Datasets:

ds3lab
/

ac-sgd-arxiv21

Dataset card Data Studio Files Files and versions Community

Split (2)

train · 30k rows

abstract stringlengths 42 2.09k
We show how phase-space simulations of Gaussian quantum states in a photonic network permit verification of measurable correlations of Gaussian boson sampling (GBS) quantum computers. Our results agree with experiments for up to 100-th order correlations, provided decoherence is included. We extend this to more than 16,000 modes, and describe how to simulate genuine multipartite entanglement.
A fundamental task in AI is to assess (in)dependence between mixed-type variables (text, image, sound). We propose a Bayesian kernelised correlation test of (in)dependence using a Dirichlet process model. The new measure of (in)dependence allows us to answer some fundamental questions: Based on data, are (mixed-type) variables independent? How likely is dependence/independence to hold? How high is the probability that two mixed-type variables are more than just weakly dependent? We theoretically show the properties of the approach, as well as algorithms for fast computation with it. We empirically demonstrate the effectiveness of the proposed method by analysing its performance and by comparing it with other frequentist and Bayesian approaches on a range of datasets and tasks with mixed-type variables.
Superconductivity in a crystalline lattice without inversion is subject to complex spin-orbit-coupling effects, which can lead to mixed-parity pairing and an unusual magnetic response. In this study, the properties of a layered superconductor with alternating Rashba spin-orbit coupling in the stacking of layers, hence (globally) possessing a center of inversion, is analyzed in an applied magnetic field, using a generalized Ginzburg-Landau model. The superconducting order parameter consists of an even- and an odd-parity pairing component which exchange their roles as dominant pairing channel upon increasing the magnetic field. This leads to an unusual kink feature in the upper critical field and a first-order phase transition within the mixed phase. We investigate various signatures of this internal phase transition. The physics we discuss here could explain the recently found $H$--$T$ phase diagram of the heavy Fermion superconductor CeRh$_2$As$_2$.
Recent studies strive to incorporate various human rationales into neural networks to improve model performance, but few pay attention to the quality of the rationales. Most existing methods distribute their models' focus to distantly-labeled rationale words entirely and equally, while ignoring the potential important non-rationale words and not distinguishing the importance of different rationale words. In this paper, we propose two novel auxiliary loss functions to make better use of distantly-labeled rationales, which encourage models to maintain their focus on important words beyond labeled rationales (PINs) and alleviate redundant training on non-helpful rationales (NoIRs). Experiments on two representative classification tasks show that our proposed methods can push a classification model to effectively learn crucial clues from non-perfect rationales while maintaining the ability to spread its focus to other unlabeled important words, thus significantly outperform existing methods.
A novel framework is proposed to extract near-threshold resonant states from finite-volume energy levels of lattice QCD and is applied to elucidate structures of the positive parity $D_s$. The quark model, the quark-pair-creation mechanism and $D^{()}K$ interaction are incorporated into the Hamiltonian effective field theory. The bare $1^+$ $c\bar s$ states are almost purely given by the states with heavy-quark spin bases. The physical $D^_{s0}(2317)$ and $D^_{s1}(2460)$ are the mixtures of bare $c\bar s$ core and $D^{()}K$ component, while the $D^_{s1}(2536)$ and $D^_{s2}(2573)$ are almost dominated by bare $c\bar{s}$. Furthermore, our model reproduces the clear level crossing of the $D^*_{s1}(2536)$ with the scattering state at a finite volume.
As the worldwide population gets increasingly aged, in-home telemedicine and mobile-health solutions represent promising services to promote active and independent aging and to contribute to a paradigm shift towards patient-centric healthcare. In this work, we present ACTA (Advanced Cognitive Training for Aging), a prototype mobile-health solution to provide advanced cognitive training for senior citizens with mild cognitive impairments. We disclose here the conceptualization of ACTA as the integration of two promising rehabilitation strategies: the "Nudge theory", from the cognitive domain, and the neurofeedback, from the neuroscience domain. Moreover, in ACTA we exploit the most advanced machine learning techniques to deliver customized and fully adaptive support to the elderly, while training in an ecological environment. ACTA represents the next-step beyond SENIOR, an earlier mobile-health project for cognitive training based on Nudge theory, currently ongoing in Lombardy Region. Beyond SENIOR, ACTA represents a highly-usable, accessible, low-cost, new-generation mobile-health solution to promote independent aging and effective motor-cognitive training support, while empowering the elderly in their own aging.
Analysis of large observational data sets generated by a reactive system is a common challenge in debugging system failures and determining their root cause. One of the major problems is that these observational data suffer from survivorship bias. Examples include analyzing traffic logs from networks, and simulation logs from circuit design. In such applications, users want to detect non-spurious correlations from observational data and obtain actionable insights about them. In this paper, we introduce log to Neuro-symbolic (Log2NS), a framework that combines probabilistic analysis from machine learning (ML) techniques on observational data with certainties derived from symbolic reasoning on an underlying formal model. We apply the proposed framework to network traffic debugging by employing the following steps. To detect patterns in network logs, we first generate global embedding vector representations of entities such as IP addresses, ports, and applications. Next, we represent large log flow entries as clusters that make it easier for the user to visualize and detect interesting scenarios that will be further analyzed. To generalize these patterns, Log2NS provides an ability to query from static logs and correlation engines for positive instances, as well as formal reasoning for negative and unseen instances. By combining the strengths of deep learning and symbolic methods, Log2NS provides a very powerful reasoning and debugging tool for log-based data. Empirical evaluations on a real internal data set demonstrate the capabilities of Log2NS.
Gaze estimation methods learn eye gaze from facial features. However, among rich information in the facial image, real gaze-relevant features only correspond to subtle changes in eye region, while other gaze-irrelevant features like illumination, personal appearance and even facial expression may affect the learning in an unexpected way. This is a major reason why existing methods show significant performance degradation in cross-domain/dataset evaluation. In this paper, we tackle the cross-domain problem in gaze estimation. Different from common domain adaption methods, we propose a domain generalization method to improve the cross-domain performance without touching target samples. The domain generalization is realized by gaze feature purification. We eliminate gaze-irrelevant factors such as illumination and identity to improve the cross-domain performance. We design a plug-and-play self-adversarial framework for the gaze feature purification. The framework enhances not only our baseline but also existing gaze estimation methods directly and significantly. To the best of our knowledge, we are the first to propose domain generalization methods in gaze estimation. Our method achieves not only state-of-the-art performance among typical gaze estimation methods but also competitive results among domain adaption methods. The code is released in https://github.com/yihuacheng/PureGaze.
The High Altitude Water Cherenkov (HAWC) Gamma-Ray Observatory surveys the very high energy sky in the 300 GeV to $>100$ TeV energy range. HAWC has detected two blazars above $11\sigma$, Markarian 421 (Mrk 421) and Markarian 501 (Mrk 501). The observations are comprised of data taken in the period between June 2015 and July 2018, resulting in a $\sim 1038$ days of exposure. In this work we report the time-averaged spectral analysis for both sources above 0.5 TeV. Taking into account the flux attenuation due to the extragalactic background light (EBL), the intrinsic spectrum of Mrk 421 is described by a power law with an exponential energy cut-off with index $\alpha=2.26\pm(0.12)_{stat}(_{-0.2}^{+0.17})_{sys}$ and energy cut-off $E_c=5.1\pm(1.6)_{stat}(_{-2.5}^{+1.4})_{sys}$ TeV, while the intrinsic spectrum of Mrk 501 is better described by a simple power law with index $\alpha=2.61\pm(0.11)_{stat}(_{-0.07}^{+0.01})_{sys}$. The maximum energies at which the Mrk 421 and Mrk 501 signals are detected are 9 and 12 TeV, respectively. This makes these some of the highest energy detections to date for spectra averaged over years-long timescales. Since the observation of gamma radiation from blazars provides information about the physical processes that take place in their relativistic jets, it is important to study the broad-band spectral energy distribution (SED) of these objects. To this purpose, contemporaneous data from the Large Area Telescope on board the {\em Fermi} satellite and literature data, in the radio to X-ray range, were used to build time-averaged SEDs that were modeled within a synchrotron self-Compton leptonic scenario to derive the physical parameters that describe the nature of the respective jets.
We describe the structure of finite Boolean inverse monoids and apply our results to the representation theory of finite inverse semigroups. We then generalize to semisimple Boolean inverse semigroups.
The score of a vertex $x$ in an oriented graph is defined to be its outdegree, \emph{i.e.}, the number of arcs with initial vertex $x$. The score sequence of an oriented graph is the sequence of all scores arranged in nondecreasing order. An oriented complete bipartite graph is called a bitournament. The score sequence of a bitournament consists of two nondecreasing sequences of nonnegative integers, one for each of the two partite sets. Moon has characterized the score sequences of bitournaments. This paper introduces the concept of trimming a sequence and gives a characterization of score sequences of bitournaments utilizing this concept.
In the present work, $k_T$-factorization formalism is applied to compute the exclusive dilepton production by timelike Compton scattering (TCS) in $eA$, $pA$ and $AA$ collisions. The nuclear effects are investigated considering heavy and light ions. The production cross section in terms of invariant mass and rapidity distribution of the lepton pair is shown. The analysis is done for electron-ion collisions at the Large Hadron-Electron Collider (LHeC), its high-energy upgrade (HE-LHeC) and at the Future Circular Collider (FCC) in lepton-hadron mode. Additionally, ultraperipheral heavy ion collisions at future runs of the Large Hadron Collider (LHC) and at the FCC (hadron-hadron mode) are also considered.
This paper surveys 60 English Machine Reading Comprehension datasets, with a view to providing a convenient resource for other researchers interested in this problem. We categorize the datasets according to their question and answer form and compare them across various dimensions including size, vocabulary, data source, method of creation, human performance level, and first question word. Our analysis reveals that Wikipedia is by far the most common data source and that there is a relative lack of why, when, and where questions across datasets.
The security of mobile robotic networks (MRNs) has been an active research topic in recent years. This paper demonstrates that the observable interaction process of MRNs under formation control will present increasingly severe threats. Specifically, we find that an external attack robot, who has only partial observation over MRNs while not knowing the system dynamics or access, can learn the interaction rules from observations and utilize them to replace a target robot, destroying the cooperation performance of MRNs. We call this novel attack as sneak, which endows the attacker with the intelligence of learning knowledge and is hard to be tackled by traditional defense techniques. The key insight is to separately reveal the internal interaction structure within robots and the external interaction mechanism with the environment, from the coupled state evolution influenced by the model-unknown rules and unobservable part of the MRN. To address this issue, we first provide general interaction process modeling and prove the learnability of the interaction rules. Then, with the learned rules, we design an Evaluate-Cut-Restore (ECR) attack strategy considering the partial interaction structure and geometric pattern. We also establish the sufficient conditions for a successful sneak with maximum control impacts over the MRN. Extensive simulations illustrate the feasibility and effectiveness of the proposed attack.
We present a fully Eulerian hybrid immersed-boundary/phase-field model to simulate wetting and contact line motion over any arbitrary geometry. The solid wall is described with a volume-penalisation ghost-cell immersed boundary whereas the interface between the two fluids by a diffuse-interface method. The contact line motion on the complex wall is prescribed via slip velocity in the momentum equation and static/dynamic contact angle condition for the order parameter of the Cahn-Hilliard model. This combination requires accurate computations of the normal and tangential gradients of the scalar order parameter and of the components of the velocity. However, the present algorithm requires the computation of averaging weights and other geometrical variables as a preprocessing step. Several validation tests are reported in the manuscript, together with 2D simulations of a droplet spreading over a sinusoidal wall with different contact angles and slip length and a spherical droplet spreading over a sphere, showing that the proposed algorithm is capable to deal with the three-phase contact line motion over any complex wall. The Eulerian feature of the algorithm facilitates the implementation and provides a straight-forward and potentially highly scalable parallelisation. The employed parallelisation of the underlying Navier-Stokes solver can be efficiently used for the multiphase part as well. The procedure proposed here can be directly employed to impose any types of boundary conditions (Neumann, Dirichlet and mixed) for any field variable evolving over a complex geometry, modelled with an immersed-boundary approach (for instance, modelling deformable biological membranes, red blood cells, solidification, evaporation and boiling, to name a few)
Many systems nowadays require protection against security or safety threats. A physical protection system (PPS) integrates people, procedures, and equipment to protect assets or facilities. PPSs have targeted various systems, including airports, rail transport, highways, hospitals, bridges, the electricity grid, dams, power plants, seaports, oil refineries, and water systems. Hence, PPSs are characterized by a broad set of features, from which part is common, while other features are variant and depend on the particular system to be developed. The notion of PPS has been broadly addressed in the literature, and even domain-specific PPS development methods have been proposed. However, the common and variant features are fragmented across many studies. This situation seriously impedes the identification of the required features and likewise the guidance of the systems engineering process of PPSs. To enhance the understanding and support the guidance of the development of PPS, in this paper, we provide a feature-driven survey of PPSs. The approach applies a systematic domain analysis process based on the state-of-the-art of PPSs. It presents a family feature model that defines the common and variant features and herewith the configuration space of PPSs
Multi-task learning is an important trend of machine learning in facing the era of artificial intelligence and big data. Despite a large amount of researches on learning rate estimates of various single-task machine learning algorithms, there is little parallel work for multi-task learning. We present mathematical analysis on the learning rate estimate of multi-task learning based on the theory of vector-valued reproducing kernel Hilbert spaces and matrix-valued reproducing kernels. For the typical multi-task regularization networks, an explicit learning rate dependent both on the number of sample data and the number of tasks is obtained. It reveals that the generalization ability of multi-task learning algorithms is indeed affected as the number of tasks increases.
We present multi-scale and multi-wavelength data of the Galactic HII region G25.4-0.14 (hereafter G25.4NW, distance ~5.7 kpc). The SHARC-II 350 micron continuum map displays a hub-filament configuration containing five parsec scale filaments and a central compact hub. Through the 5 GHz radio continuum map, four ionized clumps (i.e., Ia-Id) are identified toward the central hub, and are powered by massive OB-stars. The Herschel temperature map depicts the warm dust emission (i.e., Td ~23-39 K) toward the hub. High resolution Atacama Large Millimeter/submillimeter Array (ALMA) 1.3 mm continuum map (resolution ~0".82 X 0".58) reveals three cores (c1-c3; mass ~80-130 Msun) toward the ionized clumps Ia, and another one (c4; mass ~70 Msun) toward the ionized clump Ib. A compact near-infrared (NIR) emission feature (extent ~0.2 pc) is investigated toward the ionized clump Ia excited by an O8V-type star, and contains at least three embedded K-band stars. In the direction of the ionized clump Ia, the ALMA map also shows an elongated feature (extent ~0.2 pc) hosting the cores c1-c3. All these findings together illustrate the existence of a small cluster of massive stars in the central hub. Considering the detection of the hub-filament morphology and the spatial locations of the mm cores, a global non-isotropic collapse (GNIC) scenario appears to be applicable in G25.4NW, which includes the basic ingredients of the global hierarchical collapse and clump-fed accretion models. Overall, the GNIC scenario explains the birth of massive stars in G25.4NW.
In this work, we introduce a control variate approximation technique for low error approximate Deep Neural Network (DNN) accelerators. The control variate technique is used in Monte Carlo methods to achieve variance reduction. Our approach significantly decreases the induced error due to approximate multiplications in DNN inference, without requiring time-exhaustive retraining compared to state-of-the-art. Leveraging our control variate method, we use highly approximated multipliers to generate power-optimized DNN accelerators. Our experimental evaluation on six DNNs, for Cifar-10 and Cifar-100 datasets, demonstrates that, compared to the accurate design, our control variate approximation achieves same performance and 24% power reduction for a merely 0.16% accuracy loss.
We study unbendable rational curves, i.e., nonsingular rational curves in a complex manifold of dimension $n$ with normal bundles isomorphic to $\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus p} \oplus \mathcal{O}_{\mathbb{P}^1}^{\oplus (n-1-p)}$ for some nonnegative integer $p$. Well-known examples arise from algebraic geometry as general minimal rational curves of uniruled projective manifolds. After describing the relations between the differential geometric properties of the natural distributions on the deformation spaces of unbendable rational curves and the projective geometric properties of their varieties of minimal rational tangents, we concentrate on the case of $p=1$ and $n \leq 5$, which is the simplest nontrivial situation. In this case, the families of unbendable rational curves fall essentially into two classes: Goursat type or Cartan type. Those of Goursat type arise from ordinary differential equations and those of Cartan type have special features related to contact geometry. We show that the family of lines on any nonsingular cubic 4-fold is of Goursat type, whereas the family of lines on a general quartic 5-fold is of Cartan type, in the proof of which the projective geometry of varieties of minimal rational tangents plays a key role.
Deep Neural Networks (DNNs) are witnessing increased adoption in multiple domains owing to their high accuracy in solving real-world problems. However, this high accuracy has been achieved by building deeper networks, posing a fundamental challenge to the low latency inference desired by user-facing applications. Current low latency solutions trade-off on accuracy or fail to exploit the inherent temporal locality in prediction serving workloads. We observe that caching hidden layer outputs of the DNN can introduce a form of late-binding where inference requests only consume the amount of computation needed. This enables a mechanism for achieving low latencies, coupled with an ability to exploit temporal locality. However, traditional caching approaches incur high memory overheads and lookup latencies, leading us to design learned caches - caches that consist of simple ML models that are continuously updated. We present the design of GATI, an end-to-end prediction serving system that incorporates learned caches for low-latency DNN inference. Results show that GATI can reduce inference latency by up to 7.69X on realistic workloads.
India has a maternal mortality ratio of 113 and child mortality ratio of 2830 per 100,000 live births. Lack of access to preventive care information is a major contributing factor for these deaths, especially in low resource households. We partner with ARMMAN, a non-profit based in India employing a call-based information program to disseminate health-related information to pregnant women and women with recent child deliveries. We analyze call records of over 300,000 women registered in the program created by ARMMAN and try to identify women who might not engage with these call programs that are proven to result in positive health outcomes. We built machine learning based models to predict the long term engagement pattern from call logs and beneficiaries' demographic information, and discuss the applicability of this method in the real world through a pilot validation. Through a pilot service quality improvement study, we show that using our model's predictions to make interventions boosts engagement metrics by 61.37%. We then formulate the intervention planning problem as restless multi-armed bandits (RMABs), and present preliminary results using this approach.
The evolution of deformation from plasticity to localization to damage is investigated in ferritic-pearlitic steel through nanometer-resolution microstructure-correlated SEM-DIC (u-DIC) strain mapping, enabled through highly accurate microstructure-to-strain alignment. We reveal the key plasticity mechanisms in ferrite and pearlite as well as their evolution into localization and damage and their relation to the microstructural arrangement. Notably, two contrasting mechanisms were identified that control whether damage initiation in pearlite occurs and, through connection of localization hotspots in ferrite grains, potentially results in macroscale fracture: (i) cracking of pearlite bridges with relatively clean lamellar structure by brittle fracture of cementite lamellae due to build-up of strain concentrations in nearby ferrite, versus (ii) large plasticity without damage in pearlite bridges with a more "open", chaotic pearlite morphology, which enables plastic percolation paths in the interlamellar ferrite channels. Based on these insights, recommendations for damage resistant ferritic-pearlitic steels are proposed.
This is a survey of recent results on central and non-central limit theorems for quadratic functionals of stationary processes. The underlying processes are Gaussian, linear or L\'evy-driven linear processes with memory, and are defined either in discrete or continuous time. We focus on limit theorems for Toeplitz and tapered Toeplitz type quadratic functionals of stationary processes with applications in parametric and nonparametric statistical estimation theory. We discuss questions concerning Toeplitz matrices and operators, Fej\'er-type singular integrals, and L\'evy-It\^o-type and Stratonovich-type multiple stochastic integrals. These are the main tools for obtaining limit theorems.
Although deep convolution neural networks (DCNN) have achieved excellent performance in human pose estimation, these networks often have a large number of parameters and computations, leading to the slow inference speed. For this issue, an effective solution is knowledge distillation, which transfers knowledge from a large pre-trained network (teacher) to a small network (student). However, there are some defects in the existing approaches: (I) Only a single teacher is adopted, neglecting the potential that a student can learn from multiple teachers. (II) The human segmentation mask can be regarded as additional prior information to restrict the location of keypoints, which is never utilized. (III) A student with a small number of parameters cannot fully imitate heatmaps provided by datasets and teachers. (IV) There exists noise in heatmaps generated by teachers, which causes model degradation. To overcome these defects, we propose an orderly dual-teacher knowledge distillation (ODKD) framework, which consists of two teachers with different capabilities. Specifically, the weaker one (primary teacher, PT) is used to teach keypoints information, the stronger one (senior teacher, ST) is utilized to transfer segmentation and keypoints information by adding the human segmentation mask. Taking dual-teacher together, an orderly learning strategy is proposed to promote knowledge absorbability. Moreover, we employ a binarization operation which further improves the learning ability of the student and reduces noise in heatmaps. Experimental results on COCO and OCHuman keypoints datasets show that our proposed ODKD can improve the performance of different lightweight models by a large margin, and HRNet-W16 equipped with ODKD achieves state-of-the-art performance for lightweight human pose estimation.
RGB-D salient object detection (SOD) is usually formulated as a problem of classification or regression over two modalities, i.e., RGB and depth. Hence, effective RGBD feature modeling and multi-modal feature fusion both play a vital role in RGB-D SOD. In this paper, we propose a depth-sensitive RGB feature modeling scheme using the depth-wise geometric prior of salient objects. In principle, the feature modeling scheme is carried out in a depth-sensitive attention module, which leads to the RGB feature enhancement as well as the background distraction reduction by capturing the depth geometry prior. Moreover, to perform effective multi-modal feature fusion, we further present an automatic architecture search approach for RGB-D SOD, which does well in finding out a feasible architecture from our specially designed multi-modal multi-scale search space. Extensive experiments on seven standard benchmarks demonstrate the effectiveness of the proposed approach against the state-of-the-art.
This paper attempts to study the optimal stopping time for semi-Markov processes (SMPs) under the discount optimization criteria with unbounded cost rates. In our work, we introduce an explicit construction of the equivalent semi-Markov decision processes (SMDPs). The equivalence is embodied in the value functions of SMPs and SMDPs, that is, every stopping time of SMPs can induce a policy of SMDPs such that the value functions are equal, and vice versa. The existence of the optimal stopping time of SMPs is proved by this equivalence relation. Next, we give the optimality equation of the value function and develop an effective iterative algorithm for computing it. Moreover, we show that the optimal and {\epsilon}-optimal stopping time can be characterized by the hitting time of the special sets. Finally, to illustrate the validity of our results, an example of a maintenance system is presented in the end.
Autoregressive models are widely used for tasks such as image and audio generation. The sampling process of these models, however, does not allow interruptions and cannot adapt to real-time computational resources. This challenge impedes the deployment of powerful autoregressive models, which involve a slow sampling process that is sequential in nature and typically scales linearly with respect to the data dimension. To address this difficulty, we propose a new family of autoregressive models that enables anytime sampling. Inspired by Principal Component Analysis, we learn a structured representation space where dimensions are ordered based on their importance with respect to reconstruction. Using an autoregressive model in this latent space, we trade off sample quality for computational efficiency by truncating the generation process before decoding into the original data space. Experimentally, we demonstrate in several image and audio generation tasks that sample quality degrades gracefully as we reduce the computational budget for sampling. The approach suffers almost no loss in sample quality (measured by FID) using only 60\% to 80\% of all latent dimensions for image data. Code is available at https://github.com/Newbeeer/Anytime-Auto-Regressive-Model .
A novel spin orientation mechanism - dynamic electron spin polarization has been recently suggested in Phys. Rev. Lett. $\mathbf{125}$, 156801 (2020). It takes place for unpolarized optical excitation in weak magnetic fields of the order of a few millitesla. In this paper we demonstrate experimentally and theoretically that the dynamic electron spin polarization degree changes sign as a function of time, strength of the applied magnetic field and its direction. The studies are performed on indirect band-gap (In,Al)As/AlAs quantum dots and their results are explained in the framework of a theoretical model developed for our experimental setting.
We investigate the computability of algebraic closure and definable closure with respect to a collection of formulas. We show that for a computable collection of formulas of quantifier rank at most $n$, in any given computable structure, both algebraic and definable closure with respect to that collection are $\Sigma^0_{n+2}$ sets. We further show that these bounds are tight.
We prove the existence of an eddy heat diffusion coefficient coming from an idealized model of turbulent fluid. A difficulty lies in the presence of a boundary, with also turbulent mixing and the eddy diffusion coefficient going to zero at the boundary. Nevertheless enhanced diffusion takes place.
A simplification strategy for Segmented Mirror Splitters (SMS) used as beam combiners is presented. These devices are useful for compact beam division and combination of linear and 2-D arrays. However, the standard design requires unique thin-film coating sections for each input beam and thus potential for scaling to high beam-counts is limited due to manufacturing complexity. Taking advantage of the relative insensitivity of the beam combination process to amplitude variations, numerical techniques are used to optimize highly-simplified designs with only one, two or three unique coatings. It is demonstrated that with correctly chosen coating reflectivities, the simplified optics are capable of high combination efficiency for several tens of beams. The performance of these optics as beamsplitters in multicore fiber amplifier systems is analyzed, and inhomogeneous power distribution of the simplified designs is noted as a potential source of combining loss in such systems. These simplified designs may facilitate further scaling of filled-aperture coherently combined systems.
In this paper, we provide causal evidence on abortions and risky health behaviors as determinants of mental health development among young women. Using administrative in- and outpatient records from Sweden, we apply a novel grouped fixed-effects estimator proposed by Bonhomme and Manresa (2015) to allow for time-varying unobserved heterogeneity. We show that the positive association obtained from standard estimators shrinks to zero once we control for grouped time-varying unobserved heterogeneity. We estimate the group-specific profiles of unobserved heterogeneity, which reflect differences in unobserved risk to be diagnosed with a mental health condition. We then analyze mental health development and risky health behaviors other than unwanted pregnancies across groups. Our results suggest that these are determined by the same type of unobserved heterogeneity, which we attribute to the same unobserved process of decision-making. We develop and estimate a theoretical model of risky choices and mental health, in which mental health disparity across groups is generated by different degrees of self-control problems. Our findings imply that mental health concerns cannot be used to justify restrictive abortion policies. Moreover, potential self-control problems should be targeted as early as possible to combat future mental health consequences.
Recent years have witnessed a renewed interest in Boolean function in explaining binary classifiers in the field of explainable AI (XAI). The standard approach of Boolean function is propositional logic. We study a family of classifier models, axiomatize it and show completeness of our axiomatics. Moreover, we prove that satisfiability checking for our modal language relative to such a class of models is NP-complete. We leverage the language to formalize counterfactual conditional as well as a variety of notions of explanation including abductive, contrastive and counterfactual explanations, and biases. Finally, we present two extensions of our language: a dynamic extension by the notion of assignment enabling classifier change and an epistemic extension in which the classifier's uncertainty about the actual input can be represented.
We discuss a probe of the contribution of wind-related shocks to the radio emission in otherwise radio-quiet quasars. Given 1) the non-linear correlation between UV and X-ray luminosity in quasars, 2) that such correlation leads to higher likelihood of radiation-line-driven winds in more luminous quasars, and 3) that luminous quasars are more abundant at high redshift, deep radio observations of high-redshift quasars are needed to probe potential contributions from accretion disk winds. We target a sample of 50 $z\simeq 1.65$ color-selected quasars that span the range of expected accretion disk wind properties as traced by broad CIV emission. 3-GHz observations with the Very Large Array to an rms of $\approx10\mu$Jy beam$^{-1}$ probe to star formation rates of $\approx400\,M_{\rm Sun}\,{\rm yr}^{-1}$, leading to 22 detections. Supplementing these pointed observations are survey data of 388 sources from the LOFAR Two-metre Sky Survey Data Release 1 that reach comparable depth (for a typical radio spectral index), where 123 sources are detected. These combined observations reveal a radio detection fraction that is a non-linear function of \civ\ emission-line properties and suggest that the data may require multiple origins of radio emission in radio-quiet quasars. We find evidence for radio emission from weak jets or coronae in radio-quiet quasars with low Eddingtion ratios, with either (or both) star formation and accretion disk winds playing an important role in optically luminous quasars and correlated with increasing Eddington ratio. Additional pointed radio observations are needed to fully establish the nature of radio emission in radio-quiet quasars.
Like adiabatic time-dependent density-functional theory (TD-DFT), the Bethe-Salpeter equation (BSE) formalism of many-body perturbation theory, in its static approximation, is "blind" to double (and higher) excitations, which are ubiquitous, for example, in conjugated molecules like polyenes. Here, we apply the spin-flip \textit{ansatz} (which considers the lowest triplet state as the reference configuration instead of the singlet ground state) to the BSE formalism in order to access, in particular, double excitations. The present scheme is based on a spin-unrestricted version of the $GW$ approximation employed to compute the charged excitations and screened Coulomb potential required for the BSE calculations. Dynamical corrections to the static BSE optical excitations are taken into account via an unrestricted generalization of our recently developed (renormalized) perturbative treatment. The performance of the present spin-flip BSE formalism is illustrated by computing excited-state energies of the beryllium atom, the hydrogen molecule at various bond lengths, and cyclobutadiene in its rectangular and square-planar geometries.
Question answering from semi-structured tables can be seen as a semantic parsing task and is significant and practical for pushing the boundary of natural language understanding. Existing research mainly focuses on understanding contents from unstructured evidence, e.g., news, natural language sentences, and documents. The task of verification from structured evidence, such as tables, charts, and databases, is still less explored. This paper describes sattiy team's system in SemEval-2021 task 9: Statement Verification and Evidence Finding with Tables (SEM-TAB-FACT). This competition aims to verify statements and to find evidence from tables for scientific articles and to promote the proper interpretation of the surrounding article. In this paper, we exploited ensemble models of pre-trained language models over tables, TaPas and TaBERT, for Task A and adjust the result based on some rules extracted for Task B. Finally, in the leaderboard, we attain the F1 scores of 0.8496 and 0.7732 in Task A for the 2-way and 3-way evaluation, respectively, and the F1 score of 0.4856 in Task B.
A generalized method of alternating resolvents was introduced by Boikanyo and Moro{\c s}anu as a way to approximate common zeros of two maximal monotone operators. In this paper we analyse the strong convergence of this algorithm under two different sets of conditions. As a consequence we obtain effective rates of metastability (in the sense of Terence Tao) and quasi-rates of asymptotic regularity. Furthermore, we bypass the need for sequential weak compactness in the original proofs. Our quantitative results are obtained using proof-theoretical techniques in the context of the proof mining program.
At the latest since the advent of the Internet, disinformation and conspiracy theories have become ubiquitous. Recent examples like QAnon and Pizzagate prove that false information can lead to real violence. In this motivation statement for the Workshop on Human Aspects of Misinformation at CHI 2021, I explain my research agenda focused on 1. why people believe in disinformation, 2. how people can be best supported in recognizing disinformation, and 3. what the potentials and risks of different tools designed to fight disinformation are.
We investigate the asymptotic risk of a general class of overparameterized likelihood models, including deep models. The recent empirical success of large-scale models has motivated several theoretical studies to investigate a scenario wherein both the number of samples, $n$, and parameters, $p$, diverge to infinity and derive an asymptotic risk at the limit. However, these theorems are only valid for linear-in-feature models, such as generalized linear regression, kernel regression, and shallow neural networks. Hence, it is difficult to investigate a wider class of nonlinear models, including deep neural networks with three or more layers. In this study, we consider a likelihood maximization problem without the model constraints and analyze the upper bound of an asymptotic risk of an estimator with penalization. Technically, we combine a property of the Fisher information matrix with an extended Marchenko-Pastur law and associate the combination with empirical process techniques. The derived bound is general, as it describes both the double descent and the regularized risk curves, depending on the penalization. Our results are valid without the linear-in-feature constraints on models and allow us to derive the general spectral distributions of a Fisher information matrix from the likelihood. We demonstrate that several explicit models, such as parallel deep neural networks, ensemble learning, and residual networks, are in agreement with our theory. This result indicates that even large and deep models have a small asymptotic risk if they exhibit a specific structure, such as divisibility. To verify this finding, we conduct a real-data experiment with parallel deep neural networks. Our results expand the applicability of the asymptotic risk analysis, and may also contribute to the understanding and application of deep learning.
We study Lagrangian systems with a finite number of degrees of freedom that are non-local in time. We obtain an extension of Noether theorem and Noether identities to this kind of Lagrangians. A Hamiltonian formalism is then set up for this systems. $n$-order local Lagrangians can be treated as a particular case and the standard results for them are recovered. The method is then applied to several other cases, namely two examples of non-local oscillators and the p-adic particle.
Quantum algorithms for computing classical nonlinear maps are widely known for toy problems but might not suit potential applications to realistic physics simulations. Here, we propose how to compute a general differentiable invertible nonlinear map on a quantum computer using only linear unitary operations. The price of this universality is that the original map is represented adequately only on a finite number of iterations. More iterations produce spurious echos, which are unavoidable in any finite unitary emulation of generic non-conservative dynamics. Our work is intended as the first survey of these issues and possible ways to overcome them in the future. We propose how to monitor spurious echos via auxiliary measurements, and we illustrate our results with numerical simulations.
Assessing the exploitability of software vulnerabilities at the time of disclosure is difficult and error-prone, as features extracted via technical analysis by existing metrics are poor predictors for exploit development. Moreover, exploitability assessments suffer from a class bias because "not exploitable" labels could be inaccurate. To overcome these challenges, we propose a new metric, called Expected Exploitability (EE), which reflects, over time, the likelihood that functional exploits will be developed. Key to our solution is a time-varying view of exploitability, a departure from existing metrics, which allows us to learn EE using data-driven techniques from artifacts published after disclosure, such as technical write-ups, proof-of-concept exploits, and social media discussions. Our analysis reveals that prior features proposed for related exploit prediction tasks are not always beneficial for predicting functional exploits, and we design novel feature sets to capitalize on previously under-utilized artifacts. This view also allows us to investigate the effect of the label biases on the classifiers. We characterize the noise-generating process for exploit prediction, showing that our problem is subject to class- and feature-dependent label noise, considered the most challenging type. By leveraging domain-specific observations, we then develop techniques to incorporate noise robustness into learning EE. On a dataset of 103,137 vulnerabilities, we show that EE increases precision from 49\% to 86\% over existing metrics, including two state-of-the-art exploit classifiers, while the performance of our metric also improving over time. EE scores capture exploitation imminence, by distinguishing exploits which are going to be developed in the near future.
We introduce the "inverse bandit" problem of estimating the rewards of a multi-armed bandit instance from observing the learning process of a low-regret demonstrator. Existing approaches to the related problem of inverse reinforcement learning assume the execution of an optimal policy, and thereby suffer from an identifiability issue. In contrast, our paradigm leverages the demonstrator's behavior en route to optimality, and in particular, the exploration phase, to obtain consistent reward estimates. We develop simple and efficient reward estimation procedures for demonstrations within a class of upper-confidence-based algorithms, showing that reward estimation gets progressively easier as the regret of the algorithm increases. We match these upper bounds with information-theoretic lower bounds that apply to any demonstrator algorithm, thereby characterizing the optimal tradeoff between exploration and reward estimation. Extensive empirical evaluations on both synthetic data and simulated experimental design data from the natural sciences corroborate our theoretical results.
Interevent times in temporal contact data from humans and animals typically obey heavy-tailed distributions, and this property impacts contagion and other dynamical processes on networks. We theoretically show that distributions of interevent times heavier-tailed than exponential distributions are a consequence of the most basic metapopulation model used in epidemiology and ecology, in which individuals move from a patch to another according to the simple random walk. Our results hold true irrespectively of the network structure and also for more realistic mobility rules such as high-order random walks and the recurrent mobility patterns used for modeling human dynamics.
We present observations of a region of the Galactic plane taken during the Early Science Program of the Australian Square Kilometre Array Pathfinder (ASKAP). In this context, we observed the SCORPIO field at 912 MHz with an uncompleted array consisting of 15 commissioned antennas. The resulting map covers a square region of ~40 deg^2, centred on (l, b)=(343.5{\deg}, 0.75{\deg}), with a synthesized beam of 24"x21" and a background rms noise of 150-200 {\mu}Jy/beam, increasing to 500-600 {\mu}Jy/beam close to the Galactic plane. A total of 3963 radio sources were detected and characterized in the field using the CAESAR source finder. We obtained differential source counts in agreement with previously published data after correction for source extraction and characterization uncertainties, estimated from simulated data. The ASKAP positional and flux density scale accuracy were also investigated through comparison with previous surveys (MGPS, NVSS) and additional observations of the SCORPIO field, carried out with ATCA at 2.1 GHz and 10" spatial resolution. These allowed us to obtain a measurement of the spectral index for a subset of the catalogued sources and an estimated fraction of (at least) 8% of resolved sources in the reported catalogue. We cross-matched our catalogued sources with different astronomical databases to search for possible counterparts, finding ~150 associations to known Galactic objects. Finally, we explored a multiparametric approach for classifying previously unreported Galactic sources based on their radio-infrared colors.
In this paper, a general class of mixture of some densities is proposed. The proposed class contains some of classical and weighted distributions as special cases. Formulas for each of cumulative distribution function, reliability function, hazard rate function, rth raw moments function, characteristic function, stress-strength reliability and Tsallis entropy of order are derived.
Attribution methods provide an insight into the decision-making process of machine learning models, especially deep neural networks, by assigning contribution scores to each individual feature. However, the attribution problem has not been well-defined, which lacks a unified guideline to the contribution assignment process. Furthermore, existing attribution methods often built upon various empirical intuitions and heuristics. There still lacks a general theoretical framework that not only can offer a good description of the attribution problem, but also can be applied to unifying and revisiting existing attribution methods. To bridge the gap, in this paper, we propose a Taylor attribution framework, which models the attribution problem as how to decide individual payoffs in a coalition. Then, we reformulate fourteen mainstream attribution methods into the Taylor framework and analyze these attribution methods in terms of rationale, fidelity, and limitation in the framework. Moreover, we establish three principles for a good attribution in the Taylor attribution framework, i.e., low approximation error, correct Taylor contribution assignment, and unbiased baseline selection. Finally, we empirically validate the Taylor reformulations and reveal a positive correlation between the attribution performance and the number of principles followed by the attribution method via benchmarking on real-world datasets.
We develop some basic concepts in the theory of higher categories internal to an arbitrary $\infty$-topos. We define internal left and right fibrations and prove a version of the Grothendieck construction and of Yoneda's lemma for internal categories.
Much of interesting complex biological behaviour arises from collective properties. Important information about collective behaviour lies in the time and space structure of fluctuations around average properties, and two-point correlation functions are a fundamental tool to study these fluctuations. We give a self-contained presentation of definitions and techniques for computation of correlation functions aimed at providing students and researchers outside the field of statistical physics a practical guide to calculating correlation functions from experimental and simulation data. We discuss some properties of correlations in critical systems, and the effect of finite system size, which is particularly relevant for most biological experimental systems. Finally we apply these to the case of the dynamical transition in a simple neuronal model,
In this paper, we propose a time-dependent Susceptible-Exposed-Infectious-Recovered-Died (SEIRD) reaction-diffusion system for the COVID-19 pandemic and we deal with its derivation from a kinetic model. The derivation is obtained by mathematical description delivered at the micro-scale of individuals. Our approach is based on the micro-macro decomposition which leads to an equivalent formulation of the kinetic model which couples the microscopic equations with the macroscopic equations. We develop a numerical asymptotic preservation scheme to solve the kinetic model. The proposed approach is validated by various numerical tests where particular attention is paid to the Moroccan situation against the actual pandemic.
Luminescent multifunctional nanomaterials are important because of their potential impact on the development of key technologies such as smart luminescent sensors and solid-state lightings. To be technologically viable, the luminescent material needs to fulfil a number of requirements such as facile and cost-effective fabrication, a high quantum yield, structural robustness, and long-term material stability. To achieve these requirements, an eco-friendly and scalable synthesis of a highly photoluminescent, multistimuli responsive and electroluminescent silver-based metal-organic framework (Ag-MOF), termed "OX-2" was developed. Its exceptional photophysical and mechanically resilient properties that can be reversibly switched by temperature and pressure make this material stood out over other competing luminescent materials. The potential use of OX-2 MOF as a good electroluminescent material was tested by constructing a proof-of-concept MOF-LED (light emitting diode) device, further contributing to the rare examples of electroluminescent MOFs. The results reveal the huge potential for exploiting the Ag MOF as a multitasking platform to engineer innovative photonic technologies.
I find that several models for information sharing in social networks can be interpreted as age-dependent multi-type branching processes, and build them independently following Sewastjanow. This allows to characterize criticality in (real and random) social networks. For random networks, I develop a moment-closure method that handles the high-dimensionality of these models: By modifying the timing of sharing with followers, all users can be represented by a single representative, while leaving the total progeny unchanged. Thus I compute the exact popularity distribution, revealing a viral character of critical models expressed by fat tails of order minus three half.
We address the task of domain generalization, where the goal is to train a predictive model such that it is able to generalize to a new, previously unseen domain. We choose a hierarchical generative approach within the framework of variational autoencoders and propose a domain-unsupervised algorithm that is able to generalize to new domains without domain supervision. We show that our method is able to learn representations that disentangle domain-specific information from class-label specific information even in complex settings where domain structure is not observed during training. Our interpretable method outperforms previously proposed generative algorithms for domain generalization as well as other non-generative state-of-the-art approaches in several hierarchical domain settings including sequential overlapped near continuous domain shift. It also achieves competitive performance on the standard domain generalization benchmark dataset PACS compared to state-of-the-art approaches which rely on observing domain-specific information during training, as well as another domain unsupervised method. Additionally, we proposed model selection purely based on Evidence Lower Bound (ELBO) and also proposed weak domain supervision where implicit domain information can be added into the algorithm.
Generative adversarial networks (GAN) is a framework for generating fake data based on given reals but is unstable in the optimization. In order to stabilize GANs, the noise enlarges the overlap of the real and fake distributions at the cost of significant variance. The data smoothing may reduce the dimensionality of data but suppresses the capability of GANs to learn high-frequency information. Based on these observations, we propose a data representation for GANs, called noisy scale-space, that recursively applies the smoothing with noise to data in order to preserve the data variance while replacing high-frequency information by random data, leading to a coarse-to-fine training of GANs. We also present a synthetic data-set using the Hadamard bases that enables us to visualize the true distribution of data. We experiment with a DCGAN with the noise scale-space (NSS-GAN) using major data-sets in which NSS-GAN overtook state-of-the-arts in most cases independent of the image content.
This paper discusses the design, implementation and field trials of WiMesh - a resilient Wireless Mesh Network (WMN) based disaster communication system purpose-built for underdeveloped and rural parts of the world. Mesh networking is a mature area, and the focus of this paper is not on proposing novel models, protocols or other mesh solutions. Instead, the paper focuses on the identification of important design considerations and justifications for several design trade offs in the context of mesh networking for disaster communication in developing countries with very limited resources. These trade-offs are discussed in the context of key desirable traits including security, low cost, low power, size, availability, customization, portability, ease of installation and deployment, and coverage area among others. We discuss at length the design, implementation, and field trial results of the WiMesh system which enables users spread over large geographical regions, to communicate with each other despite the lack of cellular coverage, power, and other communication infrastructure by leveraging multi-hop mesh networking and Wi-Fi equipped handheld devices. Lessons learned along with real-world results are shared for WiMesh deployment in a remote rural mountainous village of Pakistan, and the source code is shared with the research community.
Learning by interaction is the key to skill acquisition for most living organisms, which is formally called Reinforcement Learning (RL). RL is efficient in finding optimal policies for endowing complex systems with sophisticated behavior. All paradigms of RL utilize a system model for finding the optimal policy. Modeling dynamics can be done by formulating a mathematical model or system identification. Dynamic models are usually exposed to aleatoric and epistemic uncertainties that can divert the model from the one acquired and cause the RL algorithm to exhibit erroneous behavior. Accordingly, the RL process sensitive to operating conditions and changes in model parameters and lose its generality. To address these problems, Intensive system identification for modeling purposes is needed for each system even if the model dynamics structure is the same, as the slight deviation in the model parameters can render the model useless in RL. The existence of an oracle that can adaptively predict the rest of the trajectory regardless of the uncertainties can help resolve the issue. The target of this work is to present a framework for facilitating the system identification of different instances of the same dynamics class by learning a probability distribution of the dynamics conditioned on observed data with variational inference and show its reliability in robustly solving different instances of control problems with the same model in model-based RL with maximum sample efficiency.
Distributed quantum metrology can enhance the sensitivity for sensing spatially distributed parameters beyond the classical limits. Here we demonstrate distributed quantum phase estimation with discrete variables to achieve Heisenberg limit phase measurements. Based on parallel entanglement in modes and particles, we demonstrate distributed quantum sensing for both individual phase shifts and an averaged phase shift, with an error reduction up to 1.4 dB and 2.7 dB below the shot-noise limit. Furthermore, we demonstrate a combined strategy with parallel mode entanglement and multiple passes of the phase shifter in each mode. In particular, our experiment uses six entangled photons with each photon passing the phase shifter up to six times, and achieves a total number of photon passes N=21 at an error reduction up to 4.7 dB below the shot-noise limit. Our research provides a faithful verification of the benefit of entanglement and coherence for distributed quantum sensing in general quantum networks.
We consider a simple model of a stochastic heat engine, which consists of a single Brownian particle moving in a one-dimensional periodically breathing harmonic potential. Overdamped limit is assumed. Expressions of second moments (variances and covariances ) of heat and work are obtained in the form of integrals, whose integrands contain functions satisfying certain differential equations. The results in the quasi-static limit are simple functions of temperatures of hot and cold thermal baths. The coefficient of variation of the work is suggested to give an approximate probability for the work to exceeds a certain threshold. During derivation, we get the expression of the cumulant-generating function.
We apply general moment identities for Poisson stochastic integrals with random integrands to the computation of the moments of Markovian growth-collapse processes. This extends existing formulas for mean and variance available in the literature to closed form moments expressions of all orders. In comparison with other methods based on differential equations, our approach yields polynomial expressions in the time parameter. We also treat the case of the associated embedded chain.
Studies on stratospheric ozone have attracted much attention due to its serious impacts on climate changes and its important role as a tracer of Earth's global circulation. Tropospheric ozone as a main atmospheric pollutant damages human health as well as the growth of vegetation. Yet there is still a lack of a theoretical framework to fully describe the variation of ozone. To understand ozone's spatiotemporal variance, we introduce the eigen microstate method to analyze the global ozone mass mixing ratio (OMMR) between 1979-01-01 and 2020-06-30 at 37 pressure layers. We find that eigen microstates at different geopotential heights can capture different climate phenomena and modes. Without deseasonalization, the first eigen microstates capture the seasonal effect and reveal that the phase of the intra-annual cycle moves with the geopotential heights. After deseasonalization, by contrast, the collective patterns from the overall trend, ENSO, QBO, and tropopause pressure are identified by the first few significant eigen microstates. The theoretical framework proposed here can also be applied to other complex Earth systems.
Effective and causal observable functions for low-order lifting linearization of nonlinear controlled systems are learned from data by using neural networks. While Koopman operator theory allows us to represent a nonlinear system as a linear system in an infinite-dimensional space of observables, exact linearization is guaranteed only for autonomous systems with no input, and finding effective observable functions for approximation with a low-order linear system remains an open question. Dual-Faceted Linearization uses a set of effective observables for low-order lifting linearization, but the method requires knowledge of the physical structure of the nonlinear system. Here, a data-driven method is presented for generating a set of nonlinear observable functions that can accurately approximate a nonlinear control system to a low-order linear control system. A caveat in using data of measured variables as observables is that the measured variables may contain input to the system, which incurs a causality contradiction when lifting the system, i.e. taking derivatives of the observables. The current work presents a method for eliminating such anti-causal components of the observables and lifting the system using only causal observables. The method is applied to excavation automation, a complex nonlinear dynamical system, to obtain a low-order lifted linear model for control design.
With the widespread use and adoption of mobile platforms like Android a new software quality concern has emerged -- energy consumption. However, developing energy-efficient software and applications requires knowledge and likewise proper tooling to support mobile developers. To this aim, we present an approach to examine the energy evolution of software revisions based on their API interactions. The approach stems from the assumption that the utilization of an API has direct implications on the energy being consumed during runtime. Based on an empirical evaluation, we show initial results that API interactions serve as a flexible, lightweight, and effective way to compare software revisions regarding their energy evolution. Given our initial results we envision that in future using our approach mobile developers will be able to gain insights on the energy implications of changes in source code in the course of the software development life-cycle.
In this work we develop a novel characterization of marginal causal effect and causal bias in the continuous treatment setting. We show they can be expressed as an expectation with respect to a conditional probability distribution, which can be estimated via standard statistical and probabilistic methods. All terms in the expectations can be computed via automatic differentiation, also for highly non-linear models. We further develop a new complete criterion for identifiability of causal effects via covariate adjustment, showing the bias equals zero if the criterion is met. We study the effectiveness of our framework in three different scenarios: linear models under confounding, overcontrol and endogenous selection bias; a non-linear model where full identifiability cannot be achieved because of missing data; a simulated medical study of statins and atherosclerotic cardiovascular disease.
In weather disasters, first responders access dedicated communication channels different from civilian commercial channels to facilitate rescues. However, rescues in recent disasters have increasingly involved civilian and volunteer forces, requiring civilian channels not to be overloaded with traffic. We explore seven enhancements to the wording of Wireless Emergency Alerts (WEAs) and their effectiveness in getting smartphone users to comply, including reducing frivolous mobile data consumption during critical weather disasters. We conducted a between-subjects survey (N=898), in which participants were either assigned no alert (control) or an alert framed as Basic Information, Altruism, Multimedia, Negative Feedback, Positive Feedback, Reward, or Punishment. We find that Basic Information alerts resulted in the largest reduction of multimedia and video services usage; we also find that Punishment alerts have the lowest absolute compliance. This work has implications for creating more effective WEAs and providing a better understanding of how wording can affect emergency alert compliance.
Malaria is an infectious disease with an immense global health burden. Plasmodium vivax is the most geographically widespread species of malaria. Relapsing infections, caused by the activation of liver-stage parasites known as hypnozoites, are a critical feature of the epidemiology of Plasmodium vivax. Hypnozoites remain dormant in the liver for weeks or months after inoculation, but cause relapsing infections upon activation. Here, we introduce a dynamic probability model of the activation-clearance process governing both potential relapses and the size of the hypnozoite reservoir. We begin by modelling activation-clearance dynamics for a single hypnozoite using a continuous-time Markov chain. We then extend our analysis to consider activation-clearance dynamics for a single mosquito bite, which can simultaneously establish multiple hypnozoites, under the assumption of independent hypnozoite behaviour. We derive analytic expressions for the time to first relapse and the time to hypnozoite clearance for mosquito bites establishing variable numbers of hypnozoites, both of which are quantities of epidemiological significance. Our results extend those in the literature, which were limited due to an assumption of non-independence. Our within-host model can be embedded readily in multi-scale models and epidemiological frameworks, with analytic solutions increasing the tractability of statistical inference and analysis. Our work therefore provides a foundation for further work on immune development and epidemiological-scale analysis, both of which are important for achieving the goal of malaria elimination.
Academic neural models for coreference resolution (coref) are typically trained on a single dataset, OntoNotes, and model improvements are benchmarked on that same dataset. However, real-world applications of coref depend on the annotation guidelines and the domain of the target dataset, which often differ from those of OntoNotes. We aim to quantify transferability of coref models based on the number of annotated documents available in the target dataset. We examine eleven target datasets and find that continued training is consistently effective and especially beneficial when there are few target documents. We establish new benchmarks across several datasets, including state-of-the-art results on PreCo.
In many randomized clinical trials of therapeutics for COVID-19, the primary outcome is an ordinal categorical variable, and interest focuses on the odds ratio (active agent vs. control) under the assumption of a proportional odds model. Although at the final analysis the outcome will be determined for all subjects, at an interim analysis, the status of some participants may not yet be determined, e.g., because ascertainment of the outcome may not be possible until some pre-specified follow-up time. Accordingly, the outcome from these subjects can be viewed as censored. A valid interim analysis can be based on data only from those subjects with full follow up; however, this approach is inefficient, as it does not exploit additional information that may be available on those for whom the outcome is not yet available at the time of the interim analysis. Appealing to the theory of semiparametrics, we propose an estimator for the odds ratio in a proportional odds model with censored, time-lagged categorical outcome that incorporates additional baseline and time-dependent covariate information and demonstrate that it can result in considerable gains in efficiency relative to simpler approaches. A byproduct of the approach is a covariate-adjusted estimator for the odds ratio based on the full data that would be available at a final analysis.
Scoring rules measure the deviation between a probabilistic forecast and reality. Strictly proper scoring rules have the property that for any forecast, the mathematical expectation of the score of a forecast p by the lights of p is strictly better than the mathematical expectation of any other forecast q by the lights of p. Probabilistic forecasts need not satisfy the axioms of the probability calculus, but Predd, et al. (2009) have shown that given a finite sample space and any strictly proper additive and continuous scoring rule, the score for any forecast that does not satisfy the axioms of probability is strictly dominated by the score for some probabilistically consistent forecast. Recently, this result has been extended to non-additive continuous scoring rules. In this paper, a condition weaker than continuity is given that suffices for the result, and the condition is proved to be optimal.
We prove that, if $\mathcal{GP}$ is the class of all Gorenstein projective modules over a ring $R$, then $\mathfrak{GP}=(\mathcal{GP},\mathcal{GP}^\perp)$ is a cotorsion pair. Moreover, $\mathfrak{GP}$ is complete when all projective modules are $\lambda$-pure-injective for some infinite regular cardinal $\lambda$ (in particular, if $R$ is right $\Sigma$-pure-injective). We obtain these results, on the one hand, studying the class of totally acyclic complexes over $R$. We prove that, when $R$ is $\Sigma$-pure-injective, this class is deconstructible and forms a coreflective subcategory of the homotopy category of the projective modules. On the other hand, we use results about $\lambda$-pure-injective modules for infinite regular cardinals $\lambda$. Finally, under different set-theoretical hypotheses, we show that for an arbitrary ring $R$, the following hold: (1) There exists an infinite regular cardinal number $\lambda$ such that every projective module is $\lambda$-pure-injective (and $\mathfrak{GP}$ is complete). (2) $R$ is right pure-semisimple if and only if there exists a regular uncountable $\lambda$ such that $\mathrm{Mod}$-$R$ has enough $\lambda$-pure-injective objects.
Intensity mapping of the 21cm signal of neutral hydrogen will yield exciting insights into the Epoch of Reionisation and the nature of the first galaxies. However, the large amount of data that will be generated by the next generation of radio telescopes, such as the Square Kilometre Array (SKA), as well as the numerous observational obstacles to overcome, require analysis techniques tuned to extract the reionisation history and morphology. In this context, we introduce a one-point statistic, to which we refer as the local variance, $\sigma_\mathrm{loc}$, that describes the distribution of the mean differential 21cm brightness temperatures measured in two-dimensional maps along the frequency direction of a light-cone. The local variance takes advantage of what is usually considered an observational bias, the sample variance. We find the redshift-evolution of the local variance to not only probe the reionisation history of the observed patches of the sky, but also trace the ionisation morphology. This estimator provides a promising tool to constrain the midpoint of reionisation as well as gaining insight into the ionising properties of early galaxies.
Temperature fluctuations of a finite system follows the Landau bound $\delta T^2 = T^2/C(T)$ where $C(T)$ is the heat capacity of the system. In turn, the same bound sets a limit to the precision of temperature estimation when the system itself is used as a thermometer. In this paper, we employ graph theory and the concept of Fisher information to assess the role of topology on the thermometric performance of a given system. We find that low connectivity is a resource to build precise thermometers working at low temperatures, whereas highly connected systems are suitable for higher temperatures. Upon modelling the thermometer as a set of vertices for the quantum walk of an excitation, we compare the precision achievable by position measurement to the optimal one, which itself corresponds to energy measurement.
Deep neural networks have been widely used for feature learning in facial expression recognition systems. However, small datasets and large intra-class variability can lead to overfitting. In this paper, we propose a method which learns an optimized compact network topology for real-time facial expression recognition utilizing localized facial landmark features. Our method employs a spatio-temporal bilinear layer as backbone to capture the motion of facial landmarks during the execution of a facial expression effectively. Besides, it takes advantage of Monte Carlo Dropout to capture the model's uncertainty which is of great importance to analyze and treat uncertain cases. The performance of our method is evaluated on three widely used datasets and it is comparable to that of video-based state-of-the-art methods while it has much less complexity.
Accurate evaluation of the treatment result on X-ray images is a significant and challenging step in root canal therapy since the incorrect interpretation of the therapy results will hamper timely follow-up which is crucial to the patients' treatment outcome. Nowadays, the evaluation is performed in a manual manner, which is time-consuming, subjective, and error-prone. In this paper, we aim to automate this process by leveraging the advances in computer vision and artificial intelligence, to provide an objective and accurate method for root canal therapy result assessment. A novel anatomy-guided multi-branch Transformer (AGMB-Transformer) network is proposed, which first extracts a set of anatomy features and then uses them to guide a multi-branch Transformer network for evaluation. Specifically, we design a polynomial curve fitting segmentation strategy with the help of landmark detection to extract the anatomy features. Moreover, a branch fusion module and a multi-branch structure including our progressive Transformer and Group Multi-Head Self-Attention (GMHSA) are designed to focus on both global and local features for an accurate diagnosis. To facilitate the research, we have collected a large-scale root canal therapy evaluation dataset with 245 root canal therapy X-ray images, and the experiment results show that our AGMB-Transformer can improve the diagnosis accuracy from 57.96% to 90.20% compared with the baseline network. The proposed AGMB-Transformer can achieve a highly accurate evaluation of root canal therapy. To our best knowledge, our work is the first to perform automatic root canal therapy evaluation and has important clinical value to reduce the workload of endodontists.
Inference of population structure from genetic data plays an important role in population and medical genetics studies. The traditional EIGENSTRAT method has been widely used for computing and selecting top principal components that capture population structure information (Price et al., 2006). With the advancement and decreasing cost of sequencing technology, whole-genome sequencing data provide much richer information about the underlying population structures. However, the EIGENSTRAT method was originally developed for analyzing array-based genotype data and thus may not perform well on sequencing data for two reasons. First, the number of genetic variants $p$ is much larger than the sample size $n$ in sequencing data such that the sample-to-marker ratio $n/p$ is nearly zero, violating the assumption of the Tracy-Widom test used in the EIGENSTRAT method. Second, the EIGENSTRAT method might not be able to handle the linkage disequilibrium (LD) well in sequencing data. To resolve those two critical issues, we propose a new statistical method called ERStruct to estimate the number of latent sub-populations based on sequencing data. We propose to use the ratio of successive eigenvalues as a more robust testing statistic, and then we approximate the null distribution of our proposed test statistic using modern random matrix theory. Simulation studies found that our proposed ERStruct method has outperformed the traditional Tracy-Widom test on sequencing data. We further use two public data sets from the HapMap 3 and the 1000 Genomes Projects to demonstrate the performance of our ERStruct method. We also implement our ERStruct in a MATLAB toolbox which is now publicly available on GitHub through https://github.com/bglvly/ERStruct.
It is well known that the classic Allen-Cahn equation satisfies the maximum bound principle (MBP), that is, the absolute value of its solution is uniformly bounded for all time by certain constant under suitable initial and boundary conditions. In this paper, we consider numerical solutions of the modified Allen-Cahn equation with a Lagrange multiplier of nonlocal and local effects, which not only shares the same MBP as the original Allen-Cahn equation but also conserves the mass exactly. We reformulate the model equation with a linear stabilizing technique, then construct first- and second-order exponential time differencing schemes for its time integration. We prove the unconditional MBP preservation and mass conservation of the proposed schemes in the time discrete sense and derive their error estimates under some regularity assumptions. Various numerical experiments in two and three dimensions are also conducted to verify the theoretical results.
We prove that the set of Segre-degenerate points of a real-analytic subvariety $X$ in ${\mathbb{C}}^n$ is a closed semianalytic set. It is a subvariety if $X$ is coherent. More precisely, the set of points where the germ of the Segre variety is of dimension $k$ or greater is a closed semianalytic set in general, and for a coherent $X$, it is a real-analytic subvariety of $X$. For a hypersurface $X$ in ${\mathbb{C}}^n$, the set of Segre-degenerate points, $X_{[n]}$, is a semianalytic set of dimension at most $2n-4$. If $X$ is coherent, then $X_{[n]}$ is a complex subvariety of (complex) dimension $n-2$. Example hypersurfaces are given showing that $X_{[n]}$ need not be a subvariety and that it also needs not be complex; $X_{[n]}$ can, for instance, be a real line.
In the house credit process, banks and lenders rely on a fast and accurate estimation of a real estate price to determine the maximum loan value. Real estate appraisal is often based on relational data, capturing the hard facts of the property. Yet, models benefit strongly from including image data, capturing additional soft factors. The combination of the different data types requires a multi-view learning method. Therefore, the question arises which strengths and weaknesses different multi-view learning strategies have. In our study, we test multi-kernel learning, multi-view concatenation and multi-view neural networks on real estate data and satellite images from Asheville, NC. Our results suggest that multi-view learning increases the predictive performance up to 13% in MAE. Multi-view neural networks perform best, however result in intransparent black-box models. For users seeking interpretability, hybrid multi-view neural networks or a boosting strategy are a suitable alternative.
The electronic density of states (DOS) highlights fundamental properties of materials that oftentimes dictate their properties, such as the band gap and Van Hove singularities. In this short note, we discuss how sharp features of the density of states can be obscured by smearing methods (such as the Gaussian and Fermi smearing methods) when calculating the DOS. While the common approach to reach a "converged" density of states of a material is to increase the discrete k-point mesh density, we show that the DOS calculated by smearing methods can appear to converge but not to the correct DOS. Employing the tetrahedron method for Brillouin zone integration resolves key features of the density of states far better than smearing methods.
In this study, the in-plane Bloch wave propagation and bandgaps in a finitely stretched square lattice were investigated numerically and theoretically. To be specific, the elastic band diagram was calculated for an infinite periodic structure with a cruciform hyperelastic unit cell under uniaxial or biaxial tension. In addition, an elastodynamic "tight binding" model was proposed to investigate the formation and evolution of the band structure. The elastic waves were found to propagate largely under "easy" modes in the pre-stretched soft lattice, and finite stretch tuned the symmetry of the band structure, but also "purify" the propagation modes. Moreover, the uniaxial stretch exhibits the opposite impacts on the two "easy" modes. The effect of the biaxial stretch was equated with the superposition of the uniaxial stretches in the tessellation directions. The mentioned effects on the band structure could be attributed to the competition between the effective shear moduli and lengths for different beam components. Next, the finite stretch could tune the directional bandgap of the soft lattice, and the broadest elastic wave bandgaps could be anticipated in an equi-biaxial stretch. In this study, an avenue was opened to design and implement elastic wave control devices with weight efficiency and tunability. Furthermore, the differences between the physical system and the corresponding simplified theoretical model (e.g., the theoretically predicted flat bands) did not exist in the numerical calculations.
In this paper, a sparse Kronecker-product (SKP) coding scheme is proposed for unsourced multiple access. Specifically, the data of each active user is encoded as the Kronecker product of two component codewords with one being sparse and the other being forward-error-correction (FEC) coded. At the receiver, an iterative decoding algorithm is developed, consisting of matrix factorization for the decomposition of the Kronecker product and soft-in soft-out decoding for the component sparse code and the FEC code. The cyclic redundancy check (CRC) aided interference cancellation technique is further incorporated for performance improvement. Numerical results show that the proposed scheme outperforms the state-of-the-art counterparts, and approaches the random coding bound within a gap of only 0.1 dB at the code length of 30000 when the number of active users is less than 75, and the error rate can be made very small even if the number of active users is relatively large.
Deep neural networks for medical image reconstruction are traditionally trained using high-quality ground-truth images as training targets. Recent work on Noise2Noise (N2N) has shown the potential of using multiple noisy measurements of the same object as an alternative to having a ground-truth. However, existing N2N-based methods are not suitable for learning from the measurements of an object undergoing nonrigid deformation. This paper addresses this issue by proposing the deformation-compensated learning (DeCoLearn) method for training deep reconstruction networks by compensating for object deformations. A key component of DeCoLearn is a deep registration module, which is jointly trained with the deep reconstruction network without any ground-truth supervision. We validate DeCoLearn on both simulated and experimentally collected magnetic resonance imaging (MRI) data and show that it significantly improves imaging quality.
This paper develops a theory of polynomial maps from commutative semigroups to arbitrary groups and proves that it has desirable formal properties when the target group is locally nilpotent. We apply this theory to solve Waring's Problem for Heisenberg groups in a sequel to this paper.
Conditional image synthesis aims to create an image according to some multi-modal guidance in the forms of textual descriptions, reference images, and image blocks to preserve, as well as their combinations. In this paper, instead of investigating these control signals separately, we propose a new two-stage architecture, M6-UFC, to unify any number of multi-modal controls. In M6-UFC, both the diverse control signals and the synthesized image are uniformly represented as a sequence of discrete tokens to be processed by Transformer. Different from existing two-stage autoregressive approaches such as DALL-E and VQGAN, M6-UFC adopts non-autoregressive generation (NAR) at the second stage to enhance the holistic consistency of the synthesized image, to support preserving specified image blocks, and to improve the synthesis speed. Further, we design a progressive algorithm that iteratively improves the non-autoregressively generated image, with the help of two estimators developed for evaluating the compliance with the controls and evaluating the fidelity of the synthesized image, respectively. Extensive experiments on a newly collected large-scale clothing dataset M2C-Fashion and a facial dataset Multi-Modal CelebA-HQ verify that M6-UFC can synthesize high-fidelity images that comply with flexible multi-modal controls.
The scaling up of quantum hardware is the fundamental challenge ahead in order to realize the disruptive potential of quantum technology in information science. Among the plethora of hardware platforms, photonics stands out by offering a modular approach, where the main challenge is to construct sufficiently high-quality building blocks and develop methods to efficiently interface them. Importantly, the subsequent scaling-up will make full use of the mature integrated photonic technology provided by photonic foundry infrastructure to produce small foot-print quantum processors of immense complexity. A fully coherent and deterministic photon-emitter interface is a key enabler of quantum photonics, and can today be realized with solid-state quantum emitters with specifications reaching the quantitative benchmark referred to as Quantum Advantage. This light-matter interaction primer realizes a range of quantum photonic resources and functionalities, including on-demand single-photon and multi-photon entanglement sources, and photon-photon nonlinear quantum gates. We will present the current state-of-the-art in single-photon quantum hardware and the main photonic building blocks required in order to scale up. Furthermore, we will point out specific promising applications of the hardware building blocks within quantum communication and photonic quantum computing, laying out the road ahead for quantum photonics applications that could offer a genuine quantum advantage.
This paper presents an NLP (Natural Language Processing) approach to detecting spoilers in book reviews, using the University of California San Diego (UCSD) Goodreads Spoiler dataset. We explored the use of LSTM, BERT, and RoBERTa language models to perform spoiler detection at the sentence-level. This was contrasted with a UCSD paper which performed the same task, but using handcrafted features in its data preparation. Despite eschewing the use of handcrafted features, our results from the LSTM model were able to slightly exceed the UCSD team's performance in spoiler detection.
An iterative numerical method to compute the conformal mapping in the context of propagating water waves over uneven topographies is investigated. The map flattens the fluid domain onto a canonical strip in which computations are performed. The accuracy of the method is tested by using the MATLAB Schwarz-Christoffel toolbox mapping as a benchmark. Besides, we give a numerical alternative to compute the inverse of the conformal map.
We consider an investment process that includes a number of features, each of which can be active or inactive. Our goal is to attribute or decompose an achieved performance to each of these features, plus a baseline value. There are many ways to do this, which lead to potentially different attributions in any specific case. We argue that a specific attribution method due to Shapley is the preferred method, and discuss methods that can be used to compute this attribution exactly, or when that is not practical, approximately.
The evolution of circumstellar discs is highly influenced by their surroundings, in particular by external photoevaporation due to nearby stars and dynamical truncations. The impact of these processes on disc populations depends on the dynamical evolution of the star-forming region. Here we implement a simple model of molecular cloud collapse and star formation to obtain primordial positions and velocities of young stars and follow their evolution in time, including that of their circumstellar discs. Our disc model takes into account viscous evolution, internal and external photoevaporation, dust evolution, and dynamical truncations. The disc evolution is resolved simultaneously with the star cluster dynamics and stellar evolution. Our results show that an extended period of star formation allows for massive discs formed later in the simulations to survive for several million years. This could explain massive discs surviving in regions of high UV radiation.
This article proposes a framework for the study of periodic maps $T$ from a (typically finite) set $X$ to itself when the set $X$ is equipped with one or more real- or complex-valued functions. The main idea, inspired by the time-evolution operator construction from ergodic theory, is the introduction of a vector space that contains the given functions and is closed under composition with $T$, along with a time-evolution operator on that vector space. I show that the invariant functions and 0-mesic functions span complementary subspaces associated respectively with the eigenvalue 1 and the other eigenvalues. Alongside other examples, I give an explicit description of the spectrum of the evolution operator when $X$ is the set of $k$-element multisets with elements in $\{0,1,\dots,n-1\}$, $T$ increments each element of a multiset by 1 mod $n$, and $g_i: X \rightarrow \mathbb{R}$ (with $1 \leq i \leq k$) maps a multiset to its $i$th smallest element.
In the upcoming decades large facilities, such as the SKA, will provide resolved observations of the kinematics of millions of galaxies. In order to assist in the timely exploitation of these vast datasets we explore the use of a self-supervised, physics aware neural network capable of Bayesian kinematic modelling of galaxies. We demonstrate the network's ability to model the kinematics of cold gas in galaxies with an emphasis on recovering physical parameters and accompanying modelling errors. The model is able to recover rotation curves, inclinations and disc scale lengths for both CO and HI data which match well with those found in the literature. The model is also able to provide modelling errors over learned parameters thanks to the application of quasi-Bayesian Monte-Carlo dropout. This work shows the promising use of machine learning, and in particular self-supervised neural networks, in the context of kinematically modelling galaxies. This work represents the first steps in applying such models for kinematic fitting and we propose that variants of our model would seem especially suitable for enabling emission-line science from upcoming surveys with e.g. the SKA, allowing fast exploitation of these large datasets.
Context: Petri net slicing is a technique to reduce the size of a Petri net so that it can ease the analysis or understanding of the original Petri net. Objective: Presenting two new Petri net slicing algorithms to isolate those places and transitions of a Petri net (the slice) that may contribute tokens to one or more places given (the slicing criterion). Method: The two algorithms proposed are formalized. The completeness of the first algorithm and the minimality of the second algorithm are formally proven. Both algorithms together with other three state-of-the-art algorithms have been implemented and integrated into a single tool so that we have been able to carry out a fair empirical evaluation. Results: Besides the two new Petri net slicing algorithms, a public, free, and open-source implementation of five algorithms is reported. The results of an empirical evaluation of the new algorithms and the slices that they produce are also presented. Conclusions: The first algorithm collects all places and transitions that may influence (in any computation) the slicing criterion, while the second algorithm collects a minimum set of places and transitions that may influence (in some computation) the slicing criterion. Therefore, the net computed by the first algorithm can reproduce any computation that contributes tokens to any place of interest. In contrast, the second algorithm loses this possibility but it often produces a much more reduced subnet (which still can reproduce some computations that contribute tokens to some places of interest). The first algorithm is proven complete, and the second one is proven minimal.
The recently developed generalized Fourier-Galerkin method is complemented by a numerical continuation with respect to the kinetic energy, which extends the framework to the investigation of modal interactions resulting in folds of the nonlinear modes. In order to enhance the practicability regarding the investigation of complex large-scale systems, it is proposed to provide analytical gradients and exploit sparsity of the nonlinear part of the governing algebraic equations. A novel reduced order model (ROM) is developed for those regimes where internal resonances are absent. The approach allows for an accurate approximation of the multi-harmonic content of the resonant mode and accounts for the contributions of the off-resonant modes in their linearized forms. The ROM facilitates the efficient analysis of self-excited limit cycle oscillations, frequency response functions and the direct tracing of forced resonances. The ROM is equipped with a large parameter space including parameters associated with linear damping and near-resonant harmonic forcing terms. An important objective of this paper is to demonstrate the broad applicability of the proposed overall methodology. This is achieved by selected numerical examples including finite element models of structures with strongly nonlinear, non-conservative contact constraints.
In this position paper, we explore the adoption of a Smart City with a socio-technical perspective. A Smart city is a transformational technological process leading to profound modifications of existing urban regimes and infrastructure components. In this study, we consider a Smart City as a socio-technical system where the interplay between technologies and users ensures the sustainable development of smart city initiatives that improve the quality of life and solve important socio-economic problems. The adoption of a Smart City required a participative approach where users are involved during the adoption process to joint optimise both systems. Thus, we contribute to socio-technical research showing how a participative approach based on press relationships to facilitate information exchange between municipal actors and citizens worked as a success factor for the smart city adoption. We also discuss the limitations of this approach.
We investigate the role played by density inhomogeneities and dissipation on the final outcome of collapse of a self-gravitating sphere. By imposing a perturbative scheme on the thermodynamical variables and gravitational potentials we track the evolution of the collapse process starting off with an initially static perfect fluid sphere which is shear-free. The collapsing core dissipates energy in the form of a radial heat flux with the exterior spacetime being filled with a superposition of null energy and an anisotropic string distribution. The ensuing dynamical process slowly evolves into a shear-like regime with contributions from the heat flux and density fluctuations. We show that the anisotropy due to the presence of the strings drives the stellar fluid towards instability with this effect being enhanced by the density inhomogeneity. An interesting and novel consequence of this collapse scenario is the delay in the formation of the horizon.
Shapley Values, a solution to the credit assignment problem in cooperative game theory, are a popular type of explanation in machine learning, having been used to explain the importance of features, embeddings, and even neurons. In NLP, however, leave-one-out and attention-based explanations still predominate. Can we draw a connection between these different methods? We formally prove that -- save for the degenerate case -- attention weights and leave-one-out values cannot be Shapley Values. $\textit{Attention flow}$ is a post-processed variant of attention weights obtained by running the max-flow algorithm on the attention graph. Perhaps surprisingly, we prove that attention flows are indeed Shapley Values, at least at the layerwise level. Given the many desirable theoretical qualities of Shapley Values -- which has driven their adoption among the ML community -- we argue that NLP practitioners should, when possible, adopt attention flow explanations alongside more traditional ones.
Currently there has been increasing demand for real-time training on resource-limited IoT devices such as smart sensors, which realizes standalone online adaptation for streaming data without data transfers to remote servers. OS-ELM (Online Sequential Extreme Learning Machine) has been one of promising neural-network-based online algorithms for on-chip learning because it can perform online training at low computational cost and is easy to implement as a digital circuit. Existing OS-ELM digital circuits employ fixed-point data format and the bit-widths are often manually tuned, however, this may cause overflow or underflow which can lead to unexpected behavior of the circuit. For on-chip learning systems, an overflow/underflow-free design has a great impact since online training is continuously performed and the intervals of intermediate variables will dynamically change as time goes by. In this paper, we propose an overflow/underflow-free bit-width optimization method for fixed-point digital circuits of OS-ELM. Experimental results show that our method realizes overflow/underflow-free OS-ELM digital circuits with 1.0x - 1.5x more area cost compared to the baseline simulation method where overflow or underflow can happen.
In 1914 Bohr proved that there is an $r_0 \in(0,1)$ such that if a power series $\sum_{m=0}^\infty c_m z^m$ is convergent in the open unit disc and $\|\sum_{m=0}^\infty c_m z^m\|<1$ then, $\sum_{m=0}^\infty \|c_m z^m\|<1$ for $\|z\|<r_0$. The largest value of such $r_0$ is called the Bohr radius. In this article, we find Bohr radius for some univalent harmonic mappings having different dilatations and in addition, also compute Bohr radius for the functions convex in one direction.
Edge computing was introduced as a technical enabler for the demanding requirements of new network technologies like 5G. It aims to overcome challenges related to centralized cloud computing environments by distributing computational resources to the edge of the network towards the customers. The complexity of the emerging infrastructures increases significantly, together with the ramifications of outages on critical use cases such as self-driving cars or health care. Artificial Intelligence for IT Operations (AIOps) aims to support human operators in managing complex infrastructures by using machine learning methods. This paper describes the system design of an AIOps platform which is applicable in heterogeneous, distributed environments. The overhead of a high-frequency monitoring solution on edge devices is evaluated and performance experiments regarding the applicability of three anomaly detection algorithms on edge devices are conducted. The results show, that it is feasible to collect metrics with a high frequency and simultaneously run specific anomaly detection algorithms directly on edge devices with a reasonable overhead on the resource utilization.
Pushing is an essential non-prehensile manipulation skill used for tasks ranging from pre-grasp manipulation to scene rearrangement, reasoning about object relations in the scene, and thus pushing actions have been widely studied in robotics. The effective use of pushing actions often requires an understanding of the dynamics of the manipulated objects and adaptation to the discrepancies between prediction and reality. For this reason, effect prediction and parameter estimation with pushing actions have been heavily investigated in the literature. However, current approaches are limited because they either model systems with a fixed number of objects or use image-based representations whose outputs are not very interpretable and quickly accumulate errors. In this paper, we propose a graph neural network based framework for effect prediction and parameter estimation of pushing actions by modeling object relations based on contacts or articulations. Our framework is validated both in real and simulated environments containing different shaped multi-part objects connected via different types of joints and objects with different masses. Our approach enables the robot to predict and adapt the effect of a pushing action as it observes the scene. Further, we demonstrate 6D effect prediction in the lever-up action in the context of robot-based hard-disk disassembly.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.