text
null | inputs
dict | prediction
null | prediction_agent
null | annotation
list | annotation_agent
null | multi_label
bool 1
class | explanation
null | id
stringlengths 1
5
| metadata
null | status
stringclasses 2
values | event_timestamp
null | metrics
null |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
null |
{
"abstract": " Inference in log-linear models scales linearly in the size of output space in\nthe worst-case. This is often a bottleneck in natural language processing and\ncomputer vision tasks when the output space is feasibly enumerable but very\nlarge. We propose a method to perform inference in log-linear models with\nsublinear amortized cost. Our idea hinges on using Gumbel random variable\nperturbations and a pre-computed Maximum Inner Product Search data structure to\naccess the most-likely elements in sublinear amortized time. Our method yields\nprovable runtime and accuracy guarantees. Further, we present empirical\nexperiments on ImageNet and Word Embeddings showing significant speedups for\nsampling, inference, and learning in log-linear models.\n",
"title": "Fast Amortized Inference and Learning in Log-linear Models with Randomly Perturbed Nearest Neighbor Search"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
17401
| null |
Validated
| null | null |
null |
{
"abstract": " When an ordered spin system of a given dimensionality undergoes a second\norder phase transition the dependence of the order parameter i.e. magnetization\non temperature can be well-described by thermal excitations of elementary\ncollective spin excitations (magnons). However, the behavior of magnons\nthemselves, as a function of temperature and across the transition temperature\nTC, is an unknown issue. Utilizing spin-polarized high resolution electron\nenergy loss spectroscopy we monitor the high-energy (terahertz) magnons,\nexcited in an ultrathin ferromagnet, as a function of temperature. We show that\nthe magnons' energy and lifetime decrease with temperature. The\ntemperature-induced renormalization of the magnons' energy and lifetime depends\non the wave vector. We provide quantitative results on the temperature-induced\ndamping and discuss the possible mechanism e.g., multi-magnon scattering. A\ncareful investigation of physical quantities determining the magnons'\npropagation indicates that terahertz magnons sustain their propagating\ncharacter even at temperatures far above TC.\n",
"title": "Temperature Dependence of Magnetic Excitations: Terahertz Magnons above the Curie Temperature"
}
| null | null |
[
"Physics"
] | null | true | null |
17402
| null |
Validated
| null | null |
null |
{
"abstract": " We report the bright solitons of the generalized Gross-Pitaevskii (GP)\nequation with some types of physically relevant parity-time-(PT-) and\nnon-PT-symmetric potentials. We find that the constant momentum coefficient can\nmodulate the linear stability and complicated transverse power-flows (not\nalways from the gain toward loss) of nonlinear modes. However, the varying\nmomentum coefficient Gamma(x) can modulate both unbroken linear PT-symmetric\nphases and stability of nonlinear modes. Particularly, the nonlinearity can\nexcite the unstable linear mode (i.e., broken linear PT-symmetric phase) to\nstable nonlinear modes. Moreover, we also find stable bright solitons in the\npresence of non-PT-symmetric harmonic-Gaussian potential. The interactions of\ntwo bright solitons are also illustrated in PT-symmetric potentials. Finally,\nwe consider nonlinear modes and transverse power-flows in the three-dimensional\n(3D) GP equation with the generalized PT-symmetric Scarf-II potential\n",
"title": "On stable solitons and interactions of the generalized Gross-Pitaevskii equation with PT-and non-PT-symmetric potentials"
}
| null | null | null | null | true | null |
17403
| null |
Default
| null | null |
null |
{
"abstract": " The most recent experimental advances could provide ways for the fabrication\nof several atomic thick and planar forms of boron atoms. For the first time, we\nexplore the mechanical properties of five types of boron films with various\nvacancy ratios ranging from 0.1 to 0.15, using molecular dynamics simulations\nwith ReaxFF force field. It is found that the Young's modulus and tensile\nstrength decrease with increasing the temperature. We found that boron sheets\nexhibit an anisotropic mechanical response due to the different arrangement of\natoms along the armchair and zigzag directions. At room temperature, 2D Young's\nmodulus and fracture stress of these five sheets appear in the range 63 N/m and\n12 N/m, respectively. In addition, the strains at tensile strength are in the\nranges of 9, 11, and 10 percent at 1, 300, and 600 K, respectively. This\ninvestigation not only reveals the remarkable stiffness of 2D boron, but\nestablishes relations between the mechanical properties of the boron sheets to\nthe loading direction, temperature and atomic structures.\n",
"title": "Mechanical properties of borophene films: A reactive molecular dynamics investigation"
}
| null | null | null | null | true | null |
17404
| null |
Default
| null | null |
null |
{
"abstract": " Stronger selection implies faster evolution---that is, the greater the force,\nthe faster the change. This apparently self-evident proposition, however, is\nderived under the assumption that genetic variation within a population is\nprimarily supplied by mutation (i.e.\\ mutation-driven evolution). Here, we show\nthat this proposition does not actually hold for recombination-driven\nevolution, i.e.\\ evolution in which genetic variation is primarily created by\nrecombination rather than mutation. By numerically investigating population\ngenetics models of recombination, migration and selection, we demonstrate that\nstronger selection can slow down evolution on a perfectly smooth fitness\nlandscape. Through simple analytical calculation, this apparently\ncounter-intuitive result is shown to stem from two opposing effects of natural\nselection on the rate of evolution. On the one hand, natural selection tends to\nincrease the rate of evolution by increasing the fixation probability of fitter\ngenotypes. On the other hand, natural selection tends to decrease the rate of\nevolution by decreasing the chance of recombination between immigrants and\nresident individuals. As a consequence of these opposing effects, there is a\nfinite selection pressure maximizing the rate of evolution. Hence, stronger\nselection can imply slower evolution if genetic variation is primarily supplied\nby recombination.\n",
"title": "Stronger selection can slow down evolution driven by recombination on a smooth fitness landscape"
}
| null | null |
[
"Physics"
] | null | true | null |
17405
| null |
Validated
| null | null |
null |
{
"abstract": " L.J. Savage once hoped to show that \"the superficially incompatible systems\nof ideas associated on the one hand with [subjective Bayesianism] and on the\nother hand with [classical statistics]...lend each other mutual support and\nclarification.\" By 1972, however, he had largely \"lost faith in the devices\" of\nclassical statistics. One aspect of those \"devices\" that he found objectionable\nis that differences among the \"stopping rules\" that are used to decide when to\nend an experiment which are \"noninformative\" from a Bayesian perspective can\naffect decisions made using a classical approach. Two experiments that produce\nthe same data using different stopping rules seem to differ only in the\nintentions of the experimenters regarding whether or not they would have\ncarried on if the data had been different, which seem irrelevant to the\nevidential import of the data and thus to facts about what actions the data\nwarrant.\nI argue that classical and Bayesian ideas about stopping rules do in fact\n\"lend each other\" the kind of \"mutual support and clarification\" that Savage\nhad originally hoped to find. They do so in a kind of case that is common in\nscientific practice, in which those who design an experiment have different\ninterests from those who will make decisions in light of its results. I show\nthat, in cases of this kind, Bayesian principles provide qualified support for\nthe classical statistical practice of \"penalizing\" \"biased\" stopping rules.\nHowever, they require this practice in a narrower range of circumstances than\nclassical principles do, and for different reasons. I argue that classical\narguments for this practice are compelling in precisely the class of cases in\nwhich Bayesian principles also require it, and thus that we should regard\nBayesian principles as clarifying classical statistical ideas about stopping\nrules rather than the reverse.\n",
"title": "Differences Among Noninformative Stopping Rules Are Often Relevant to Bayesian Decisions"
}
| null | null | null | null | true | null |
17406
| null |
Default
| null | null |
null |
{
"abstract": " Stochastic Gradient Descent (SGD) is the central workhorse for training\nmodern CNNs. Although giving impressive empirical performance it can be slow to\nconverge. In this paper we explore a novel strategy for training a CNN using an\nalternation strategy that offers substantial speedups during training. We make\nthe following contributions: (i) replace the ReLU non-linearity within a CNN\nwith positive hard-thresholding, (ii) reinterpret this non-linearity as a\nbinary state vector making the entire CNN linear if the multi-layer support is\nknown, and (iii) demonstrate that under certain conditions a global optima to\nthe CNN can be found through local descent. We then employ a novel alternation\nstrategy (between weights and support) for CNN training that leads to\nsubstantially faster convergence rates, nice theoretical properties, and\nachieving state of the art results across large scale datasets (e.g. ImageNet)\nas well as other standard benchmarks.\n",
"title": "CNNs are Globally Optimal Given Multi-Layer Support"
}
| null | null | null | null | true | null |
17407
| null |
Default
| null | null |
null |
{
"abstract": " The Kontsevich integral is a powerful link invariant, taking values in spaces\nof Jacobi diagrams. In this paper, we extend the Kontsevich integral to\nconstruct a functor on the category of bottom tangles in handlebodies. This\nfunctor gives a universal finite type invariant of bottom tangles, and refines\na functorial version of the Le-Murakami-Ohtsuki 3-manifold invariant for\nLagrangian cobordisms of surfaces.\n",
"title": "The Kontsevich integral for bottom tangles in handlebodies"
}
| null | null | null | null | true | null |
17408
| null |
Default
| null | null |
null |
{
"abstract": " In this note we prove a selection of commutativity theorems for various\nclasses of semigroups. For instance, if in a separative or completely regular\nsemigroup $S$ we have $x^p y^p = y^p x^p$ and $x^q y^q = y^q x^q$ for all\n$x,y\\in S$ where $p$ and $q$ are relatively prime, then $S$ is commutative. In\na separative or inverse semigroup $S$, if there exist three consecutive\nintegers $i$ such that $(xy)^i = x^i y^i$ for all $x,y\\in S$, then $S$ is\ncommutative. Finally, if $S$ is a separative or inverse semigroup satisfying\n$(xy)^3=x^3y^3$ for all $x,y\\in S$, and if the cubing map $x\\mapsto x^3$ is\ninjective, then $S$ is commutative.\n",
"title": "Commutativity theorems for groups and semigroups"
}
| null | null | null | null | true | null |
17409
| null |
Default
| null | null |
null |
{
"abstract": " Short Message Service (SMS) spam is a serious problem in Vietnam because of\nthe availability of very cheap pre-paid SMS packages. There are some systems to\ndetect and filter spam messages for English, most of which use machine learning\ntechniques to analyze the content of messages and classify them. For\nVietnamese, there is some research on spam email filtering but none focused on\nSMS. In this work, we propose the first system for filtering Vietnamese spam\nSMS. We first propose an appropriate preprocessing method since existing tools\nfor Vietnamese preprocessing cannot give good accuracy on our dataset. We then\nexperiment with vector representations and classifiers to find the best model\nfor this problem. Our system achieves an accuracy of 94% when labelling spam\nmessages while the misclassification rate of legitimate messages is relatively\nsmall, about only 0.4%. This is an encouraging result compared to that of\nEnglish and can be served as a strong baseline for future development of\nVietnamese SMS spam prevention systems.\n",
"title": "Content-based Approach for Vietnamese Spam SMS Filtering"
}
| null | null | null | null | true | null |
17410
| null |
Default
| null | null |
null |
{
"abstract": " Measuring and analyzing the performance of software has reached a high\ncomplexity, caused by more advanced processor designs and the intricate\ninteraction between user programs, the operating system, and the processor's\nmicroarchitecture. In this report, we summarize our experience about how\nperformance characteristics of software should be measured when running on a\nLinux operating system and a modern processor. In particular, (1) We provide a\ngeneral overview about hardware and operating system features that may have a\nsignificant impact on timing and how they interact, (2) we identify sources of\nerrors that need to be controlled in order to obtain unbiased measurement\nresults, and (3) we propose a measurement setup for Linux to minimize errors.\nAlthough not the focus of this report, we describe the measurement process\nusing hardware performance counters, which can faithfully reflect the real\nbottlenecks on a given processor. Our experiments confirm that our measurement\nsetup has a large impact on the results. More surprisingly, however, they also\nsuggest that the setup can be negligible for certain analysis methods.\nFurthermore, we found that our setup maintains significantly better performance\nunder background load conditions, which means it can be used to improve\nsoftware in high-performance applications.\n",
"title": "Measuring Software Performance on Linux"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17411
| null |
Validated
| null | null |
null |
{
"abstract": " We present a method for calculating the complex Green function $G_{ij}\n(\\omega)$ at any real frequency $\\omega$ between any two sites $i$ and $j$ on a\nlattice. Starting from numbers of walks on square, cubic, honeycomb,\ntriangular, bcc, fcc, and diamond lattices, we derive Chebyshev expansion\ncoefficients for $G_{ij} (\\omega)$. The convergence of the Chebyshev series can\nbe accelerated by constructing functions $f(\\omega)$ that mimic the van Hove\nsingularities in $G_{ij} (\\omega)$ and subtracting their Chebyshev coefficients\nfrom the original coefficients. We demonstrate this explicitly for the square\nlattice and bcc lattice. Our algorithm achieves typical accuracies of 6--9\nsignificant figures using 1000 series terms.\n",
"title": "A general method for calculating lattice Green functions on the branch cut"
}
| null | null | null | null | true | null |
17412
| null |
Default
| null | null |
null |
{
"abstract": " Literature on the modeling and simulation of complex adaptive systems (cas)\nhas primarily advanced vertically in different scientific domains with\nscientists developing a variety of domain-specific approaches and applications.\nHowever, while cas researchers are inher-ently interested in an\ninterdisciplinary comparison of models, to the best of our knowledge, there is\ncurrently no single unified framework for facilitating the development,\ncomparison, communication and validation of models across different scientific\ndomains. In this thesis, we propose first steps towards such a unified\nframework using a combination of agent-based and complex network-based modeling\napproaches and guidelines formulated in the form of a set of four levels of\nusage, which allow multidisciplinary researchers to adopt a suitable framework\nlevel on the basis of available data types, their research study objectives and\nexpected outcomes, thus allowing them to better plan and conduct their\nrespective re-search case studies.\n",
"title": "Towards A Novel Unified Framework for Developing Formal, Network and Validated Agent-Based Simulation Models of Complex Adaptive Systems"
}
| null | null |
[
"Computer Science",
"Physics"
] | null | true | null |
17413
| null |
Validated
| null | null |
null |
{
"abstract": " In this paper, we study a new type of spatial sparse recovery problem, that\nis to infer the fine-grained spatial distribution of certain density data in a\nregion only based on the aggregate observations recorded for each of its\nsubregions. One typical example of this spatial sparse recovery problem is to\ninfer spatial distribution of cellphone activities based on aggregate mobile\ntraffic volumes observed at sparsely scattered base stations. We propose a\nnovel Constrained Spatial Smoothing (CSS) approach, which exploits the local\ncontinuity that exists in many types of spatial data to perform sparse recovery\nvia finite-element methods, while enforcing the aggregated observation\nconstraints through an innovative use of the ADMM algorithm. We also improve\nthe approach to further utilize additional geographical attributes. Extensive\nevaluations based on a large dataset of phone call records and a demographical\ndataset from the city of Milan show that our approach significantly outperforms\nvarious state-of-the-art approaches, including Spatial Spline Regression (SSR).\n",
"title": "Recover Fine-Grained Spatial Data from Coarse Aggregation"
}
| null | null | null | null | true | null |
17414
| null |
Default
| null | null |
null |
{
"abstract": " In the game theory literature, there appears to be little research on\nequilibrium selection for normal-form games with an infinite strategy space and\ndiscontinuous utility functions. Moreover, many existing selection methods are\nnot applicable to games involving both cooperative and noncooperative scenarios\n(e.g., \"games on signed graphs\"). With the purpose of equilibrium selection,\nthe power allocation game developed in \\cite{allocation}, which is a static,\nresource allocation game on signed graphs, will be reformulated into an\nextensive form. Results about the subgame perfect Nash equilibria in the\nextensive-form game will be given. This appears to be the first time that\nsubgame perfection based on time-varying graphs is used for equilibrium\nselection in network games. This idea of subgame perfection proposed in the\npaper may be extrapolated to other network games, which will be illustrated\nwith a simple example of congestion games.\n",
"title": "The Power Allocation Game on Dynamic Networks: Subgame Perfection"
}
| null | null | null | null | true | null |
17415
| null |
Default
| null | null |
null |
{
"abstract": " Superior performance and ease of implementation have fostered the adoption of\nConvolutional Neural Networks (CNNs) for a wide array of inference and\nreconstruction tasks. CNNs implement three basic blocks: convolution, pooling\nand pointwise nonlinearity. Since the two first operations are well-defined\nonly on regular-structured data such as audio or images, application of CNNs to\ncontemporary datasets where the information is defined in irregular domains is\nchallenging. This paper investigates CNNs architectures to operate on signals\nwhose support can be modeled using a graph. Architectures that replace the\nregular convolution with a so-called linear shift-invariant graph filter have\nbeen recently proposed. This paper goes one step further and, under the\nframework of multiple-input multiple-output (MIMO) graph filters, imposes\nadditional structure on the adopted graph filters, to obtain three new (more\nparsimonious) architectures. The proposed architectures result in a lower\nnumber of model parameters, reducing the computational complexity, facilitating\nthe training, and mitigating the risk of overfitting. Simulations show that the\nproposed simpler architectures achieve similar performance as more complex\nmodels.\n",
"title": "MIMO Graph Filters for Convolutional Neural Networks"
}
| null | null | null | null | true | null |
17416
| null |
Default
| null | null |
null |
{
"abstract": " Wavelet frame systems are known to be effective in capturing singularities\nfrom noisy and degraded images. In this paper, we introduce a new edge driven\nwavelet frame model for image restoration by approximating images as piecewise\nsmooth functions. With an implicit representation of image singularities sets,\nthe proposed model inflicts different strength of regularization on smooth and\nsingular image regions and edges. The proposed edge driven model is robust to\nboth image approximation and singularity estimation. The implicit formulation\nalso enables an asymptotic analysis of the proposed models and a rigorous\nconnection between the discrete model and a general continuous variational\nmodel. Finally, numerical results on image inpainting and deblurring show that\nthe proposed model is compared favorably against several popular image\nrestoration models.\n",
"title": "An Edge Driven Wavelet Frame Model for Image Restoration"
}
| null | null | null | null | true | null |
17417
| null |
Default
| null | null |
null |
{
"abstract": " A recently proposed exact algorithm for the maximum independent set problem\nis analyzed. The typical running time is improved exponentially in some\nparameter regions compared to simple binary search. The algorithm also\novercomes the core transition point, where the conventional leaf removal\nalgorithm fails, and works up to the replica symmetry breaking (RSB) transition\npoint. This suggests that a leaf removal core itself is not enough for typical\nhardness in the random maximum independent set problem, providing further\nevidence for RSB being the obstacle for algorithms in general.\n",
"title": "An exact algorithm exhibiting RS-RSB/easy-hard correspondence for the maximum independent set problem"
}
| null | null | null | null | true | null |
17418
| null |
Default
| null | null |
null |
{
"abstract": " Privatizing data is a useful strategy for increasing parallelism in a shared\nmemory multithreaded program. Independent cores can compute independently on\nduplicates of shared data, combining their results at the end of their\ncomputations. Conventional approaches to privatization, however, rely on\nexplicit static or dynamic memory allocation for duplicated state, increasing\nmemory footprint and contention for cache resources, especially in shared\ncaches. In this work, we describe CCache, a system for on-demand privatization\nof data manipulated by commutative operations. CCache garners the benefits of\nprivatization, without the increase in memory footprint or cache occupancy.\nEach core in CCache dynamically privatizes commutatively manipulated data,\noperating on a copy. Periodically or at the end of its computation, the core\nmerges its value with the value resident in memory, and when all cores have\nmerged, the in-memory copy contains the up-to-date value. We describe a\nlow-complexity architectural implementation of CCache that extends a\nconventional multicore to support on-demand privatization without using\nadditional memory for private copies. We evaluate CCache on several high-value\napplications, including random access key-value store, clustering, breadth\nfirst search and graph ranking, showing speedups upto 3.2X.\n",
"title": "Flexible Support for Fast Parallel Commutative Updates"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17419
| null |
Validated
| null | null |
null |
{
"abstract": " Cellular Electron Cryo-Tomography (CECT) is a powerful imaging technique for\nthe 3D visualization of cellular structure and organization at submolecular\nresolution. It enables analyzing the native structures of macromolecular\ncomplexes and their spatial organization inside single cells. However, due to\nthe high degree of structural complexity and practical imaging limitations,\nsystematic macromolecular structural recovery inside CECT images remains\nchallenging. Particularly, the recovery of a macromolecule is likely to be\nbiased by its neighbor structures due to the high molecular crowding. To reduce\nthe bias, here we introduce a novel 3D convolutional neural network inspired by\nFully Convolutional Network and Encoder-Decoder Architecture for the supervised\nsegmentation of macromolecules of interest in subtomograms. The tests of our\nmodels on realistically simulated CECT data demonstrate that our new approach\nhas significantly improved segmentation performance compared to our baseline\napproach. Also, we demonstrate that the proposed model has generalization\nability to segment new structures that do not exist in training data.\n",
"title": "Deep learning based supervised semantic segmentation of Electron Cryo-Subtomograms"
}
| null | null | null | null | true | null |
17420
| null |
Default
| null | null |
null |
{
"abstract": " Large-scale deep convolutional neural networks (CNNs) are widely used in\nmachine learning applications. While CNNs involve huge complexity, VLSI (ASIC\nand FPGA) chips that deliver high-density integration of computational\nresources are regarded as a promising platform for CNN's implementation. At\nmassive parallelism of computational units, however, the external memory\nbandwidth, which is constrained by the pin count of the VLSI chip, becomes the\nsystem bottleneck. Moreover, VLSI solutions are usually regarded as a lack of\nthe flexibility to be reconfigured for the various parameters of CNNs. This\npaper presents CNN-MERP to address these issues. CNN-MERP incorporates an\nefficient memory hierarchy that significantly reduces the bandwidth\nrequirements from multiple optimizations including on/off-chip data allocation,\ndata flow optimization and data reuse. The proposed 2-level reconfigurability\nis utilized to enable fast and efficient reconfiguration, which is based on the\ncontrol logic and the multiboot feature of FPGA. As a result, an external\nmemory bandwidth requirement of 1.94MB/GFlop is achieved, which is 55% lower\nthan prior arts. Under limited DRAM bandwidth, a system throughput of\n1244GFlop/s is achieved at the Vertex UltraScale platform, which is 5.48 times\nhigher than the state-of-the-art FPGA implementations.\n",
"title": "CNN-MERP: An FPGA-Based Memory-Efficient Reconfigurable Processor for Forward and Backward Propagation of Convolutional Neural Networks"
}
| null | null | null | null | true | null |
17421
| null |
Default
| null | null |
null |
{
"abstract": " We report evidence for an enstrophy cascade in large-scale point-vortex\nsimulations of decaying two-dimensional quantum turbulence. Devising a method\nto generate quantum vortex configurations with kinetic energy narrowly\nlocalized near a single length scale, the dynamics are found to be\nwell-characterised by a superfluid Reynolds number, $\\mathrm{Re_s}$, that\ndepends only on the number of vortices and the initial kinetic energy scale.\nUnder free evolution the vortices exhibit features of a classical enstrophy\ncascade, including a $k^{-3}$ power-law kinetic energy spectrum, and steady\nenstrophy flux associated with inertial transport to small scales. Clear\nsignatures of the cascade emerge for $N\\gtrsim 500$ vortices. Simulating up to\nvery large Reynolds numbers ($N = 32, 768$ vortices), additional features of\nthe classical theory are observed: the Kraichnan-Batchelor constant is found to\nconverge to $C' \\approx 1.6$, and the width of the $k^{-3}$ range scales as\n$\\mathrm{Re_s}^{1/2}$. The results support a universal phenomenology\nunderpinning classical and quantum fluid turbulence.\n",
"title": "Enstrophy Cascade in Decaying Two-Dimensional Quantum Turbulence"
}
| null | null | null | null | true | null |
17422
| null |
Default
| null | null |
null |
{
"abstract": " We study contextual multi-armed bandit problems under linear realizability on\nrewards and uncertainty (or noise) on features. For the case of identical noise\non features across actions, we propose an algorithm, coined {\\em NLinRel},\nhaving $O\\left(T^{\\frac{7}{8}} \\left(\\log{(dT)}+K\\sqrt{d}\\right)\\right)$ regret\nbound for $T$ rounds, $K$ actions, and $d$-dimensional feature vectors. Next,\nfor the case of non-identical noise, we observe that popular linear hypotheses\nincluding {\\em NLinRel} are impossible to achieve such sub-linear regret.\nInstead, under assumption of Gaussian feature vectors, we prove that a greedy\nalgorithm has $O\\left(T^{\\frac23}\\sqrt{\\log d}\\right)$ regret bound with\nrespect to the optimal linear hypothesis. Utilizing our theoretical\nunderstanding on the Gaussian case, we also design a practical variant of {\\em\nNLinRel}, coined {\\em Universal-NLinRel}, for arbitrary feature distributions.\nIt first runs {\\em NLinRel} for finding the `true' coefficient vector using\nfeature uncertainties and then adjust it to minimize its regret using the\nstatistical feature information. We justify the performance of {\\em\nUniversal-NLinRel} on both synthetic and real-world datasets.\n",
"title": "Contextual Multi-armed Bandits under Feature Uncertainty"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
17423
| null |
Validated
| null | null |
null |
{
"abstract": " Variational inference for latent variable models is prevalent in various\nmachine learning problems, typically solved by maximizing the Evidence Lower\nBound (ELBO) of the true data likelihood with respect to a variational\ndistribution. However, freely enriching the family of variational distribution\nis challenging since the ELBO requires variational likelihood evaluations of\nthe latent variables. In this paper, we propose a novel framework to enrich the\nvariational family by incorporating auxiliary variables to the variational\nfamily. The resulting inference network doesn't require density evaluations for\nthe auxiliary variables and thus complex implicit densities over the auxiliary\nvariables can be constructed by neural networks. It can be shown that the\nactual variational posterior of the proposed approach is essentially modeling a\nrich probabilistic mixture of simple variational posterior indexed by auxiliary\nvariables, thus a flexible inference model can be built. Empirical evaluations\non several density estimation tasks demonstrates the effectiveness of the\nproposed method.\n",
"title": "Asymmetric Variational Autoencoders"
}
| null | null | null | null | true | null |
17424
| null |
Default
| null | null |
null |
{
"abstract": " To understand the self-sustenance of subcritical turbulence in spectrally\nstable shear flows, we performed direct numerical simulations of homogeneous\nshear turbulence for different aspect ratios of the flow domain and analyzed\nthe dynamical processes in Fourier space. There are no exponentially growing\nmodes in such flows and the turbulence is energetically supported only by the\nlinear growth of perturbation harmonics due to the shear flow non-normality.\nThis non-normality-induced, or nonmodal growth is anisotropic in spectral\nspace, which, in turn, leads to anisotropy of nonlinear processes in this\nspace. As a result, a transverse (angular) redistribution of harmonics in\nFourier space appears to be the main nonlinear process in these flows, rather\nthan direct or inverse cascades. We refer to this type of nonlinear\nredistribution as the nonlinear transverse cascade. It is demonstrated that the\nturbulence is sustained by a subtle interplay between the linear nonmodal\ngrowth and the nonlinear transverse cascade that exemplifies a well-known\nbypass scenario of subcritical turbulence. These two basic processes mainly\noperate at large length scales, comparable to the domain size. Therefore, this\ncentral, small wave number area of Fourier space is crucial in the\nself-sustenance; we defined its size and labeled it as the vital area of\nturbulence. Outside the vital area, the nonmodal growth and the transverse\ncascade are of secondary importance. Although the cascades and the\nself-sustaining process of turbulence are qualitatively the same at different\naspect ratios, the number of harmonics actively participating in this process\nvaries, but always remains quite large. This implies that the self-sustenance\nof subcritical turbulence cannot be described by low-order models.\n",
"title": "Dynamics of homogeneous shear turbulence: A key role of the nonlinear transverse cascade in the bypass concept"
}
| null | null |
[
"Physics"
] | null | true | null |
17425
| null |
Validated
| null | null |
null |
{
"abstract": " Learning would be a convincing method to achieve coordination on an\nequilibrium. But does learning converge, and to what? We answer this question\nin generic 2-player, 2-strategy games, using Experience-Weighted Attraction\n(EWA), which encompasses many extensively studied learning algorithms. We\nexhaustively characterize the parameter space of EWA learning, for any payoff\nmatrix, and we understand the generic properties that imply convergent or\nnon-convergent behaviour in 2 x 2 games.\nIrrational choice and lack of incentives imply convergence to a mixed\nstrategy in the centre of the strategy simplex, possibly far from the Nash\nEquilibrium (NE). In the opposite limit, in which the players quickly modify\ntheir strategies, the behaviour depends on the payoff matrix: (i) a strong\ndiscrepancy between the pure strategies describes dominance-solvable games,\nwhich show convergence to a unique fixed point close to the NE; (ii) a\npreference towards profiles of strategies along the main diagonal describes\ncoordination games, with multiple stable fixed points corresponding to the NE;\n(iii) a cycle of best responses defines discoordination games, which commonly\nyield limit cycles or low-dimensional chaos.\nWhile it is well known that mixed strategy equilibria may be unstable, our\napproach is novel from several perspectives: we fully analyse EWA and provide\nexplicit thresholds that define the onset of instability; we find an emerging\ntaxonomy of the learning dynamics, without focusing on specific classes of\ngames ex-ante; we show that chaos can occur even in the simplest games; we make\na precise theoretical prediction that can be tested against data on\nexperimental learning of discoordination games.\n",
"title": "A taxonomy of learning dynamics in 2 x 2 games"
}
| null | null | null | null | true | null |
17426
| null |
Default
| null | null |
null |
{
"abstract": " This paper addresses the trajectory tracking control problem for\nunderactuated VTOL UAVs. According to the different actuation mechanisms, the\nmost common UAV platforms can achieve only a partial decoupling of attitude and\nposition tasks. Since position tracking is of utmost importance for\napplications involving aerial vehicles, we propose a control scheme in which\nposition tracking is the primary objective. To this end, this work introduces\nthe concept of attitude planner, a dynamical system through which the desired\nattitude reference is processed to guarantee the satisfaction of the primary\nobjective: the attitude tracking task is considered as a secondary objective\nwhich can be realized as long as the desired trajectory satisfies specific\ntrackability conditions. Two numerical simulations are performed by applying\nthe proposed control law to a hexacopter with and without tilted propellers,\nwhich accounts for unmodeled dynamics and external disturbances not included in\nthe control design model.\n",
"title": "Dynamic attitude planning for trajectory tracking in underactuated VTOL UAVs"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17427
| null |
Validated
| null | null |
null |
{
"abstract": " Though there has been a significant amount of work investigating the early\nstages of low-mass star formation in recent years, the evolution of the mass\nassembly rate onto the central protostar remains largely unconstrained.\nExamining in depth the variation in this rate is critical to understanding the\nphysics of star formation. Instabilities in the outer and inner circumstellar\ndisk can lead to episodic outbursts. Observing these brightness variations at\ninfrared or submillimetre wavelengths sets constraints on the current accretion\nmodels. The JCMT Transient Survey is a three-year project dedicated to studying\nthe continuum variability of deeply embedded protostars in eight nearby\nstar-forming regions at a one month cadence. We use the SCUBA-2 instrument to\nsimultaneously observe these regions at wavelengths of 450 $\\mu$m and 850\n$\\mu$m. In this paper, we present the data reduction techniques, image\nalignment procedures, and relative flux calibration methods for 850 $\\mu$m\ndata. We compare the properties and locations of bright, compact emission\nsources fitted with Gaussians over time. Doing so, we achieve a spatial\nalignment of better than 1\" between the repeated observations and an\nuncertainty of 2-3\\% in the relative peak brightness of significant, localised\nemission. This combination of imaging performance is unprecedented in\nground-based, single dish submillimetre observations. Finally, we identify a\nfew sources that show possible and confirmed brightness variations. These\nsources will be closely monitored and presented in further detail in additional\nstudies throughout the duration of the survey.\n",
"title": "The JCMT Transient Survey: Data Reduction and Calibration Methods"
}
| null | null | null | null | true | null |
17428
| null |
Default
| null | null |
null |
{
"abstract": " This paper gives new results for synchronization strings, a powerful\ncombinatorial object that allows to efficiently deal with insertions and\ndeletions in various communication settings:\n$\\bullet$ We give a deterministic, linear time synchronization string\nconstruction, improving over an $O(n^5)$ time randomized construction.\nIndependently of this work, a deterministic $O(n\\log^2\\log n)$ time\nconstruction was just put on arXiv by Cheng, Li, and Wu. We also give a\ndeterministic linear time construction of an infinite synchronization string,\nwhich was not known to be computable before. Both constructions are highly\nexplicit, i.e., the $i^{th}$ symbol can be computed in $O(\\log i)$ time.\n$\\bullet$ This paper also introduces a generalized notion we call\nlong-distance synchronization strings that allow for local and very fast\ndecoding. In particular, only $O(\\log^3 n)$ time and access to logarithmically\nmany symbols is required to decode any index.\nWe give several applications for these results:\n$\\bullet$ For any $\\delta<1$ and $\\epsilon>0$ we provide an insdel correcting\ncode with rate $1-\\delta-\\epsilon$ which can correct any $O(\\delta)$ fraction\nof insdel errors in $O(n\\log^3n)$ time. This near linear computational\nefficiency is surprising given that we do not even know how to compute the\n(edit) distance between the decoding input and output in sub-quadratic time. We\nshow that such codes can not only efficiently recover from $\\delta$ fraction of\ninsdel errors but, similar to [Schulman, Zuckerman; TransInf'99], also from any\n$O(\\delta/\\log n)$ fraction of block transpositions and replications.\n$\\bullet$ We show that highly explicitness and local decoding allow for\ninfinite channel simulations with exponentially smaller memory and decoding\ntime requirements. These simulations can be used to give the first near linear\ntime interactive coding scheme for insdel errors.\n",
"title": "Synchronization Strings: Explicit Constructions, Local Decoding, and Applications"
}
| null | null | null | null | true | null |
17429
| null |
Default
| null | null |
null |
{
"abstract": " We apply both distance-based (Jin and Matteson, 2017) and kernel-based\n(Pfister et al., 2016) mutual dependence measures to independent component\nanalysis (ICA), and generalize dCovICA (Matteson and Tsay, 2017) to MDMICA,\nminimizing empirical dependence measures as an objective function in both\ndeflation and parallel manners. Solving this minimization problem, we introduce\nLatin hypercube sampling (LHS) (McKay et al., 2000), and a global optimization\nmethod, Bayesian optimization (BO) (Mockus, 1994) to improve the initialization\nof the Newton-type local optimization method. The performance of MDMICA is\nevaluated in various simulation studies and an image data example. When the ICA\nmodel is correct, MDMICA achieves competitive results compared to existing\napproaches. When the ICA model is misspecified, the estimated independent\ncomponents are less mutually dependent than the observed components using\nMDMICA, while they are prone to be even more mutually dependent than the\nobserved components using other approaches.\n",
"title": "Independent Component Analysis via Energy-based and Kernel-based Mutual Dependence Measures"
}
| null | null |
[
"Statistics"
] | null | true | null |
17430
| null |
Validated
| null | null |
null |
{
"abstract": " We study the local geometry of testing a mean vector within a\nhigh-dimensional ellipse against a compound alternative. Given samples of a\nGaussian random vector, the goal is to distinguish whether the mean is equal to\na known vector within an ellipse, or equal to some other unknown vector in the\nellipse. Such ellipse testing problems lie at the heart of several\napplications, including non-parametric goodness-of-fit testing, signal\ndetection in cognitive radio, and regression function testing in reproducing\nkernel Hilbert spaces. While past work on such problems has focused on the\ndifficulty in a global sense, we study difficulty in a way that is localized to\neach vector within the ellipse. Our main result is to give sharp upper and\nlower bounds on the localized minimax testing radius in terms of an explicit\nformula involving the Kolmogorov width of the ellipse intersected with a\nEuclidean ball. When applied to particular examples, our general theorems yield\ninteresting rates that were not known before: as a particular case, for testing\nin Sobolev ellipses of smoothness $\\alpha$, we demonstrate rates that vary from\n$(\\sigma^2)^{\\frac{4 \\alpha}{4 \\alpha + 1}}$, corresponding to the classical\nglobal rate, to the faster rate $(\\sigma^2)^{\\frac{8\n\\alpha}{8 \\alpha + 1}}$, achievable for vectors at favorable locations within\nthe ellipse. We also show that the optimal test for this problem is achieved by\na linear projection test that is based on an explicit lower-dimensional\nprojection of the observation vector.\n",
"title": "The local geometry of testing in ellipses: Tight control via localized Kolmogorov widths"
}
| null | null | null | null | true | null |
17431
| null |
Default
| null | null |
null |
{
"abstract": " Two of the most fundamental prototypes of greedy optimization are the\nmatching pursuit and Frank-Wolfe algorithms. In this paper, we take a unified\nview on both classes of methods, leading to the first explicit convergence\nrates of matching pursuit methods in an optimization sense, for general sets of\natoms. We derive sublinear ($1/t$) convergence for both classes on general\nsmooth objectives, and linear convergence on strongly convex objectives, as\nwell as a clear correspondence of algorithm variants. Our presented algorithms\nand rates are affine invariant, and do not need any incoherence or sparsity\nassumptions.\n",
"title": "A Unified Optimization View on Generalized Matching Pursuit and Frank-Wolfe"
}
| null | null | null | null | true | null |
17432
| null |
Default
| null | null |
null |
{
"abstract": " We study the problem of subsampling in differential privacy (DP), a question\nthat is the centerpiece behind many successful differentially private machine\nlearning algorithms. Specifically, we provide a tight upper bound on the\nRényi Differential Privacy (RDP) (Mironov, 2017) parameters for algorithms\nthat: (1) subsample the dataset, and then (2) applies a randomized mechanism M\nto the subsample, in terms of the RDP parameters of M and the subsampling\nprobability parameter. Our results generalize the moments accounting technique,\ndeveloped by Abadi et al. (2016) for the Gaussian mechanism, to any subsampled\nRDP mechanism.\n",
"title": "Subsampled Rényi Differential Privacy and Analytical Moments Accountant"
}
| null | null | null | null | true | null |
17433
| null |
Default
| null | null |
null |
{
"abstract": " Assigning a satisfactory truly concurrent semantics to Petri nets with\nconfusion and distributed decisions is a long standing problem, especially if\none wants to fully replace nondeterminism with probability distributions and no\nstochastic structure is desired/allowed. Here we propose a general solution\nbased on a recursive, static decomposition of (finite, occurrence) nets in loci\nof decision, called structural branching cells (s-cells). Each s-cell exposes a\nset of alternatives, called transactions, that can be equipped with a general\nprobabilistic distribution. The solution is formalised as a transformation from\na given Petri net to another net whose transitions are the transactions of the\ns-cells and whose places are the places of the original net, with some\nauxiliary structure for bookkeeping. The resulting net is confusion-free,\nnamely if a transition is enabled, then all its conflicting alternatives are\nalso enabled. Thus sets of conflicting alternatives can be equipped with\nprobability distributions, while nonintersecting alternatives are purely\nconcurrent and do not introduce any nondeterminism: they are Church-Rosser and\ntheir probability distributions are independent. The validity of the\nconstruction is witnessed by a tight correspondence result with the recent\napproach by Abbes and Benveniste (AB) based on recursively stopped\nconfigurations in event structures. Some advantages of our approach over AB's\nare that: i) s-cells are defined statically and locally in a compositional way,\nwhereas AB's branching cells are defined dynamically and globally; ii) their\nrecursively stopped configurations correspond to possible executions, but the\nexisting concurrency is not made explicit. Instead, our resulting nets are\nequipped with an original concurrency structure exhibiting a so-called complete\nconcurrency property.\n",
"title": "Concurrency and Probability: Removing Confusion, Compositionally"
}
| null | null | null | null | true | null |
17434
| null |
Default
| null | null |
null |
{
"abstract": " We propose a technique for multi-task learning from demonstration that trains\nthe controller of a low-cost robotic arm to accomplish several complex picking\nand placing tasks, as well as non-prehensile manipulation. The controller is a\nrecurrent neural network using raw images as input and generating robot arm\ntrajectories, with the parameters shared across the tasks. The controller also\ncombines VAE-GAN-based reconstruction with autoregressive multimodal action\nprediction. Our results demonstrate that it is possible to learn complex\nmanipulation tasks, such as picking up a towel, wiping an object, and\ndepositing the towel to its previous position, entirely from raw images with\ndirect behavior cloning. We show that weight sharing and reconstruction-based\nregularization substantially improve generalization and robustness, and\ntraining on multiple tasks simultaneously increases the success rate on all\ntasks.\n",
"title": "Vision-Based Multi-Task Manipulation for Inexpensive Robots Using End-To-End Learning from Demonstration"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17435
| null |
Validated
| null | null |
null |
{
"abstract": " We present a probabilistic language model for time-stamped text data which\ntracks the semantic evolution of individual words over time. The model\nrepresents words and contexts by latent trajectories in an embedding space. At\neach moment in time, the embedding vectors are inferred from a probabilistic\nversion of word2vec [Mikolov et al., 2013]. These embedding vectors are\nconnected in time through a latent diffusion process. We describe two scalable\nvariational inference algorithms--skip-gram smoothing and skip-gram\nfiltering--that allow us to train the model jointly over all times; thus\nlearning on all data while simultaneously allowing word and context vectors to\ndrift. Experimental results on three different corpora demonstrate that our\ndynamic model infers word embedding trajectories that are more interpretable\nand lead to higher predictive likelihoods than competing methods that are based\non static models trained separately on time slices.\n",
"title": "Dynamic Word Embeddings"
}
| null | null | null | null | true | null |
17436
| null |
Default
| null | null |
null |
{
"abstract": " The multiple scattering of ultra relativistic electrons in an amorphous\nmatter leads to the suppression of the soft part of radiation spectrum (the\nLandau-Pomeranchuk-Migdal effect), and also can change essentially the angular\ndistribution of the emitted photons. A similar effect must take place in a\ncrystal for the coherent radiation of relativistic electron. The results of the\ntheoretical investigation of angular distributions and polarization of\nradiation by a relativistic electron passing through a thin (in comparison with\na coherence length) crystal at a small angle to the crystal axis are presented.\nThe electron trajectories in crystal were simulated using the binary collision\nmodel which takes into account both coherent and incoherent effects at\nscattering. The angular distribution of radiation and polarization were\ncalculated as a sum of radiation from each electron. It is shown that there are\nnontrivial angular distributions of the emitted photons and their polarization\nthat are connected to the superposition of the coherent scattering of electrons\nby atomic rows (\"doughnut scattering\" effect) and the suppression of radiation\n(similar to the Landau-Pomeranchuk-Migdal effect in an amorphous matter). It is\nalso shown that circular polarization of radiation in the considered case is\nidentically zero.\n",
"title": "Multiple scattering effect on angular distribution and polarization of radiation by relativistic electrons in a thin crystal"
}
| null | null | null | null | true | null |
17437
| null |
Default
| null | null |
null |
{
"abstract": " We give necessary and sufficient conditions for a 4-manifold to be a branched\ncovering of $CP^2$, $S^2\\times S^2$, $S^2 \\mathbin{\\tilde\\times} S^2$ and $S^3\n\\times S^1$, which are expressed in terms of the Betti numbers and the\nintersection form of the 4-manifold.\n",
"title": "Branched coverings of $CP^2$ and other basic 4-manifolds"
}
| null | null | null | null | true | null |
17438
| null |
Default
| null | null |
null |
{
"abstract": " The photoelectric effect established by Einstein is well known, which\nindicates that electrons on lower energy levels can jump up to higher levels by\nabsorbing photons, or jump down from higher levels to lower levels and give out\nphotons1-3. However, how do photons act on electrons and further on atoms have\nkept unknown up to now. Here we show the results that photons collide on\nelectrons with energy-transmission in semiconductors and pass their momenta to\nelectrons, which make the electrons jump up from lower energy levels to higher\nlevels. We found that (i) photons have rest mass of 7.287exp(-38) kg and\n2.886exp(-35) kg, in vacuum and silicon respectively; (ii) excited by photons\nwith energy of 1.12eV, electrons in silicon may jump up from the top of valance\nband to the bottom of conduction band with initial speed of 2.543exp(3) m/s and\ntaking time of 4.977exp(-17) s; (iii) acted by photons with energy of 4.6eV,\nthe atoms who lose electrons may be catapulted out of the semiconductors by the\nextruded neighbor atoms, and taking time of 2.224exp(-15) s. These results make\nreasonable explanation to rapid thermal annealing, laser ablation and laser\ncutting.\n",
"title": "Instantaneous effects of photons on electrons in semiconductors"
}
| null | null | null | null | true | null |
17439
| null |
Default
| null | null |
null |
{
"abstract": " We apply sequence-to-sequence model to mitigate the impact of speech\nrecognition errors on open domain end-to-end dialog generation. We cast the\ntask as a domain adaptation problem where ASR transcriptions and original text\nare in two different domains. In this paper, our proposed model includes two\nindividual encoders for each domain data and make their hidden states similar\nto ensure the decoder predict the same dialog text. The method shows that the\nsequence-to-sequence model can learn the ASR transcriptions and original text\npair having the same meaning and eliminate the speech recognition errors.\nExperimental results on Cornell movie dialog dataset demonstrate that the\ndomain adaption system help the spoken dialog system generate more similar\nresponses with the original text answers.\n",
"title": "Mitigating the Impact of Speech Recognition Errors on Chatbot using Sequence-to-Sequence Model"
}
| null | null | null | null | true | null |
17440
| null |
Default
| null | null |
null |
{
"abstract": " We investigate the use of 183 GHz H2O masers for characterization of the\nphysical conditions and mass loss process in the circumstellar envelopes of\nevolved stars. We used APEX SEPIA Band 5 to observe the 183 GHz H2O line\ntowards 2 Red Supergiant and 3 Asymptotic Giant Branch stars. Simultaneously,\nwe observed lines in 28SiO v0, 1, 2 and 3, and for 29SiO v0 and 1. We detected\nthe 183 GHz H2O line towards all the stars with peak flux densities greater\nthan 100 Jy, including a new detection from VY CMa. Towards all 5 targets, the\nwater line had indications of being due to maser emission and had higher peak\nflux densities than for the SiO lines. The SiO lines appear to originate from\nboth thermal and maser processes. Comparison with simulations and models\nindicate that 183 GHz maser emission is likely to extend to greater radii in\nthe circumstellar envelopes than SiO maser emission and to similar or greater\nradii than water masers at 22, 321 and 325 GHz. We speculate that a prominent\nblue-shifted feature in the W Hya 183 GHz spectrum is amplifying the stellar\ncontinuum, and is located at a similar distance from the star as mainline OH\nmaser emission. From a comparison of the individual polarizations, we find that\nthe SiO maser linear polarization fraction of several features exceeds the\nmaximum fraction allowed under standard maser assumptions and requires strong\nanisotropic pumping of the maser transition and strongly saturated maser\nemission. The low polarization fraction of the H2O maser however, fits with the\nexpectation for a non-saturated maser. 183 GHz H2O masers can provide strong\nprobes of the mass loss process of evolved stars. Higher angular resolution\nobservations of this line using ALMA Band 5 will enable detailed investigation\nof the emission location in circumstellar envelopes and can also provide\ninformation on magnetic field strength and structure.\n",
"title": "Simultaneous 183 GHz H2O Maser and SiO Observations Towards Evolved Stars Using APEX SEPIA Band 5"
}
| null | null | null | null | true | null |
17441
| null |
Default
| null | null |
null |
{
"abstract": " The free energy principle has been proposed as a unifying theory of brain\nfunction. It is closely related, and in some cases subsumes, earlier unifying\nideas such as Bayesian inference, predictive coding, and active learning. This\narticle clarifies these connections, teasing apart distinctive and shared\npredictions.\n",
"title": "What does the free energy principle tell us about the brain?"
}
| null | null | null | null | true | null |
17442
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we study the setting where features are added or change\ninterpretation over time, which has applications in multiple domains such as\nretail, manufacturing, finance. In particular, we propose an approach to\nprovably determine the time instant from which the new/changed features start\nbecoming relevant with respect to an output variable in an agnostic\n(supervised) learning setting. We also suggest an efficient version of our\napproach which has the same asymptotic performance. Moreover, our theory also\napplies when we have more than one such change point. Independent post analysis\nof a change point identified by our method for a large retailer revealed that\nit corresponded in time with certain unflattering news stories about a brand\nthat resulted in the change in customer behavior. We also applied our method to\ndata from an advanced manufacturing plant identifying the time instant from\nwhich downstream features became relevant. To the best of our knowledge this is\nthe first work that formally studies change point detection in a distribution\nindependent agnostic setting, where the change point is based on the changing\nrelationship between input and output.\n",
"title": "Learning with Changing Features"
}
| null | null | null | null | true | null |
17443
| null |
Default
| null | null |
null |
{
"abstract": " Contextual bandit algorithms are sensitive to the estimation method of the\noutcome model as well as the exploration method used, particularly in the\npresence of rich heterogeneity or complex outcome models, which can lead to\ndifficult estimation problems along the path of learning. We study a\nconsideration for the exploration vs. exploitation framework that does not\narise in multi-armed bandits but is crucial in contextual bandits; the way\nexploration and exploitation is conducted in the present affects the bias and\nvariance in the potential outcome model estimation in subsequent stages of\nlearning. We develop parametric and non-parametric contextual bandits that\nintegrate balancing methods from the causal inference literature in their\nestimation to make it less prone to problems of estimation bias. We provide the\nfirst regret bound analyses for contextual bandits with balancing in the domain\nof linear contextual bandits that match the state of the art regret bounds. We\ndemonstrate the strong practical advantage of balanced contextual bandits on a\nlarge number of supervised learning datasets and on a synthetic example that\nsimulates model mis-specification and prejudice in the initial training data.\nAdditionally, we develop contextual bandits with simpler assignment policies by\nleveraging sparse model estimation methods from the econometrics literature and\ndemonstrate empirically that in the early stages they can improve the rate of\nlearning and decrease regret.\n",
"title": "Estimation Considerations in Contextual Bandits"
}
| null | null | null | null | true | null |
17444
| null |
Default
| null | null |
null |
{
"abstract": " Smart solar inverters can be used to store, monitor and manage a home's solar\nenergy. We describe a smart solar inverter system with battery which can either\noperate in an automatic mode or receive commands over a network to charge and\ndischarge at a given rate. In order to make battery storage financially viable\nand advantageous to the consumers, effective battery scheduling algorithms can\nbe employed. Particularly, when time-of-use tariffs are in effect in the region\nof the inverter, it is possible in some cases to schedule the battery to save\nmoney for the individual customer, compared to the \"automatic\" mode. Hence,\nthis paper presents and evaluates the performance of a novel battery scheduling\nalgorithm for residential consumers of solar energy. The proposed battery\nscheduling algorithm optimizes the cost of electricity over next 24 hours for\nresidential consumers. The cost minimization is realized by controlling the\ncharging/discharging of battery storage system based on the predictions for\nload and solar power generation values. The scheduling problem is formulated as\na linear programming problem. We performed computer simulations over 83\ninverters using several months of hourly load and PV data. The simulation\nresults indicate that key factors affecting the viability of optimization are\nthe tariffs and the PV to Load ratio at each inverter. Depending on the tariff,\nsavings of between 1% and 10% can be expected over the automatic approach. The\nprediction approach used in this paper is also shown to out-perform basic\n\"persistence\" forecasting approaches. We have also examined the approaches for\nimproving the prediction accuracy and optimization effectiveness.\n",
"title": "Using solar and load predictions in battery scheduling at the residential level"
}
| null | null | null | null | true | null |
17445
| null |
Default
| null | null |
null |
{
"abstract": " Despite recent advances in face recognition using deep learning, severe\naccuracy drops are observed for large pose variations in unconstrained\nenvironments. Learning pose-invariant features is one solution, but needs\nexpensively labeled large-scale data and carefully designed feature learning\nalgorithms. In this work, we focus on frontalizing faces in the wild under\nvarious head poses, including extreme profile views. We propose a novel deep 3D\nMorphable Model (3DMM) conditioned Face Frontalization Generative Adversarial\nNetwork (GAN), termed as FF-GAN, to generate neutral head pose face images. Our\nframework differs from both traditional GANs and 3DMM based modeling.\nIncorporating 3DMM into the GAN structure provides shape and appearance priors\nfor fast convergence with less training data, while also supporting end-to-end\ntraining. The 3DMM-conditioned GAN employs not only the discriminator and\ngenerator loss but also a new masked symmetry loss to retain visual quality\nunder occlusions, besides an identity loss to recover high frequency\ninformation. Experiments on face recognition, landmark localization and 3D\nreconstruction consistently show the advantage of our frontalization method on\nfaces in the wild datasets.\n",
"title": "Towards Large-Pose Face Frontalization in the Wild"
}
| null | null | null | null | true | null |
17446
| null |
Default
| null | null |
null |
{
"abstract": " Citizen science projects recruit members of the public as volunteers to\nprocess and produce datasets. These datasets must win the trust of the\nscientific community. The task of securing credibility involves, in part,\napplying standard scientific procedures to clean these datasets. However,\neffective management of volunteer behavior also makes a significant\ncontribution to enhancing data quality. Through a case study of Galaxy Zoo, a\ncitizen science project set up to generate datasets based on volunteer\nclassifications of galaxy morphologies, this paper explores how those involved\nin running the project manage volunteers. The paper focuses on how methods for\ncrediting volunteer contributions motivate volunteers to provide higher quality\ncontributions and to behave in a way that better corresponds to statistical\nassumptions made when combining volunteer contributions into datasets. These\nmethods have made a significant contribution to the success of the project in\nsecuring trust in these datasets, which have been well used by other\nscientists. Implications for practice are then presented for citizen science\nprojects, providing a list of considerations to guide choices regarding how to\ncredit volunteer contributions to improve the quality and trustworthiness of\ncitizen science-produced datasets.\n",
"title": "Managing the Public to Manage Data: Citizen Science and Astronomy"
}
| null | null | null | null | true | null |
17447
| null |
Default
| null | null |
null |
{
"abstract": " We study the category of representations of $\\mathfrak{sl}_{m+2n}$ in\npositive characteristic, whose p-character is a nilpotent whose Jordan type is\nthe two-row partition (m+n,n). In a previous paper with Anno, we used\nBezrukavnikov-Mirkovic-Rumynin's theory of positive characteristic localization\nand exotic t-structures to give a geometric parametrization of the simples\nusing annular crossingless matchings. Building on this, here we give\ncombinatorial dimension formulae for the simple objects, and compute the\nJordan-Holder multiplicities of the simples inside the baby Vermas (in special\ncase where n=1, i.e. that a subregular nilpotent, these were known from work of\nJantzen). We use Cautis-Kamnitzer's geometric categorification of the tangle\ncalculus to study the images of the simple objects under the [BMR] equivalence.\nThe dimension formulae may be viewed as a positive characteristic analogue of\nthe combinatorial character formulae for simple objects in parabolic category O\nfor $\\mathfrak{sl}_{m+2n}$, due to Lascoux and Schutzenberger.\n",
"title": "Modular representations in type A with a two-row nilpotent central character"
}
| null | null | null | null | true | null |
17448
| null |
Default
| null | null |
null |
{
"abstract": " Complex networks have been found to provide a good representation of the\nstructure of knowledge, as understood in terms of discoverable concepts and\ntheir relationships. In this context, the discovery process can be modeled as\nagents walking in a knowledge space. Recent studies proposed more realistic\ndynamics, including the possibility of agents being influenced by others with\nhigher visibility or by their own memory. However, rather than dealing with\nthese two concepts separately, as previously approached, in this study we\npropose a multi-agent random walk model for knowledge acquisition that\nincorporates both concepts. More specifically, we employed the true self\navoiding walk alongside a new dynamics based on jumps, in which agents are\nattracted by the influence of others. That was achieved by using a Lévy\nflight influenced by a field of attraction emanating from the agents. In order\nto evaluate our approach, we use a set of network models and two real networks,\none generated from Wikipedia and another from the Web of Science. The results\nwere analyzed globally and by regions. In the global analysis, we found that\nmost of the dynamics parameters do not significantly affect the discovery\ndynamics. The local analysis revealed a substantial difference of performance\ndepending on the network regions where the dynamics are occurring. In\nparticular, the dynamics at the core of networks tend to be more effective. The\nchoice of the dynamics parameters also had no significant impact to the\nacquisition performance for the considered knowledge networks, even at the\nlocal scale.\n",
"title": "Knowledge Acquisition: A Complex Networks Approach"
}
| null | null | null | null | true | null |
17449
| null |
Default
| null | null |
null |
{
"abstract": " We analyze short cadence K2 light curve of the TRAPPIST-1 system. Fourier\nanalysis of the data suggests $P_\\mathrm{rot}=3.295\\pm0.003$ days. The light\ncurve shows several flares, of which we analyzed 42 events, these have\nintegrated flare energies of $1.26\\times10^{30}-1.24\\times10^{33}$ ergs.\nApproximately 12% of the flares were complex, multi-peaked eruptions. The\nflaring and the possible rotational modulation shows no obvious correlation.\nThe flaring activity of TRAPPIST-1 probably continuously alters the atmospheres\nof the orbiting exoplanets, making these less favorable for hosting life.\n",
"title": "Frequent flaring in the TRAPPIST-1 system - unsuited for life?"
}
| null | null | null | null | true | null |
17450
| null |
Default
| null | null |
null |
{
"abstract": " As robots become increasingly prevalent in almost all areas of society, the\nfactors affecting humans trust in those robots becomes increasingly important.\nThis paper is intended to investigate the factor of robot attributes, looking\nspecifically at the relationship between anthropomorphism and human development\nof trust. To achieve this, an interaction game, Matching the Pairs, was\ndesigned and implemented on two robots of varying levels of anthropomorphism,\nPepper and Husky. Participants completed both pre- and post-test questionnaires\nthat were compared and analyzed predominantly with the use of quantitative\nmethods, such as paired sample t-tests. Post-test analyses suggested a positive\nrelationship between trust and anthropomorphism with $80\\%$ of participants\nconfirming that the robots' adoption of facial features assisted in\nestablishing trust. The results also indicated a positive relationship between\ninteraction and trust with $90\\%$ of participants confirming this for both\nrobots post-test\n",
"title": "Playing Pairs with Pepper"
}
| null | null | null | null | true | null |
17451
| null |
Default
| null | null |
null |
{
"abstract": " Let $T$ be a consistent o-minimal theory extending the theory of densely\nordered groups and let $T'$ be a consistent theory. Then there is a complete\ntheory $T^*$ extending $T$ such that $T$ is an open core of $T^*$, but every\nmodel of $T^*$ interprets a model of $T'$. If $T'$ is NIP, $T^*$ can be chosen\nto be NIP as well. From this we deduce the existence of an NIP expansion of the\nreal field that has no distal expansion.\n",
"title": "Wild theories with o-minimal open core"
}
| null | null | null | null | true | null |
17452
| null |
Default
| null | null |
null |
{
"abstract": " Standard Bayesian analyses can be difficult to perform when the full\nlikelihood, and consequently the full posterior distribution, is too complex\nand difficult to specify or if robustness with respect to data or to model\nmisspecifications is required. In these situations, we suggest to resort to a\nposterior distribution for the parameter of interest based on proper scoring\nrules. Scoring rules are loss functions designed to measure the quality of a\nprobability distribution for a random variable, given its observed value.\nImportant examples are the Tsallis score and the Hyvärinen score, which allow\nus to deal with model misspecifications or with complex models. Also the full\nand the composite likelihoods are both special instances of scoring rules.\nThe aim of this paper is twofold. Firstly, we discuss the use of scoring\nrules in the Bayes formula in order to compute a posterior distribution, named\nSR-posterior distribution, and we derive its asymptotic normality. Secondly, we\npropose a procedure for building default priors for the unknown parameter of\ninterest that can be used to update the information provided by the scoring\nrule in the SR-posterior distribution. In particular, a reference prior is\nobtained by maximizing the average $\\alpha-$divergence from the SR-posterior\ndistribution. For $0 \\leq |\\alpha|<1$, the result is a Jeffreys-type prior that\nis proportional to the square root of the determinant of the Godambe\ninformation matrix associated to the scoring rule. Some examples are discussed.\n",
"title": "Objective Bayesian inference with proper scoring rules"
}
| null | null | null | null | true | null |
17453
| null |
Default
| null | null |
null |
{
"abstract": " Large Area X-ray Proportional Counter (LAXPC) is one of the major AstroSat\npayloads. LAXPC instrument will provide high time resolution X-ray observations\nin 3 to 80 keV energy band with moderate energy resolution. A cluster of three\nco-aligned identical LAXPC detectors is used in AstroSat to provide large\ncollection area of more than 6000 cm2 . The large detection volume (15 cm\ndepth) filled with xenon gas at about 2 atmosphere pressure, results in\ndetection efficiency greater than 50%, above 30 keV. With its broad energy\nrange and fine time resolution (10 microsecond), LAXPC instrument is well\nsuited for timing and spectral studies of a wide variety of known and transient\nX-ray sources in the sky. We have done extensive calibration of all LAXPC\ndetectors using radioactive sources as well as GEANT4 simulation of LAXPC\ndetectors. We describe in brief some of the results obtained during the payload\nverification phase along with LXAPC capabilities.\n",
"title": "Large Area X-ray Proportional Counter (LAXPC) Instrument on AstroSat"
}
| null | null |
[
"Physics"
] | null | true | null |
17454
| null |
Validated
| null | null |
null |
{
"abstract": " We give a counterexample to the vector generalization of Costa's entropy\npower inequality (EPI) due to Liu, Liu, Poor and Shamai. In particular, the\nclaimed inequality can fail if the matix-valued parameter in the convex\ncombination does not commute with the covariance of the additive Gaussian\nnoise. Conversely, the inequality holds if these two matrices commute.\n",
"title": "A Counterexample to the Vector Generalization of Costa's EPI, and Partial Resolution"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17455
| null |
Validated
| null | null |
null |
{
"abstract": " In this work, we ask two questions: 1. Can we predict the type of community\ninterested in a news article using only features from the article content? and\n2. How well do these models generalize over time? To answer these questions, we\ncompute well-studied content-based features on over 60K news articles from 4\ncommunities on reddit.com. We train and test models over three different time\nperiods between 2015 and 2017 to demonstrate which features degrade in\nperformance the most due to concept drift. Our models can classify news\narticles into communities with high accuracy, ranging from 0.81 ROC AUC to 1.0\nROC AUC. However, while we can predict the community-specific popularity of\nnews articles with high accuracy, practitioners should approach these models\ncarefully. Predictions are both community-pair dependent and feature group\ndependent. Moreover, these feature groups generalize over time differently,\nwith some only degrading slightly over time, but others degrading greatly.\nTherefore, we recommend that community-interest predictions are done in a\nhierarchical structure, where multiple binary classifiers can be used to\nseparate community pairs, rather than a traditional multi-class model. Second,\nthese models should be retrained over time based on accuracy goals and the\navailability of training data.\n",
"title": "Models for Predicting Community-Specific Interest in News Articles"
}
| null | null | null | null | true | null |
17456
| null |
Default
| null | null |
null |
{
"abstract": " Hilbert's 14th problem studies the finite generation property of the\nintersection of an integral algebra of finite type with a subfield of the field\nof fractions of the algebra. It has a negative answer due to the counterexample\nof Nagata. We show that a subfinite version of Hilbert's 14th problem has a\nconfirmative answer. We then establish a graded analogue of this result, which\npermits to show that the subfiniteness of graded linear series does not depend\non the function field in which we consider it. Finally, we apply the\nsubfiniteness result to the study of geometric and arithmetic graded linear\nseries.\n",
"title": "On subfiniteness of graded linear series"
}
| null | null | null | null | true | null |
17457
| null |
Default
| null | null |
null |
{
"abstract": " We design a stochastic algorithm to train any smooth neural network to\n$\\varepsilon$-approximate local minima, using $O(\\varepsilon^{-3.25})$\nbackpropagations. The best result was essentially $O(\\varepsilon^{-4})$ by SGD.\nMore broadly, it finds $\\varepsilon$-approximate local minima of any smooth\nnonconvex function in rate $O(\\varepsilon^{-3.25})$, with only oracle access to\nstochastic gradients.\n",
"title": "Natasha 2: Faster Non-Convex Optimization Than SGD"
}
| null | null | null | null | true | null |
17458
| null |
Default
| null | null |
null |
{
"abstract": " We find evidence for a strong thermal inversion in the dayside atmosphere of\nthe highly irradiated hot Jupiter WASP-18b (T$_{eq}=2411K$, $M=10.3M_{J}$)\nbased on emission spectroscopy from Hubble Space Telescope secondary eclipse\nobservations and Spitzer eclipse photometry. We demonstrate a lack of water\nvapor in either absorption or emission at 1.4$\\mu$m. However, we infer emission\nat 4.5$\\mu$m and absorption at 1.6$\\mu$m that we attribute to CO, as well as a\nnon-detection of all other relevant species (e.g., TiO, VO). The most probable\natmospheric retrieval solution indicates a C/O ratio of 1 and a high\nmetallicity (C/H=$283^{+395}_{-138}\\times$ solar). The derived composition and\nT/P profile suggest that WASP-18b is the first example of both a planet with a\nnon-oxide driven thermal inversion and a planet with an atmospheric metallicity\ninconsistent with that predicted for Jupiter-mass planets at $>2\\sigma$. Future\nobservations are necessary to confirm the unusual planetary properties implied\nby these results.\n",
"title": "Evidence for a Dayside Thermal Inversion and High Metallicity for the Hot Jupiter WASP-18b"
}
| null | null | null | null | true | null |
17459
| null |
Default
| null | null |
null |
{
"abstract": " The magnetic-field-temperature phase diagram of solid oxygen is investigated\nby the adiabatic magnetocaloric effect (MCE) measurement with pulsed magnetic\nfields. Relatively large temperature decrease with hysteresis is observed at\njust below the $\\beta$-$\\gamma$ and $\\alpha$-$\\beta$ phase transition\ntemperatures owing to the field-induced transitions. The magnetic field\ndependences of these phase boundaries are obtained as\n$T_\\mathrm{\\beta\\gamma}(H)=43.8-1.55\\times10^{-3}H^2$ K and\n$T_\\mathrm{\\alpha\\beta}(H)=23.9-0.73\\times10^{-3}H^2$ K. The magnetic\nClausius-Clapeyron equation quantitatively explains the $H$ dependence of\n$T_\\mathrm{\\beta\\gamma}$, meanwhile, does not $T_\\mathrm{\\alpha\\beta}$. The MCE\ncurve at $T_\\mathrm{\\beta\\gamma}$ is of typical first-order, while the curve at\n$T_\\mathrm{\\alpha\\beta}$ seems to have both characteristics of first- and\nsecond-order transitions. We discuss the order of the $\\alpha$-$\\beta$ phase\ntransition and propose possible reasons for the unusual behavior.\n",
"title": "$α$-$β$ and $β$-$γ$ phase boundaries of solid oxygen observed by adiabatic magnetocaloric effect"
}
| null | null | null | null | true | null |
17460
| null |
Default
| null | null |
null |
{
"abstract": " Extended Kalman filter (EKF) does not guarantee consistent mean and\ncovariance under linearization, even though it is the main framework for\nrobotic localization. While Lie group improves the modeling of the state space\nin localization, the EKF on Lie group still relies on the arbitrary Gaussian\nassumption in face of nonlinear models. We instead use von Mises filter for\norientation estimation together with the conventional Kalman filter for\nposition estimation, and thus we are able to characterize the first two moments\nof the state estimates. Since the proposed algorithm holds a solid\nprobabilistic basis, it is fundamentally relieved from the inconsistency\nproblem. Furthermore, we extend the localization algorithm to fully circular\nrepresentation even for position, which is similar to grid patterns found in\nmammalian brains and in recurrent neural networks. The applicability of the\nproposed algorithms is substantiated not only by strong mathematical foundation\nbut also by the comparison against other common localization methods.\n",
"title": "Localization Algorithm with Circular Representation in 2D and its Similarity to Mammalian Brains"
}
| null | null | null | null | true | null |
17461
| null |
Default
| null | null |
null |
{
"abstract": " We establish new approximation results, in the sense of Lusin, of Sobolev\nfunctions by Lipschitz ones, in some classes of non-doubling metric measure\nstructures. Our proof technique relies upon estimates for heat semigroups and\napplies to Gaussian and $RCD(K, \\infty)$ spaces. As a consequence, we obtain\nquantitative stability for regular Lagrangian flows in Gaussian settings.\n",
"title": "Lusin-type approximation of Sobolev by Lipschitz functions, in Gaussian and $RCD(K,\\infty)$ spaces"
}
| null | null | null | null | true | null |
17462
| null |
Default
| null | null |
null |
{
"abstract": " In this paper, a multi-agent coordination problem with steady-state\nregulation constraints is investigated for a class of nonlinear systems. Unlike\nexisting leader-following coordination formulations, the reference signal is\nnot given by a dynamic autonomous leader but determined as the optimal solution\nof a distributed optimization problem. Furthermore, we consider a global\nconstraint having noisy data observations for the optimization problem, which\nimplies that reference signal is not trivially available with existing\noptimization algorithms. To handle those challenges, we present a\npassivity-based analysis and design approach by using only local objective\nfunction, local data observation and exchanged information from their\nneighbors. The proposed distributed algorithms are shown to achieve the optimal\nsteady-state regulation by rejecting the unknown observation disturbances for\npassive nonlinear agents, which are persuasive in various practical problems.\nApplications and simulation examples are then given to verify the effectiveness\nof our design.\n",
"title": "Distributed Coordination for a Class of Nonlinear Multi-agent Systems with Regulation Constraints"
}
| null | null | null | null | true | null |
17463
| null |
Default
| null | null |
null |
{
"abstract": " For the first time, intermodulation distortion of micro-electromechanical\ncapacitive switches in the actuated state was analyzed both theoretically and\nexperimentally. The distortion, although higher than that of switches in the\nsuspended state, was found to decrease with increasing bias voltage but to\ndepend weakly on modulation frequencies between 55 kHz and 1.1 MHz. This\ndependence could be explained by the orders-of-magnitude increase of the spring\nconstant when the switches were actuated. Additionally, the analysis suggested\nthat increasing the spring constant and decreasing the contact roughness could\nimprove the linearity of actuated switches. These results are critical to\nmicro-electromechanical capacitive switches used in tuners, filters, phase\nshifters, etc. where the linearity of both suspended and actuated states are\ncritical.\n",
"title": "Intermodulation distortion of actuated MEMS capacitive switches"
}
| null | null | null | null | true | null |
17464
| null |
Default
| null | null |
null |
{
"abstract": " InGaAs-based Gate-all-Around (GAA) FETs with moderate to high In content are\nshown experimentally and theoretically to be unsuitable for low-leakage\nadvanced CMOS nodes. The primary cause for this is the large leakage penalty\ninduced by the Parasitic Bipolar Effect (PBE), which is seen to be particularly\ndifficult to remedy in GAA architectures. Experimental evidence of PBE in\nIn70Ga30As GAA FETs is demonstrated, along with a simulation-based analysis of\nthe PBE behavior. The impact of PBE is investigated by simulation for\nalternative device architectures, such as bulk FinFETs and\nFinFETs-on-insulator. PBE is found to be non-negligible in all standard InGaAs\nFET designs. Practical PBE metrics are introduced and the design of a substrate\narchitecture for PBE suppression is elucidated. Finally, it is concluded that\nthe GAA architecture is not suitable for low-leakage InGaAs FETs; a bulk FinFET\nis better suited for the role.\n",
"title": "Parasitic Bipolar Leakage in III-V FETs: Impact of Substrate Architecture"
}
| null | null | null | null | true | null |
17465
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we study the integral of type\n\\[_{\\delta,a}\\Gamma_{\\rho,b}(x)\n=\\Gamma(\\delta,a;\\rho,b)(x)=\\int_{0}^{\\infty}t^{x-1}e^{-\\frac{t^{\\delta}}{a}-\\frac{t^{-\\rho}}{b}}dt.\\]\nDifferent authors called this integral by different names like ultra gamma\nfunction, generalized gamma function, Kratzel integral, inverse Gaussian\nintegral, reaction-rate probability integral, Bessel integral etc. We prove\nseveral identities and recurrence relation of above said integral, we called\nthis integral as Four Parameter Gamma Function. Also we evaluate relation\nbetween Four Parameter Gamma Function, p-k Gamma Function and Classical Gamma\nFunction. With some conditions we can evaluate Four Parameter Gamma Function in\nterm of Hypergeometric function.\n",
"title": "Properties of Ultra Gamma Function"
}
| null | null | null | null | true | null |
17466
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we define the notion of pullback lifting of a lifting crossed\nmodule over a crossed module morphism and interpret this notion in the category\nof group-groupoid actions as pullback action. Moreover, we give a criterion for\nthe lifting of homotopic crossed module morphisms to be homotopic, which will\nbe called homotopy lifting property for crossed module morphisms. Finally, we\ninvestigate some properties of derivations of lifting crossed modules according\nto base crossed module derivations.\n",
"title": "Further remarks on liftings of crossed modules"
}
| null | null | null | null | true | null |
17467
| null |
Default
| null | null |
null |
{
"abstract": " The simplex algorithm for linear programming is based on the fact that any\nlocal optimum with respect to the polyhedral neighborhood is also a global\noptimum. We show that a similar result carries over to submodular maximization.\nIn particular, every local optimum of a constrained monotone submodular\nmaximization problem yields a $1/2$-approximation, and we also present an\nappropriate extension to the non-monotone setting. However, reaching a local\noptimum quickly is a non-trivial task. Moreover, we describe a fast and very\ngeneral local search procedure that applies to a wide range of constraint\nfamilies, and unifies as well as extends previous methods. In our framework, we\nmatch known approximation guarantees while disentangling and simplifying\nprevious approaches. Moreover, despite its generality, we are able to show that\nour local search procedure is slightly faster than previous specialized\nmethods. Furthermore, we resolve an open question on the relation between\nlinear optimization and submodular maximization; namely, whether a linear\noptimization oracle may be enough to obtain strong approximation algorithms for\nsubmodular maximization. We show that this is not the case by providing an\nexample of a constraint family on a ground set of size $n$ for which, if only\ngiven a linear optimization oracle, any algorithm for submodular maximization\nwith a polynomial number of calls to the linear optimization oracle will have\nan approximation ratio of only $O ( \\frac{1}{\\sqrt{n}} \\cdot \\frac{\\log\nn}{\\log\\log n} )$.\n",
"title": "Submodular Maximization through the Lens of Linear Programming"
}
| null | null | null | null | true | null |
17468
| null |
Default
| null | null |
null |
{
"abstract": " In diffusion-based communication, as for molecular systems, the achievable\ndata rate is very low due to the slow nature of diffusion and the existence of\nsevere inter-symbol interference (ISI). Multiple-input multiple-output (MIMO)\ntechnique can be used to improve the data rate. Knowledge of channel impulse\nresponse (CIR) is essential for equalization and detection in MIMO systems.\nThis paper presents a training-based CIR estimation for diffusive MIMO (D-MIMO)\nchannels. Maximum likelihood and least-squares estimators are derived, and the\ntraining sequences are designed to minimize the corresponding Cramér-Rao\nbound. Sub-optimal estimators are compared to Cramér-Rao bound to validate\ntheir performance.\n",
"title": "Channel Estimation for Diffusive MIMO Molecular Communications"
}
| null | null | null | null | true | null |
17469
| null |
Default
| null | null |
null |
{
"abstract": " Modified Hamiltonian Monte Carlo (MHMC) methods combine the ideas behind two\npopular sampling approaches: Hamiltonian Monte Carlo (HMC) and importance\nsampling. As in the HMC case, the bulk of the computational cost of MHMC\nalgorithms lies in the numerical integration of a Hamiltonian system of\ndifferential equations. We suggest novel integrators designed to enhance\naccuracy and sampling performance of MHMC methods. The novel integrators belong\nto families of splitting algorithms and are therefore easily implemented. We\nidentify optimal integrators within the families by minimizing the energy error\nor the average energy error. We derive and discuss in detail the modified\nHamiltonians of the new integrators, as the evaluation of those Hamiltonians is\nkey to the efficiency of the overall algorithms. Numerical experiments show\nthat the use of the new integrators may improve very significantly the sampling\nperformance of MHMC methods, in both statistical and molecular dynamics\nproblems.\n",
"title": "Multi-stage splitting integrators for sampling with modified Hamiltonian Monte Carlo methods"
}
| null | null | null | null | true | null |
17470
| null |
Default
| null | null |
null |
{
"abstract": " This whitepaper proposes the design and adoption of a new generation of\nTensor Processing Unit which has the performance of Google's TPU, yet performs\noperations on wide precision data. The new generation TPU is made possible by\nimplementing arithmetic circuits which compute using a new general purpose,\nfractional arithmetic based on the residue number system.\n",
"title": "Proposal for a High Precision Tensor Processing Unit"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17471
| null |
Validated
| null | null |
null |
{
"abstract": " In this paper, we introduce DICOD, a convolutional sparse coding algorithm\nwhich builds shift invariant representations for long signals. This algorithm\nis designed to run in a distributed setting, with local message passing, making\nit communication efficient. It is based on coordinate descent and uses locally\ngreedy updates which accelerate the resolution compared to greedy coordinate\nselection. We prove the convergence of this algorithm and highlight its\ncomputational speed-up which is super-linear in the number of cores used. We\nalso provide empirical evidence for the acceleration properties of our\nalgorithm compared to state-of-the-art methods.\n",
"title": "DICOD: Distributed Convolutional Sparse Coding"
}
| null | null | null | null | true | null |
17472
| null |
Default
| null | null |
null |
{
"abstract": " We study uniqueness of Dirichlet problems of second order divergence-form\nelliptic systems with transversally independent coefficients on the upper\nhalf-space in absence of regularity of solutions. To this end, we develop a\nsubstitute for the fundamental solution used to invert elliptic operators on\nthe whole space by means of a representation via abstract single layer\npotentials. We also show that such layer potentials are uniquely determined.\n",
"title": "On uniqueness results for Dirichlet problems of elliptic systems without DeGiorgi-Nash-Moser regularity"
}
| null | null |
[
"Mathematics"
] | null | true | null |
17473
| null |
Validated
| null | null |
null |
{
"abstract": " At interfaces between oxide materials, lattice and electronic reconstructions\nalways play important roles in exotic phenomena. In this study, the density\nfunctional theory and maximally localized Wannier functions are employed to\ninvestigate the (LaTiO$_3$)$_n$/(LaVO$_3$)$_n$ magnetic superlattices. The\nelectron transfer from Ti$^{3+}$ to V$^{3+}$ is predicted, which violates the\nintuitive band alignment based on the electronic structures of LaTiO$_3$ and\nLaVO$_3$. Such unconventional charge transfer quenches the magnetism of\nLaTiO$_3$ layer mostly and leads to metal-insulator transition in the $n=1$\nsuperlattice when the stacking orientation is altered. In addition, the\ncompatibility among the polar structure, ferrimagnetism, and metallicity is\npredicted in the $n=2$ superlattice.\n",
"title": "(LaTiO$_3$)$_n$/(LaVO$_3$)$_n$ as a model system for unconventional charge transfer and polar metallicity"
}
| null | null | null | null | true | null |
17474
| null |
Default
| null | null |
null |
{
"abstract": " A mathematical model for emerging contaminants sorption in multispecies\nbiofilms, based on a continuum approach and mass conservation principles is\npresented. Diffusion of contaminants within the biofilm is described using a\ndiffusion-reaction equation. Binding sites formation and occupation are modeled\nby two systems of hyperbolic partial differential equations are mutually\nconnected through the two growth rate terms. The model is completed with a\nsystem of hyperbolic equations governing the microbial species growth within\nthe biofilm; a system of parabolic equations for substrates diffusion and\nreaction and a nonlinear ordinary differential equation describing the free\nboundary evolution. Two real special cases are modelled. The first one\ndescribes the dynamics of a free sorbent component diffusing and reacting in a\nmultispecies biofilm. In the second illustrative case, the fate of two\ndifferent contaminants has been modelled.\n",
"title": "Modeling sorption of emerging contaminants in biofilms"
}
| null | null | null | null | true | null |
17475
| null |
Default
| null | null |
null |
{
"abstract": " Deep generative models based on Generative Adversarial Networks (GANs) have\ndemonstrated impressive sample quality but in order to work they require a\ncareful choice of architecture, parameter initialization, and selection of\nhyper-parameters. This fragility is in part due to a dimensional mismatch or\nnon-overlapping support between the model distribution and the data\ndistribution, causing their density ratio and the associated f-divergence to be\nundefined. We overcome this fundamental limitation and propose a new\nregularization approach with low computational cost that yields a stable GAN\ntraining procedure. We demonstrate the effectiveness of this regularizer across\nseveral architectures trained on common benchmark image generation tasks. Our\nregularization turns GAN models into reliable building blocks for deep\nlearning.\n",
"title": "Stabilizing Training of Generative Adversarial Networks through Regularization"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
17476
| null |
Validated
| null | null |
null |
{
"abstract": " Approximate vanishing ideal, which is a new concept from computer algebra, is\na set of polynomials that almost takes a zero value for a set of given data\npoints. The introduction of approximation to exact vanishing ideal has played a\ncritical role in capturing the nonlinear structures of noisy data by computing\nthe approximate vanishing polynomials. However, approximate vanishing has a\ntheoretical problem, which is giving rise to the spurious vanishing problem\nthat any polynomial turns into an approximate vanishing polynomial by\ncoefficient scaling. In the present paper, we propose a general method that\nenables many basis construction methods to overcome this problem. Furthermore,\na coefficient truncation method is proposed that balances the theoretical\nsoundness and computational cost. The experiments show that the proposed method\novercomes the spurious vanishing problem and significantly increases the\naccuracy of classification.\n",
"title": "Spurious Vanishing Problem in Approximate Vanishing Ideal"
}
| null | null | null | null | true | null |
17477
| null |
Default
| null | null |
null |
{
"abstract": " X-ray Free Electron Lasers (XFELs) have been proven to generate short and\npowerful radiation pulses allowing for a wide class of novel experiments. If an\nXFEL facility supports the generation of two X-ray pulses with different\nwavelengths and controllable delay, the range of possible experiments is\nbroadened even further to include X-ray-pump/X-ray-probe applications. In this\nwork we discuss the possibility of applying a simple and cost-effective method\nfor producing two-color pulses at the SASE3 soft X-ray beamline of the European\nXFEL. The technique is based on the installation of a magnetic chicane in the\nbaseline undulator and can be accomplished in several steps. We discuss the\nscientific interest of this upgrade for the Small Quantum Systems (SQS)\ninstrument, in connection with the high-repetition rate of the European XFEL,\nand we provide start-to-end simulations up to the radiation focus on the\nsample, proving the feasibility of our concept.\n",
"title": "Opportunities for Two-color Experiments at the SASE3 undulator line of the European XFEL"
}
| null | null | null | null | true | null |
17478
| null |
Default
| null | null |
null |
{
"abstract": " Avenues of Majorana bound states (MBSs) have become one of the primary\ndirections towards a possible realization of topological quantum computation.\nFor a Y-junction of Kitaev quantum wires, we numerically investigate the\nbraiding of MBSs while considering the full quasi-particle background. The two\ncentral sources of braiding errors are found to be the fidelity loss due to the\nincomplete adiabaticity of the braiding operation as well as the hybridization\nof the MBS. The explicit extraction of the braiding phase in the low-energy\nMajorana sector from the full many-particle Hilbert space allows us to analyze\nthe breakdown of the independent-particle picture of Majorana braiding.\nFurthermore, we find nearest-neighbor interactions to significantly affect the\nbraiding performance to the better or worse, depending on the sign and\nmagnitude of the coupling.\n",
"title": "Braiding errors in interacting Majorana quantum wires"
}
| null | null | null | null | true | null |
17479
| null |
Default
| null | null |
null |
{
"abstract": " Research is a tertiary priority in the EHR, where the priorities are patient\ncare and billing. Because of this, the data is not standardized or formatted in\na manner easily adapted to machine learning approaches. Data may be missing for\na large variety of reasons ranging from individual input styles to differences\nin clinical decision making, for example, which lab tests to issue. Few\npatients are annotated at a research quality, limiting sample size and\npresenting a moving gold standard. Patient progression over time is key to\nunderstanding many diseases but many machine learning algorithms require a\nsnapshot, at a single time point, to create a usable vector form. Furthermore,\nalgorithms that produce black box results do not provide the interpretability\nrequired for clinical adoption. This chapter discusses these challenges and\nothers in applying machine learning techniques to the structured EHR (i.e.\nPatient Demographics, Family History, Medication Information, Vital Signs,\nLaboratory Tests, Genetic Testing). It does not cover feature extraction from\nadditional sources such as imaging data or free text patient notes but the\napproaches discussed can include features extracted from these sources.\n",
"title": "Machine Learning for Structured Clinical Data"
}
| null | null | null | null | true | null |
17480
| null |
Default
| null | null |
null |
{
"abstract": " We show that whereas spin-1/2 one-dimensional U(1) quantum-link models (QLMs)\nare topologically trivial, when implemented in ladder-like lattices these\nmodels may present an intriguing ground-state phase diagram, which includes a\nsymmetry protected topological (SPT) phase that may be readily revealed by\nanalyzing long-range string spin correlations along the ladder legs. We propose\na simple scheme for the realization of spin-1/2 U(1) QLMs based on\nsingle-component fermions loaded in an optical lattice with s- and p-bands,\nshowing that the SPT phase may be experimentally realized by adiabatic\npreparation.\n",
"title": "Hidden order and symmetry protected topological states in quantum link ladders"
}
| null | null | null | null | true | null |
17481
| null |
Default
| null | null |
null |
{
"abstract": " In this paper two portfolio choice models are studied: a purely possibilistic\nmodel, in which the return of a risky asset is a fuzzy number, and a mixed\nmodel in which a probabilistic background risk is added. For the two models an\napproximate formula of the optimal allocation is computed, with respect to the\npossibilistic moments associated with fuzzy numbers and the indicators of the\ninvestor risk preferences (risk aversion, prudence).\n",
"title": "The effect of prudence on the optimal allocation in possibilistic and mixed models"
}
| null | null | null | null | true | null |
17482
| null |
Default
| null | null |
null |
{
"abstract": " We study some basic properties of the class of universal operators on Hilbert\nspace, and provide new examples of universal operators and universal pairs.\n",
"title": "On universal operators and universal pairs"
}
| null | null | null | null | true | null |
17483
| null |
Default
| null | null |
null |
{
"abstract": " To realize and test advanced accelerator concepts and hardware, a beamline is\nbeing reconfigured in the Linac Extension Area (LEA) of APS linac. A\nphoto-cathode RF gun installed at the beginning of the APS linac will provide a\nlow emittance electron beam into the LEA beamline. The thermionic RF gun beam\nfor the APS storage ring, and the photo-cathode RF gun beam for LEA beamline\nwill be accelerated through the linac in an interleaved fashion. In this paper,\nthe design studies for interleaving lattice realization in APS linac is\ndescribed with initial experiment result\n",
"title": "Interleaving Lattice for the APS Linac"
}
| null | null | null | null | true | null |
17484
| null |
Default
| null | null |
null |
{
"abstract": " We investigated the magnetic behavior of metal hydrides FeH$_{x}$, CoH$_{x}$\nand NiH$_{x}$ for several concentrations of hydrogen ($x$) by using Density\nFunctional Theory calculations. Several structural phases of the metallic host:\nbcc ($\\alpha$), fcc ($\\gamma$), hcp ($\\varepsilon$), dhcp ($\\varepsilon'$),\ntetragonal structure for FeH$_{x}$ and $\\varepsilon$-$\\gamma$ phases for\nCoH$_{x}$, were studied. We found that for CoH$_{x}$ and NiH$_{x}$ the magnetic\nmoment ($m$) decreases regardless the concentration $x$. However, for FeH$_{x}$\nsystems, $m$ increases or decreases depending on the variation in $x$. In order\nto find a general trend for these changes of $m$ in magnetic metal hydrides, we\ncompare our results with the Slater-Pauling curve for ferromagnetic metallic\nbinary alloys. It is found that the $m$ of metal hydrides made of Fe, Co and Ni\nfits the shape of the Slater-Pauling curve as a function of $x$. Our results\nindicate that there are two main effects that determine the $m$ value due to\nhydrogenation: an increase of volume causes $m$ to increase, and the addition\nof an extra electron to the metal always causes it to decrease. We discuss\nthese behaviors in detail.\n",
"title": "\\textit{Ab Initio} Study of the Magnetic Behavior of Metal Hydrides: A Comparison with the Slater-Pauling Curve"
}
| null | null | null | null | true | null |
17485
| null |
Default
| null | null |
null |
{
"abstract": " Cloud computing helps reduce costs, increase business agility and deploy\nsolutions with a high return on investment for many types of applications.\nHowever, data security is of premium importance to many users and often\nrestrains their adoption of cloud technologies. Various approaches, i.e., data\nencryption, anonymization, replication and verification, help enforce different\nfacets of data security. Secret sharing is a particularly interesting\ncryptographic technique. Its most advanced variants indeed simultaneously\nenforce data privacy, availability and integrity, while allowing computation on\nencrypted data. The aim of this paper is thus to wholly survey secret sharing\nschemes with respect to data security, data access and costs in the\npay-as-you-go paradigm.\n",
"title": "Secret Sharing for Cloud Data Security"
}
| null | null | null | null | true | null |
17486
| null |
Default
| null | null |
null |
{
"abstract": " Password security can no longer provide enough security in the area of remote\nuser authentication. Considering this security drawback, researchers are trying\nto find solution with multifactor remote user authentication system. Recently,\nthree factor remote user authentication using biometric and smart card has\ndrawn a considerable attention of the researchers. However, most of the current\nproposed schemes have security flaws. They are vulnerable to attacks like user\nimpersonation attack, server masquerading attack, password guessing attack,\ninsider attack, denial of service attack, forgery attack, etc. Also, most of\nthem are unable to provide mutual authentication, session key agreement and\npassword, or smart card recovery system. Considering these drawbacks, we\npropose a secure three factor user authentication scheme using biometric and\nsmart card. Through security analysis, we show that our proposed scheme can\novercome drawbacks of existing systems and ensure high security in remote user\nauthentication.\n",
"title": "Design and Analysis of a Secure Three Factor User Authentication Scheme Using Biometric and Smart Card"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17487
| null |
Validated
| null | null |
null |
{
"abstract": " This paper presents a practical approach towards implementing pathfinding\nalgorithms on real-world and low-cost non- commercial hardware platforms. While\nusing robotics simulation platforms as a test-bed for our algorithms we easily\noverlook real- world exogenous problems that are developed by external factors.\nSuch problems involve robot wheel slips, asynchronous motors, abnormal sensory\ndata or unstable power sources. The real-world dynamics tend to be very painful\neven for executing simple algorithms like a Wavefront planner or A-star search.\nThis paper addresses designing techniques that tend to be robust as well as\nreusable for any hardware platforms; covering problems like controlling\nasynchronous drives, odometry offset issues and handling abnormal sensory\nfeedback. The algorithm implementation medium and hardware design tools have\nbeen kept general in order to present our work as a serving platform for future\nresearchers and robotics enthusiast working in the field of path planning\nrobotics.\n",
"title": "Real-World Modeling of a Pathfinding Robot Using Robot Operating System (ROS)"
}
| null | null | null | null | true | null |
17488
| null |
Default
| null | null |
null |
{
"abstract": " Recent advances in deep learning, especially deep convolutional neural\nnetworks (CNNs), have led to significant improvement over previous semantic\nsegmentation systems. Here we show how to improve pixel-wise semantic\nsegmentation by manipulating convolution-related operations that are of both\ntheoretical and practical value. First, we design dense upsampling convolution\n(DUC) to generate pixel-level prediction, which is able to capture and decode\nmore detailed information that is generally missing in bilinear upsampling.\nSecond, we propose a hybrid dilated convolution (HDC) framework in the encoding\nphase. This framework 1) effectively enlarges the receptive fields (RF) of the\nnetwork to aggregate global information; 2) alleviates what we call the\n\"gridding issue\" caused by the standard dilated convolution operation. We\nevaluate our approaches thoroughly on the Cityscapes dataset, and achieve a\nstate-of-art result of 80.1% mIOU in the test set at the time of submission. We\nalso have achieved state-of-the-art overall on the KITTI road estimation\nbenchmark and the PASCAL VOC2012 segmentation task. Our source code can be\nfound at this https URL .\n",
"title": "Understanding Convolution for Semantic Segmentation"
}
| null | null | null | null | true | null |
17489
| null |
Default
| null | null |
null |
{
"abstract": " The mechanical failure of amorphous media is a ubiquitous phenomenon from\nmaterial engineering to geology. It has been noticed for a long time that the\nphenomenon is \"scale-free\", indicating some type of criticality. In spite of\nattempts to invoke \"Self-Organized Criticality\", the physical origin of this\ncriticality, and also its universal nature, being quite insensitive to the\nnature of microscopic interactions, remained elusive. Recently we proposed that\nthe precise nature of this critical behavior is manifested by a spinodal point\nof a thermodynamic phase transition. Moreover, at the spinodal point there\nexists a divergent correlation length which is associated with the\nsystem-spanning instabilities (known also as shear bands) which are typical to\nthe mechanical yield. Demonstrating this requires the introduction of an \"order\nparameter\" that is suitable for distinguishing between disordered amorphous\nsystems, and an associated correlation function, suitable for picking up the\ngrowing correlation length. The theory, the order parameter, and the\ncorrelation functions used are universal in nature and can be applied to any\namorphous solid that undergoes mechanical yield. Critical exponents for the\ncorrelation length divergence and the system size dependence are estimated. The\nphenomenon is seen at its sharpest in athermal systems, as is explained below;\nin this paper we extend the discussion also to thermal systems, showing that at\nsufficiently high temperatures the spinodal phenomenon is destroyed by thermal\nfluctuations.\n",
"title": "Mechanical Failure in Amorphous Solids: Scale Free Spinodal Criticality"
}
| null | null |
[
"Physics"
] | null | true | null |
17490
| null |
Validated
| null | null |
null |
{
"abstract": " The validity of the strong law of large numbers for multiple sums $S_n$ of\nindependent identically distributed random variables $Z_k$, $k\\leq n$, with\n$r$-dimensional indices is equivalent to the integrability of\n$|Z|(\\log^+|Z|)^{r-1}$, where $Z$ is the typical summand. We consider the\nstrong law of large numbers for more general normalisations, without assuming\nthat the summands $Z_k$ are identically distributed, and prove a multiple sum\ngeneralisation of the Brunk--Prohorov strong law of large numbers. In the case\nof identical finite moments of irder $2q$ with integer $q\\geq1$, we show that\nthe strong law of large numbers holds with the normalisation $\\|n_1\\cdots\nn_r\\|^{1/2}(\\log n_1\\cdots\\log n_r)^{1/(2q)+\\varepsilon}$ for any\n$\\varepsilon>0$. The obtained results are also formulated in the setting of\nergodic theorems for random measures, in particular those generated by marked\npoint processes.\n",
"title": "Moment conditions in strong laws of large numbers for multiple sums and random measures"
}
| null | null |
[
"Mathematics"
] | null | true | null |
17491
| null |
Validated
| null | null |
null |
{
"abstract": " In this letter we study the mean sizes of Halpha clumps in turbulent disk\ngalaxies relative to kinematics, gas fractions, and Toomre Q. We use 100~pc\nresolution HST images, IFU kinematics, and gas fractions of a sample of rare,\nnearby turbulent disks with properties closely matched to z~1.5-2 main-sequence\ngalaxies (the DYNAMO sample). We find linear correlations of normalized mean\nclump sizes with both the gas fraction and the velocity dispersion-to-rotation\nvelocity ratio of the host galaxy. We show that these correlations are\nconsistent with predictions derived from a model of instabilities in a\nself-gravitating disk (the so-called \"violent disk instability model\"). We also\nobserve, using a two-fluid model for Q, a correlation between the size of\nclumps and self-gravity driven unstable regions. These results are most\nconsistent with the hypothesis that massive star forming clumps in turbulent\ndisks are the result of instabilities in self-gravitating gas-rich disks, and\ntherefore provide a direct connection between resolved clump sizes and this in\nsitu mechanism.\n",
"title": "Connecting Clump Sizes in Turbulent Disk Galaxies to Instability Theory"
}
| null | null | null | null | true | null |
17492
| null |
Default
| null | null |
null |
{
"abstract": " Motivated by a host of empirical evidences revealing the bursty character of\nhuman dynamics, we develop a model of human activity based on successive\nswitching between an hesitation state and a decision-realization state, with\nresidency times in the hesitation state distributed according to a heavy-tailed\nPareto distribution. This model is particularly reminiscent of an individual\nstrolling through a randomly distributed human crowd. Using a stochastic model\nbased on the concept of anomalous and non-Markovian Lévy walk, we show\nexactly that successive decision-making processes drastically slow down the\nprogression of an individual faced with randomly distributed obstacles.\nSpecifically, we prove exactly that the average displacement exhibits a\nsublinear scaling with time that finds its origins in: (i) the intrinsically\nnon-Markovian character of human activity, and (ii) the power law distribution\nof hesitation times.\n",
"title": "Anomalous slowing down of individual human activity due to successive decision-making processes"
}
| null | null | null | null | true | null |
17493
| null |
Default
| null | null |
null |
{
"abstract": " We consider the spectral Dirichlet problem for the Laplace operator in the\nplane $\\Omega^{\\circ}$ with double-periodic perforation but also in the domain\n$\\Omega^{\\bullet}$ with a semi-infinite foreign inclusion so that the\nFloquet-Bloch technique and the Gelfand transform do not apply directly. We\ndescribe waves which are localized near the inclusion and propagate along it.\nWe give a formulation of the problem with radiation conditions that provides a\nFredholm operator of index zero. The main conclusion concerns the spectra\n$\\sigma^{\\circ}$ and $\\sigma^{\\bullet}$ of the problems in $\\Omega^{\\circ}$ and\n$\\Omega^{\\bullet},$ namely we present a concrete geometry which supports the\nrelation $\\sigma^{\\circ}\\varsubsetneqq\\sigma^{\\bullet}$ due to a new non-empty\nspectral band caused by the semi-infinite inclusion called an open waveguide in\nthe double-periodic medium.\n",
"title": "The spectrum, radiation conditions and the Fredholm property for the Dirichlet Laplacian in a perforated plane with semi-infinite inclusions"
}
| null | null | null | null | true | null |
17494
| null |
Default
| null | null |
null |
{
"abstract": " We examine the meaning and the complexity of probabilistic logic programs\nthat consist of a set of rules and a set of independent probabilistic facts\n(that is, programs based on Sato's distribution semantics). We focus on two\nsemantics, respectively based on stable and on well-founded models. We show\nthat the semantics based on stable models (referred to as the \"credal\nsemantics\") produces sets of probability models that dominate infinitely\nmonotone Choquet capacities, we describe several useful consequences of this\nresult. We then examine the complexity of inference with probabilistic logic\nprograms. We distinguish between the complexity of inference when a\nprobabilistic program and a query are given (the inferential complexity), and\nthe complexity of inference when the probabilistic program is fixed and the\nquery is given (the query complexity, akin to data complexity as used in\ndatabase theory). We obtain results on the inferential and query complexity for\nacyclic, stratified, and cyclic propositional and relational programs,\ncomplexity reaches various levels of the counting hierarchy and even\nexponential levels.\n",
"title": "On the Semantics and Complexity of Probabilistic Logic Programs"
}
| null | null |
[
"Computer Science"
] | null | true | null |
17495
| null |
Validated
| null | null |
null |
{
"abstract": " Recently, Thas et al. (2012) introduced a new statistical model for the\nprobability index. This index is defined as $P(Y \\leq Y^*|X, X^*)$ where Y and\nY* are independent random response variables associated with covariates X and\nX* [...] Crucially to estimate the parameters of the model, a set of\npseudo-observations is constructed. For a sample size n, a total of $n(n-1)/2$\npairwise comparisons between observations is considered. Consequently for large\nsample sizes, it becomes computationally infeasible or even impossible to fit\nthe model as the set of pseudo-observations increases nearly quadratically. In\nthis dissertation, we provide two solutions to fit a probabilistic index model.\nThe first algorithm consists of splitting the entire data set into unique\npartitions. On each of these, we fit the model and then aggregate the\nestimates. A second algorithm is a subsampling scheme in which we select $K <<\nn$ observations without replacement and after B iterations aggregate the\nestimates. In Monte Carlo simulations, we show how the partitioning algorithm\noutperforms the latter [...] We illustrate the partitioning algorithm and the\ninterpretation of the probabilistic index model on a real data set (Przybylski\nand Weinstein, 2017) of n = 116,630 where we compare it against the ordinary\nleast squares method. By modelling the probabilistic index, we give an\nintuitive and meaningful quantification of the effect of the time adolescents\nspend using digital devices such as smartphones on self-reported mental\nwell-being. We show how moderate usage is associated with an increased\nprobability of reporting a higher mental well-being compared to random\nadolescents who do not use a smartphone. On the other hand, adolescents who\nexcessively use their smartphone are associated with a higher probability of\nreporting a lower mental well-being than randomly chosen peers who do not use a\nsmartphone.[...]\n",
"title": "Fitting Probabilistic Index Models on Large Datasets"
}
| null | null | null | null | true | null |
17496
| null |
Default
| null | null |
null |
{
"abstract": " We present the strongest constraints to date on anisotropies of CMB\npolarization rotation derived from $150$ GHz data taken by the BICEP2 & Keck\nArray CMB experiments up to and including the 2014 observing season (BK14). The\ndefinition of polarization angle in BK14 maps has gone through self-calibration\nin which the overall angle is adjusted to minimize the observed $TB$ and $EB$\npower spectra. After this procedure, the $QU$ maps lose sensitivity to a\nuniform polarization rotation but are still sensitive to anisotropies of\npolarization rotation. This analysis places constraints on the anisotropies of\npolarization rotation, which could be generated by CMB photons interacting with\naxion-like pseudoscalar fields or Faraday rotation induced by primordial\nmagnetic fields. The sensitivity of BK14 maps ($\\sim 3\\mu$K-arcmin) makes it\npossible to reconstruct anisotropies of polarization rotation angle and measure\ntheir angular power spectrum much more precisely than previous attempts. Our\ndata are found to be consistent with no polarization rotation anisotropies,\nimproving the upper bound on the amplitude of the rotation angle spectrum by\nroughly an order of magnitude compared to the previous best constraints. Our\nresults lead to an order of magnitude better constraint on the coupling\nconstant of the Chern-Simons electromagnetic term $f_a \\geq 1.7\\times\n10^2\\times (H_I/2\\pi)$ ($2\\sigma$) than the constraint derived from uniform\nrotation, where $H_I$ is the inflationary Hubble scale. The upper bound on the\namplitude of the primordial magnetic fields is 30nG ($2\\sigma$) from the\npolarization rotation anisotropies.\n",
"title": "BICEP2 / Keck Array IX: New Bounds on Anisotropies of CMB Polarization Rotation and Implications for Axion-Like Particles and Primordial Magnetic Fields"
}
| null | null |
[
"Physics"
] | null | true | null |
17497
| null |
Validated
| null | null |
null |
{
"abstract": " In this paper we introduce a system for unsupervised object discovery and\nsegmentation of RGBD-images. The system models the sensor noise directly from\ndata, allowing accurate segmentation without sensor specific hand tuning of\nmeasurement noise models making use of the recently introduced Statistical\nInlier Estimation (SIE) method. Through a fully probabilistic formulation, the\nsystem is able to apply probabilistic inference, enabling reliable segmentation\nin previously challenging scenarios. In addition, we introduce new methods for\nfiltering out false positives, significantly improving the signal to noise\nratio. We show that the system significantly outperform state-of-the-art in on\na challenging real-world dataset.\n",
"title": "Unsupervised Object Discovery and Segmentation of RGBD-images"
}
| null | null | null | null | true | null |
17498
| null |
Default
| null | null |
null |
{
"abstract": " One of the most significant challenges involved in efforts to understand the\neffects of repeated earthquake cycle activity are the computational costs of\nlarge-scale viscoelastic earthquake cycle models. Computationally intensive\nviscoelastic codes must be evaluated thousands of times and locations, and as a\nresult, studies tend to adopt a few fixed rheological structures and model\ngeometries, and examine the predicted time-dependent deformation over short\n(<10 yr) time periods at a given depth after a large earthquake. Training a\ndeep neural network to learn a computationally efficient representation of\nviscoelastic solutions, at any time, location, and for a large range of\nrheological structures, allows these calculations to be done quickly and\nreliably, with high spatial and temporal resolution. We demonstrate that this\nmachine learning approach accelerates viscoelastic calculations by more than\n50,000%. This magnitude of acceleration will enable the modeling of\ngeometrically complex faults over thousands of earthquake cycles across wider\nranges of model parameters and at larger spatial and temporal scales than have\nbeen previously possible.\n",
"title": "Enabling large-scale viscoelastic calculations via neural network acceleration"
}
| null | null | null | null | true | null |
17499
| null |
Default
| null | null |
null |
{
"abstract": " This paper examines the speaker identification potential of breath sounds in\ncontinuous speech. Speech is largely produced during exhalation. In order to\nreplenish air in the lungs, speakers must periodically inhale. When inhalation\noccurs in the midst of continuous speech, it is generally through the mouth.\nIntra-speech breathing behavior has been the subject of much study, including\nthe patterns, cadence, and variations in energy levels. However, an often\nignored characteristic is the {\\em sound} produced during the inhalation phase\nof this cycle. Intra-speech inhalation is rapid and energetic, performed with\nopen mouth and glottis, effectively exposing the entire vocal tract to enable\nmaximum intake of air. This results in vocal tract resonances evoked by\nturbulence that are characteristic of the speaker's speech-producing apparatus.\nConsequently, the sounds of inhalation are expected to carry information about\nthe speaker's identity. Moreover, unlike other spoken sounds which are subject\nto active control, inhalation sounds are generally more natural and less\naffected by voluntary influences. The goal of this paper is to demonstrate that\nbreath sounds are indeed bio-signatures that can be used to identify speakers.\nWe show that these sounds by themselves can yield remarkably accurate speaker\nrecognition with appropriate feature representations and classification\nframeworks.\n",
"title": "Speaker identification from the sound of the human breath"
}
| null | null | null | null | true | null |
17500
| null |
Default
| null | null |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.