduce DML into image co-segmentation. We propose a novel Triplet loss for Image Segmentation, called IS-Triplet loss for short, and combine it with traditional image segmentation loss. Different from the general DML task which learns the metric between pictures, we treat each pixel as a sample, and use their embedded features in high-dimensional space to form triples, then we tend to force the distance between pixels of different categories greater than of the same category by optimizing IS-Triplet loss so that the pixels from different categories are easier to be distinguished in the high-dimensional feature space. We further present an efficient triple sampling strategy to make a feasible computation of IS-Triplet loss. Finally, the IS-Triplet loss is combined with 3 traditional image segmentation losses to perform image segmentation. We apply the proposed approach to image co-segmentation and test it on the SBCoseg dataset and the Internet dataset. The experimental result shows that our approach can effectively improve the discrimination of pixels' categories in high-dimensional space and thus help traditional loss achieve better performance of image segmentation with fewer training epochs.
The domain of Embodied AI has recently witnessed substantial progress, particularly in navigating agents within their environments. These early successes have laid the building blocks for the community to tackle tasks that require agents to actively interact with objects in their environment. Object manipulation is an established research domain within the robotics community and poses several challenges including manipulator motion, grasping and long-horizon planning, particularly when dealing with oft-overlooked practical setups involving visually rich and complex scenes, manipulation using mobile agents (as opposed to tabletop manipulation), and generalization to unseen environments and objects. We propose a framework for object manipulation built upon the physics-enabled, visually rich AI2-THOR framework and present a new challenge to the Embodied AI community known as ArmPointNav. This task extends the popular point navigation task to object manipulation and offers new challenges including 3D obstacle avoidance, manipulating objects in the presence of occlusion, and multi-object manipulation that necessitates long term planning. Popular learning paradigms that are successful on PointNav challenges show promise, but leave a large room for improvement.
Houghton's groups $H_2, H_3, \ldots$ are certain infinite permutation groups acting on a countably infinite set; they have been studied, among other things, for their finiteness properties. In this note we describe all of the finite index subgroups of each Houghton group, and their isomorphism types. Using the standard notation that $d(G)$ denotes the minimal size of generating set for $G$ we then show, for each $n\in \{2, 3,\ldots\}$ and $U$ of finite index in $H_n$, that $d(U)\in\{d(H_n), d(H_n)+1\}$ and characterise when each of these cases occurs.
This article aims at developing a model based optimization for reduction of temporal unwrapping and field estimation errors in multi-echo acquisition of Gradient Echo sequence. Using the assumption that the phase is linear along the temporal dimension, the field estimation is performed by application of unity rank approximation to the Hankel matrix formed using the complex exponential of the channel combined phase at each echo time. For the purpose of maintaining consistency with the observed complex data, the linear phase evolution model is formulated as an optimization problem with a cost function that involves a fidelity term and a unity rank prior, implemented using alternating minimization. Itoh s algorithm applied to the multi-echo phase estimated from this linear phase evolution model is able to reduce the unwrapping errors as compared to the unwrapping when directly applied to the measured phase. Secondly, the improved accuracy of the frequency fit in comparison to estimation using weighted least-square regression and penalized maximum likelihood is demonstrated using numerical simulation of field perturbation due to magnetic susceptibility effect. It is shown that the field can be estimated with 80 percent reduction in mean absolute error in comparison to wLSR and 66 percent reduction with respect to penalized maximum likelihood. The improvement in performance becomes more pronounced with increasing strengths of field gradient magnitudes and echo spacing.
Willems' fundamental lemma and system level synthesis both characterize a linear dynamic system by its input/output sequences. In this work, we extend the application of the fundamental lemma from deterministic to uncertain LTI systems and then further prove this extension to be equivalent to system level synthesis. Based on this uncertain extension, a robust closed-loop data-enabled predictive control scheme is proposed, where a causal feedback control law is further derived. Two numerical experiments, including the temperature control of a single-zone building, are carried out to validate the effectiveness of the proposed data-driven controller.
Pretrained language models have shown success in many natural language processing tasks. Many works explore incorporating knowledge into language models. In the biomedical domain, experts have taken decades of effort on building large-scale knowledge bases. For example, the Unified Medical Language System (UMLS) contains millions of entities with their synonyms and defines hundreds of relations among entities. Leveraging this knowledge can benefit a variety of downstream tasks such as named entity recognition and relation extraction. To this end, we propose KeBioLM, a biomedical pretrained language model that explicitly leverages knowledge from the UMLS knowledge bases. Specifically, we extract entities from PubMed abstracts and link them to UMLS. We then train a knowledge-aware language model that firstly applies a text-only encoding layer to learn entity representation and applies a text-entity fusion encoding to aggregate entity representation. Besides, we add two training objectives as entity detection and entity linking. Experiments on the named entity recognition and relation extraction from the BLURB benchmark demonstrate the effectiveness of our approach. Further analysis on a collected probing dataset shows that our model has better ability to model medical knowledge.
Cloud computing has the capacity to transform many parts of the research ecosystem, from particular research areas to overall strategic decision making and policy. Scientometrics sits at the boundary between research and the decision making and evaluation processes of research. One of the biggest challenges in research policy and strategy is having access to data that allows iterative analysis to inform decisions. Many of these decisions are based on "global" measures such as benchmark metrics that are hard to source. In this article, Cloud technologies are explored in this context. A novel visualisation technique is presented and used as a means to explore the potential for scaling scientometrics by democratising both access to data and compute capacity using the Cloud.
We present photometric and spectroscopic observations of Supernova 2020oi (SN 2020oi), a nearby ($\sim$17 Mpc) type-Ic supernova (SN Ic) within the grand-design spiral M100. We undertake a comprehensive analysis to characterize the evolution of SN 2020oi and constrain its progenitor system. We detect flux in excess of the fireball rise model $\delta t \approx 2.5$ days from the date of explosion in multi-band optical and UV photometry from the Las Cumbres Observatory and the Neil Gehrels Swift Observatory, respectively. The derived SN bolometric luminosity is consistent with an explosion with $M_{\rm ej} = 0.81 \pm 0.03 M_{\odot}$, $E_{k}= 0.79 \pm 0.09 \times 10^{51} \rm{erg} \rm{s}^{-1}$, and $M_{\rm Ni56} = 0.08 \pm 0.02 M_{\odot}$. Inspection of the event's decline reveals the highest $\Delta m_{15,\rm{bol}}$ reported for a stripped-envelope event to date. Modeling of optical spectra near event peak indicates a partially mixed ejecta comparable in composition to the ejecta observed in SN 1994I, while the earliest spectrum shows signatures of a possible interaction with material of a distinct composition surrounding the SN progenitor. Further, Hubble Space Telescope (HST) pre-explosion imaging reveals a stellar cluster coincident with the event. From the cluster photometry, we derive the mass and age of the SN progenitor using stellar evolution models implemented in the BPASS library. Our results indicate that SN 2020oi occurred in a binary system from a progenitor of mass $M_{\rm ZAMS} \approx 9.5 \pm 1.0 M_{\odot}$, corresponding to an age of $27 \pm 7$ Myr. SN 2020oi is the dimmest SN Ic event to date for which an early-time flux excess has been observed, and the first in which an early excess is unlikely to be associated with shock-cooling.
Following the increasing interest and adoption of FaaS systems, benchmarking frameworks for determining non-functional properties have also emerged. While existing (microbenchmark) frameworks only evaluate single aspects of FaaS platforms, a more holistic, application-driven approach is still missing. In this paper, we design and present BeFaaS, an extensible application-centric benchmarking framework for FaaS environments that focuses on the evaluation of FaaS platforms through realistic and typical examples of FaaS applications. BeFaaS includes a built-in e-commerce benchmark, is extensible for new workload profiles and new platforms, supports federated benchmark runs in which the benchmark application is distributed over multiple providers, and supports a fine-grained result analysis. Our evaluation compares three major FaaS providers in single cloud provider setups and shows that BeFaaS is capable of running each benchmark automatically with minimal configuration effort and providing detailed insights for each interaction.
Confluent cell monolayers and epithelia tissues show remarkable patterns and correlations in structural arrangements and actively-driven collective flows. We simulate these properties using multiphase field models. The models are based on cell deformations and cell-cell interactions and we investigate the influence of microscopic details to incorporate active forces on emerging phenomena. We compare four different approaches, one in which the activity is determined by a random orientation, one where the activity is related to the deformation of the cells and two models with subcellular details to resolve the mechanochemical interactions underlying cell migration. The models are compared with respect to generic features, such as solid-to-liquid phase transitions, cell shape variability, emerging nematic properties, as well as vorticity correlations and flow patterns in large confluent monolayers and confinements. All results are compared with experimental data for a large variety of cell cultures. The appearing qualitative differences of the models show the importance of microscopic details and provide a route towards predictive simulations of patterns and correlations in cell colonies.
In this paper we generalize the definition of rationalizability for square roots of polynomials introduced by M. Besier and the first author to field extensions. We then show that the rationalizability of a set of field extensions is equivalent to the rationalizability of the compositum of the field extensions, providing a new strategy to prove rationalizability of sets of square roots of polynomials.
We show how to infer sharp partial regularity results for relaxed minimizers of degenerate, nonuniformly elliptic quasiconvex functionals, using tools from Nonlinear Potential Theory. In particular, in the setting of functionals with $(p,q)$-growth - according to the terminology of Marcellini [52] - we derive optimal local regularity criteria under minimal assumptions on the data.
Let $G$ be a simple graph with $2n$ vertices and a perfect matching. We denote by $f(G)$ and $F(G)$ the minimum and maximum forcing number of $G$, respectively. Hetyei obtained that the maximum number of edges of graphs $G$ with a unique perfect matching is $n^2$. We know that $G$ has a unique perfect matching if and only if $f(G)=0$. Along this line, we generalize the classical result to all graphs $G$ with $f(G)=k$ for $0\leq k\leq n-1$, and characterize corresponding extremal graphs as well. Hence we get a non-trivial lower bound of $f(G)$ in terms of the order and size. For bipartite graphs, we gain corresponding stronger results. Further, we obtain a new upper bound of $F(G)$. For bipartite graphs $G$, Che and Chen (2013) obtained that $f(G)=n-1$ if and only if $G$ is complete bipartite graph $K_{n,n}$. We completely characterize all bipartite graphs $G$ with $f(G)= n-2$.
We investigate ion-scale kinetic plasma instabilities at the collisionless shock using linear theory and nonlinear Particle-in-Cell (PIC) simulations. We focus on the Alfv\'en-ion-cyclotron (AIC), mirror, and Weibel instabilities, which are all driven unstable by the effective temperature anisotropy induced by the shock-reflected ions within the transition layer of a strictly perpendicular shock. We conduct linear dispersion analysis with a homogeneous plasma model to mimic the shock transition layer by adopting a ring distribution with finite thermal spread to represent the velocity distribution of the reflected ions. We find that, for wave propagation parallel to the ambient magnetic field, the AIC instability at lower Alfv\'en Mach numbers tends to transition to the Weibel instability at higher Alfv\'en Mach numbers. The instability property is, however, also strongly affected by the sound Mach number. We conclude that the instability at a strong shock with Alfv\'en and sound Mach numbers both in excess of $\sim 20{\rm -}40$ may be considered as Weibel-like in the sense that the reflected ions behave essentially unmagnetized. Two-dimensional PIC simulations confirm the linear theory and find that, with typical parameters of young supernova remnant shocks, the ring distribution model produces magnetic fluctuations of the order of the background magnetic field, which is smaller than those observed in previous PIC simulations for Weibel-dominated shocks. This indicates that the assumption of the gyrotropic reflected ion distribution may not be adequate to quantitatively predict nonlinear behaviors of the dynamics in high Mach number shocks.
The objective of this paper is to construct the accurate (say, to 11 decimal places) frequencies of the quasinormal modes of the 5-dimensional Schwarzschild-Tangherlini black hole using three major techniques: the Hill determinant method, the continued fractions method and the WKB-Pad\'e method and to discuss the limitations of each. It is shown that for the massless scalar, gravitational tensor, gravitational vector and electromagnetic vector perturbations considered in this paper, the Hill determinant method and the method of continued fractions (both with the convergence acceleration) always give identical results, whereas the WKB-Pad\'e method gives the results that are amazingly accurate in most cases. Notable exception are the gravitational vector perturbations ($j =2$ and $\ell = 2 $), for which the WKB-Pad\'e approach apparently does not work. Here we have interesting situation in which the WKB-based methods (WKB-Pad\'e and WKB-Borel-Le Roy) give the complex frequency that differs from the from the result obtained within the framework of the continued fraction method and the Hill determinant method. For the fundamental mode, deviation of the real part of frequency from the exact value is $0.5\%$ whereas the deviation of the imaginary part is $2.7\%.$ For $\ell \geq 3$ the accuracy of the WKB results is similar again to the accuracy obtained for other perturbations. The case of the gravitational scalar perturbations is briefly discussed.
In order to plan a safe maneuver, self-driving vehicles need to understand the intent of other traffic participants. We define intent as a combination of discrete high-level behaviors as well as continuous trajectories describing future motion. In this paper, we develop a one-stage detector and forecaster that exploits both 3D point clouds produced by a LiDAR sensor as well as dynamic maps of the environment. Our multi-task model achieves better accuracy than the respective separate modules while saving computation, which is critical to reducing reaction time in self-driving applications.
We study the the impact of Run2 LHC data on general Composite Higgs scenarios, where non-linear effects, mixing with additional scalars and new fermionic degrees of freedom could simultaneously contribute to the modification of Higgs properties. We obtain new experimental limits on the scale of compositeness, the mixing with singlets and doublets with the Higgs, and the mass and mixing angle of top-partners. We also show that for scenarios where new fermionic degrees of freedom are involved in electroweak symmetry breaking, there is an interesting interplay among Higgs coupling measurements, boosted Higgs properties, SMEFT global analyses, and direct searches for single- and double-production of vector-like quarks.
Compressive sensing (CS) is a mathematically elegant tool for reducing the sampling rate, potentially bringing context-awareness to a wider range of devices. Nevertheless, practical issues with the sampling and reconstruction algorithms prevent further proliferation of CS in real world domains, especially among heterogeneous ubiquitous devices. Deep learning (DL) naturally complements CS for adapting the sampling matrix, reconstructing the signal, and learning form the compressed samples. While the CS-DL integration has received substantial research interest recently, it has not yet been thoroughly surveyed, nor has the light been shed on practical issues towards bringing the CS-DL to real world implementations in the ubicomp domain. In this paper we identify main possible ways in which CS and DL can interplay, extract key ideas for making CS-DL efficient, identify major trends in CS-DL research space, and derive guidelines for future evolution of CS-DL within the ubicomp domain.
To respond to the increasing need for bone repair strategies, various types of biomaterials have been developed. Among those, calcium phosphate ceramics (CPCs) are promising since they possess a chemical composition similar to that of bones. To be suitable for implants, CPCs need to fulfill a number of biological and mechanical requirements. Fatigue resistance and toughness are two key mechanical properties that are still challenging to obtain in CPCs. This paper thus reviews and discusses current progress in the processing of CPCs with bioinspired microstructures for load-bearing applications. First, methods to obtain CPCs with bioinspired structure at individual lengthscales, namely nano-, micro-, and macroscale are discussed. Then, approaches to attain synergetic contribution of all lengthscales through a complex and biomimetic hierarchical structure are reviewed. The processing methods and their design capabilities are presented and the mechanical properties of the materials they can produce are analysed. Their limitations and challenges are finally discussed to suggest new directions for the fabrication of biomimetic bone implants with satisfactory properties. The paper could help biomedical researchers, materials scientists and engineers to join forces to create the next generation of bone implants.
The supernova neutrino flavor evolution in the presence of the non-trivial neutrino magnetic moment and strong magnetic field is numerically derived using the two-flavor and single-angle approximation. The novel properties of collective neutrino oscillations are studied and distinct patterns of flavor and spin-flavor spectral splits are presented. Finally we also discuss how the neutrino magnetic moment affects the observable supernova neutrino energy spectra.
Under the federated learning paradigm, a set of nodes can cooperatively train a machine learning model with the help of a centralized server. Such a server is also tasked with assigning a weight to the information received from each node, and often also to drop too-slow nodes from the learning process. Both decisions have major impact on the resulting learning performance, and can interfere with each other in counterintuitive ways. In this paper, we focus on edge networking scenarios and investigate existing and novel approaches to such model-weighting and node-dropping decisions. Leveraging a set of real-world experiments, we find that popular, straightforward decision-making approaches may yield poor performance, and that considering the quality of data in addition to its quantity can substantially improve learning.
We prove a myriad of results related to the stabilizer in an algebraic group $G$ of a generic vector in a representation $V$ of $G$ over an algebraically closed field $k$. Our results are on the level of group schemes, which carries more information than considering both the Lie algebra of $G$ and the group $G(k)$ of $k$-points. For $G$ simple and $V$ faithful and irreducible, we prove the existence of a stabilizer in general position, sometimes called a principal orbit type. We determine those $G$ and $V$ for which the stabilizer in general position is smooth, or $\dim V/G < \dim G$, or there is a $v \in V$ whose stabilizer in $G$ is trivial.
Recent development of lensless imagers has enabled three-dimensional (3D) imaging through a thin piece of optics in close proximity to a camera sensor. A general challenge of wide-field lensless imaging is the high computational complexity and slow speed to reconstruct 3D objects through iterative optimization process. Here, we demonstrated GEOMScope, a lensless 3D microscope that forms image through a single layer of microlens array and reconstructs objects through a geometrical-optics-based pixel back projection algorithm and background suppressions. Compared to others, our method allows local reconstruction, which significantly reduces the required computation resource and increases the reconstruction speed by orders of magnitude. This enables near real-time object reconstructions across a large volume of 23x23x5 mm^3, with a lateral resolution of 40 um and axial resolution of 300 um. Our system opens new avenues for broad biomedical applications such as endoscopy, which requires both miniaturized device footprint and real-time high resolution visualization.
We study the additive functional $X_n(\alpha)$ on conditioned Galton-Watson trees given, for arbitrary complex $\alpha$, by summing the $\alpha$th power of all subtree sizes. Allowing complex $\alpha$ is advantageous, even for the study of real $\alpha$, since it allows us to use powerful results from the theory of analytic functions in the proofs. For $\Re\alpha < 0$, we prove that $X_n(\alpha)$, suitably normalized, has a complex normal limiting distribution; moreover, as processes in $\alpha$, the weak convergence holds in the space of analytic functions in the left half-plane. We establish, and prove similar process-convergence extensions of, limiting distribution results for $\alpha$ in various regions of the complex plane. We focus mainly on the case where $\Re\alpha > 0$, for which $X_n(\alpha)$, suitably normalized, has a limiting distribution that is not normal but does not depend on the offspring distribution $\xi$ of the conditioned Galton-Watson tree, assuming only that $E[\xi] = 1$ and $0 < \mathrm{Var} [\xi] < \infty$. Under a weak extra moment assumption on $\xi$, we prove that the convergence extends to moments, ordinary and absolute and mixed, of all orders. At least when $\Re\alpha > \frac12$, the limit random variable $Y(\alpha)$ can be expressed as a function of a normalized Brownian excursion.
MINERvA presents a new analysis of inclusive charged-current neutrino interactions on a hydrocarbon target. We report single and double-differential cross sections in muon transverse and longitudinal momentum. These measurements are compared to neutrino interaction generator predictions from GENIE, NuWro, GiBUU, and NEUT. In addition, comparisons against models with different treatments of multi-nucleon correlations, nuclear effects, resonant pion production, and deep inelastic scattering are presented. The data recorded corresponds to $10.61\times10^{20}$ protons on target with a peak neutrino energy of approximately 6 GeV. The higher energy and larger statistics of these data extend the kinematic range for model testing beyond previous MINERvA inclusive charged-current measurements. The results are not well modeled by several generator predictions using a variety of input models.
Square-root topology is a recently emerged subfield describing a class of insulators and superconductors whose topological nature is only revealed upon squaring their Hamiltonians, i.e., the finite energy edge states of the starting square-root model inherit their topological features from the zero-energy edge states of a known topological insulator/superconductor present in the squared model. Focusing on one-dimensional models, we show how this concept can be generalized to $2^n$-root topological insulators and superconductors, with $n$ any positive integer, whose rules of construction are systematized here. Borrowing from graph theory, we introduce the concept of arborescence of $2^n$-root topological insulators/superconductors which connects the Hamiltonian of the starting model for any $n$, through a series of squaring operations followed by constant energy shifts, to the Hamiltonian of the known topological insulator/superconductor, identified as the source of its topological features. Our work paves the way for an extension of $2^n$-root topology to higher-dimensional systems.
A subset of QuantISED Sensor PIs met virtually on May 26, 2020 to discuss a response to a charge by the DOE Office of High Energy Physics. In this document, we summarize the QuantISED sensor community discussion, including a consideration of HEP science enabled by quantum sensors, describing the distinction between Quantum 1.0 and Quantum 2.0, and discussing synergies/complementarity with the new DOE NQI centers and with research supported by other SC offices. Quantum 2.0 advances in sensor technology offer many opportunities and new approaches for HEP experiments. The DOE HEP QuantISED program could support a portfolio of small experiments based on these advances. QuantISED experiments could use sensor technologies that exemplify Quantum 2.0 breakthroughs. They would strive to achieve new HEP science results, while possibly spinning off other domain science applications or serving as pathfinders for future HEP science targets. QuantISED experiments should be led by a DOE laboratory, to take advantage of laboratory technical resources, infrastructure, and expertise in the safe and efficient construction, operation, and review of experiments. The QuantISED PIs emphasized that the quest for HEP science results under the QuantISED program is distinct from the ongoing DOE HEP programs on the energy, intensity, and cosmic frontiers. There is robust evidence for the existence of particles and phenomena beyond the Standard Model, including dark matter, dark energy, quantum gravity, and new physics responsible for neutrino masses, cosmic inflation, and the cosmic preference for matter over antimatter. Where is this physics and how do we find it? The QuantISED program can exploit new capabilities provided by quantum technology to probe these kinds of science questions in new ways and over a broader range of science parameters than can be achieved with conventional techniques.
This article concerns with the global H\"older regularity of weak solutions to a class of problems involving the fractional $(p,q)$-Laplacian, denoted by $(-\Delta)^{s_1}_{p}+(-\Delta)^{s_2}_{q}$, for $1<p,q<\infty$ and $s_1,s_2\in (0,1)$. We use a suitable Caccioppoli inequality and local boundedness result in order to prove the weak Harnack type inequality. Consequently, by employing a suitable iteration process, we establish the interior H\"older regularity for local weak solutions, which need not be assumed bounded. The global H\"older regularity result we prove expands and improves the regularity results of Giacomoni, Kumar and Sreenadh (arXiv: 2102.06080) to the subquadratic case (that is, $q<2$) and more general right hand side, which requires a different and new approach. Moreover, we establish a nonlocal Harnack type inequality for weak solutions, which is of independent interest.
This work investigates the use of interactively updated label suggestions to improve upon the efficiency of gathering annotations on the task of opinion mining in German Covid-19 social media data. We develop guidelines to conduct a controlled annotation study with social science students and find that suggestions from a model trained on a small, expert-annotated dataset already lead to a substantial improvement - in terms of inter-annotator agreement(+.14 Fleiss' $\kappa$) and annotation quality - compared to students that do not receive any label suggestions. We further find that label suggestions from interactively trained models do not lead to an improvement over suggestions from a static model. Nonetheless, our analysis of suggestion bias shows that annotators remain capable of reflecting upon the suggested label in general. Finally, we confirm the quality of the annotated data in transfer learning experiments between different annotator groups. To facilitate further research in opinion mining on social media data, we release our collected data consisting of 200 expert and 2,785 student annotations.
It is well-known that the training of Deep Neural Networks (DNN) can be formalized in the language of optimal control. In this context, this paper leverages classical turnpike properties of optimal control problems to attempt a quantifiable answer to the question of how many layers should be considered in a DNN. The underlying assumption is that the number of neurons per layer -- i.e., the width of the DNN -- is kept constant. Pursuing a different route than the classical analysis of approximation properties of sigmoidal functions, we prove explicit bounds on the required depths of DNNs based on asymptotic reachability assumptions and a dissipativity-inducing choice of the regularization terms in the training problem. Numerical results obtained for the two spiral task data set for classification indicate that the proposed estimates can provide non-conservative depth bounds.
The bulk properties of nodal line materials have been an important research topic in recent years. In this paper, we study the orbital magnetic susceptibility and the Hall conductivity of nodal line materials using the formalism with thermal Green's functions and find characteristic singular behaviors of them. It is shown that, in the vicinity of the gapless nodal line, the orbital magnetic susceptibility shows a $\delta$-function singularity and the Hall conductivity shows a step function behavior in their chemical potential dependences. Furthermore, these singular behaviors are found to show strong field angle dependences corresponding to the orientation of the nodal line in the momentum space. These singular behaviors and strong angle dependences will give clear evidence for the presence of the nodal line and its orientation and can be used to experimentally detect nodal line materials.
The past 20 years have witnessed a renewal of interest in the subject of double-diffusive processes in astrophysics, and their impact on stellar evolution. This lecture aims to summarize the state of the field as of early 2019, although the reader should bear in mind that it is rapidly evolving. An Annual Review of Fluid Mechanics article entitled "Double-diffusive convection at low Prandtl number" (Garaud, 2018) contains a reasonably comprehensive review of the topic, up to the summer of 2017. I focus here on presenting what I hope are clear derivations of some of the most important results with an astrophysical audience in mind, and discuss their implications for stellar evolution, both in an observational context, and in relation to previous work on the subject.
Variation of positional information, measured by the two-body excess entropy $\mathsf{S}_\mathrm{2}$, is studied across the liquid-solid equilibrium transition in a simple two-dimensional system. Analysis reveals a master relation between $\mathsf{S}_\mathrm{2}$ and the freezing temperature $T_{\mathrm{f}}$, from which a scaling law is extracted: $-\mathsf{S}_\mathrm{2}\sim |T_{\mathrm{f}} - T| ^{-1/3}$. Theoretical and practical implications of the observed universality are discussed.
Large pre-trained multilingual models like mBERT, XLM-R achieve state of the art results on language understanding tasks. However, they are not well suited for latency critical applications on both servers and edge devices. It's important to reduce the memory and compute resources required by these models. To this end, we propose pQRNN, a projection-based embedding-free neural encoder that is tiny and effective for natural language processing tasks. Without pre-training, pQRNNs significantly outperform LSTM models with pre-trained embeddings despite being 140x smaller. With the same number of parameters, they outperform transformer baselines thereby showcasing their parameter efficiency. Additionally, we show that pQRNNs are effective student architectures for distilling large pre-trained language models. We perform careful ablations which study the effect of pQRNN parameters, data augmentation, and distillation settings. On MTOP, a challenging multilingual semantic parsing dataset, pQRNN students achieve 95.9\% of the performance of an mBERT teacher while being 350x smaller. On mATIS, a popular parsing task, pQRNN students on average are able to get to 97.1\% of the teacher while again being 350x smaller. Our strong results suggest that our approach is great for latency-sensitive applications while being able to leverage large mBERT-like models.
In this paper, we present general one-loop form factors for $H\rightarrow \gamma^* \gamma^*$ in $R_{\xi}$ gauge, considering all cases of two on-shell, one on-shell and two off-shell for final photons. The calculations are performed in standard model and in arbitrary beyond the standard models which charged scalar particles may be exchanged in one-loop diagrams. Analytic results for the form factors are shown in general forms which are expressed in terms of the Passarino-Veltman functions. We also confirm the results in previous computations which are available for the case of two on-shell photons. The $\xi$-independent of the result is also discussed. We find that numerical results are good stability with varying $\xi=0,1$ and $\xi\rightarrow \infty$.