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": " In this work, we report X-ray photoelectron (XPS) and valence band (VB)\nspectroscopy measurements of surfactant-free silver nanoparticles and\nsilver/linear carbon chains (Ag@LCC) structures prepared by pulse laser\nablation (PLA) in water. Our measurements demonstrate significant oxidation\nonly on the surfaces of the silver nanoparticles with many covalent\ncarbon-silver bonds but only negligible traces of carbon-oxygen bonds.\nTheoretical modeling also provides evidence of the formation of robust\ncarbon-silver bonds between linear carbon chains and pure and partially\noxidized silver surfaces. A comparison of theoretical and experimental\nelectronic structures also provides evidence of the presence of non-oxidized\nlinear carbon chains on silver surfaces. To evaluate the chemical stability, we\ninvestigated the energetics of the physical adsorption of oxidative species\n(water and oxygen) and found that this adsorption is much preferrable on\noxidized or pristine silver surfaces than the adsorption of linear carbon\nchains, which makes the initial step in the oxidation of LCC energetically\nunfavorable.\n",
"title": "Atomic and electronic structures of stable linear carbon chains on Ag-nanoparticles"
}
| null | null | null | null | true | null |
13701
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we consider a degenerate pseudoparabolic equation for the\nwetting saturation of an unsaturated two-phase flow in porous media with\ndynamic capillary pressure-saturation relationship where the relaxation\nparameter depends on the saturation. Following the approach given in [12] the\nexistence of a weak solution is proved using Galerkin approximation and\nregularization techniques. A priori estimates needed for passing to the limit\nwhen the regularization parameter goes to zero are obtained by using\nappropriate test-functions, motivated by the fact that considered PDE allows a\nnatural generalization of the classical Kullback entropy. Finally, a special\ncare was given in obtaining an estimate of the mixed derivative term by\ncombining the information from the capillary pressure with obtained a priori\nestimates on the saturation.\n",
"title": "The unsaturated flow in porous media with dynamic capillary pressure"
}
| null | null | null | null | true | null |
13702
| null |
Default
| null | null |
null |
{
"abstract": " We use the ab initio Bethe Ansatz dynamics to predict the dissociation of\none-dimensional cold-atom breathers that are created by a quench from a\nfundamental soliton. We find that the dissociation is a robust quantum\nmany-body effect, while in the mean-field (MF) limit the dissociation is\nforbidden by the integrability of the underlying nonlinear Schrödinger\nequation. The analysis demonstrates the possibility to observe quantum\nmany-body effects without leaving the MF range of experimental parameters. We\nfind that the dissociation time is of the order of a few seconds for a typical\natomic-soliton setting.\n",
"title": "Dissociation of one-dimensional matter-wave breathers due to quantum many-body effects"
}
| null | null | null | null | true | null |
13703
| null |
Default
| null | null |
null |
{
"abstract": " The human eye appears to be using a low number of sensors for image\ncapturing. Furthermore, regarding the physical dimensions of\ncones-photoreceptors responsible for the sharp central vision-, we may realize\nthat these sensors are of a relatively small size and area. Nonetheless, the\neye is capable to obtain high resolution images due to visual hyperacuity and\npresents an impressive sensitivity and dynamic range when set against\nconventional digital cameras of similar characteristics. This article is based\non the hypothesis that the human eye may be benefiting from diffraction to\nimprove both image resolution and acquisition process. The developed method\nintends to explain and simulate using MATLAB software the visual hyperacuity:\nthe introduction of a controlled diffraction pattern at an initial stage,\nenables the use of a reduced number of sensors for capturing the image and\nmakes possible a subsequent processing to improve the final image resolution.\nThe results have been compared with the outcome of an equivalent system but in\nabsence of diffraction, achieving promising results. The main conclusion of\nthis work is that diffraction could be helpful for capturing images or signals\nwhen a small number of sensors available, which is far from being a\nresolution-limiting factor.\n",
"title": "Human Eye Visual Hyperacuity: A New Paradigm for Sensing?"
}
| null | null | null | null | true | null |
13704
| null |
Default
| null | null |
null |
{
"abstract": " We study the problem of matrix estimation and matrix completion under a\ngeneral framework. This framework includes several important models as special\ncases such as the gaussian mixture model, mixed membership model, bi-clustering\nmodel and dictionary learning. We consider the optimal convergence rates in a\nminimax sense for estimation of the signal matrix under the Frobenius norm and\nunder the spectral norm. As a consequence of our general result we obtain\nminimax optimal rates of convergence for various special models.\n",
"title": "Structured Matrix Estimation and Completion"
}
| null | null | null | null | true | null |
13705
| null |
Default
| null | null |
null |
{
"abstract": " We are developing position sensitive silicon detectors (PSD) which have an\nelectrode at each of four corners so that the incident position of a charged\nparticle can be obtained using signals from the electrodes. It is expected that\nthe position resolution the electromagnetic calorimeter (ECAL) of the ILD\ndetector will be improved by introducing PSD into the detection layers. In this\nstudy, we irradiated collimated laser beams to the surface of the PSD, varying\nthe incident position. We found that the incident position can be well\nreconstructed from the signals if high resistance is implemented in the p+\nlayer. We also tried to observe the signal of particles by placing a radiative\nsource on the PSD sensor.\n",
"title": "Performance evaluation of PSD for silicon ECAL"
}
| null | null | null | null | true | null |
13706
| null |
Default
| null | null |
null |
{
"abstract": " We revisit the question of reducing online learning to approximate\noptimization of the offline problem. In this setting, we give two algorithms\nwith near-optimal performance in the full information setting: they guarantee\noptimal regret and require only poly-logarithmically many calls to the\napproximation oracle per iteration. Furthermore, these algorithms apply to the\nmore general improper learning problems. In the bandit setting, our algorithm\nalso significantly improves the best previously known oracle complexity while\nmaintaining the same regret.\n",
"title": "Online Improper Learning with an Approximation Oracle"
}
| null | null |
[
"Statistics"
] | null | true | null |
13707
| null |
Validated
| null | null |
null |
{
"abstract": " Recently, Macdonald et. al. showed that many algorithmic problems for\nfinitely generated nilpotent groups including computation of normal forms, the\nsubgroup membership problem, the conjugacy problem, and computation of subgroup\npresentations can be done in Logspace. Here we follow their approach and show\nthat all these problems are complete for the uniform circuit class TC^0 -\nuniformly for all r-generated nilpotent groups of class at most c for fixed r\nand c. In order to solve these problems in TC^0, we show that the unary version\nof the extended gcd problem (compute greatest common divisors and express them\nas linear combinations) is in TC^0. Moreover, if we allow a certain binary\nrepresentation of the inputs, then the word problem and computation of normal\nforms is still in uniform TC^0, while all the other problems we examine are\nshown to be TC^0-Turing reducible to the binary extended gcd problem.\n",
"title": "TC^0 circuits for algorithmic problems in nilpotent groups"
}
| null | null | null | null | true | null |
13708
| null |
Default
| null | null |
null |
{
"abstract": " Let $G$ be a finite solvable or symmetric group and let $B$ be a $2$-block of\n$G$. We construct a canonical correspondence between the irreducible characters\nof height zero in $B$ and those in its Brauer first main correspondent. For\nsymmetric groups our bijection is compatible with restriction of characters.\n",
"title": "Alperin-McKay natural correspondences in solvable and symmetric groups for the prime $p=2$"
}
| null | null | null | null | true | null |
13709
| null |
Default
| null | null |
null |
{
"abstract": " According to a well-known principle of thermodynamics, the transfer of heat\nbetween two bodies is reversible when their temperatures are infinitesimally\nclose. As we demonstrate, a little-known alternative exists: two bodies with\ntemperatures different by an arbitrary amount can completely exchange their\ntemperatures in a reversible way if split into infinitesimal parts that are\nbrought into thermal contact sequentially.\n",
"title": "Reversible temperature exchange upon thermal contact"
}
| null | null | null | null | true | null |
13710
| null |
Default
| null | null |
null |
{
"abstract": " Almost all parameterizations of turbulence in NWP models and GCM make the\nassumption of equality of exchange coefficients for heat $K_h$ and water $K_w$.\nHowever, large uncertainties exists in old papers published in the 1950s, 1960s\nand 1970s, where the turbulent Lewis number Le_t $= K_h / K_w$ have been\nevaluated from observations and then set to Le_t$=1$.\nThe aim of this note is: 1) to trust the recommendations of Richardson\n(1919), who suggested to use the moist-air entropy as a variable on which the\nturbulence is acting; 2) to compute a new exchange coefficients $K_s$ for the\nmoist-air entropy; 3) to determine the values of the new entropy-Lewis number\nLe_ts $= K_s / K_w$ from observations (Météopole-Flux and Cabauw masts) and\nfrom LES and SCM outputs for the IHOP case (Couvreux et al., 2005).\nIt is shown that values of Le_ts significantly different from $1$ are\nfrequently observed and may have large consequences on the way the turbulence\nfluxes are computed in NWP models and GCMs.\n",
"title": "On consequences of measurements of turbulent Lewis number from observations"
}
| null | null | null | null | true | null |
13711
| null |
Default
| null | null |
null |
{
"abstract": " Recent advancements in complex network analysis are encouraging and may\nprovide useful insights when applied in software engineering domain, revealing\nproperties and structures that cannot be captured by traditional metrics. In\nthis paper, we analyzed the topological properties of Hibernate library, a\nwell-known Java-based software through the extraction of its static call graph.\nThe results reveal a complex network with small-world and scale-free\ncharacteristics while displaying a strong propensity on forming communities.\n",
"title": "Complex Networks Analysis for Software Architecture: an Hibernate Call Graph Study"
}
| null | null | null | null | true | null |
13712
| null |
Default
| null | null |
null |
{
"abstract": " In several experimental reports on nonconvex optimization problems in machine\nlearning, stochastic gradient descent (SGD) was observed to prefer minimizers\nwith flat basins in comparison to more deterministic methods, yet there is very\nlittle rigorous understanding of this phenomenon. In fact, the lack of such\nwork has led to an unverified, but widely-accepted stochastic mechanism\ndescribing why SGD prefers flatter minimizers to sharper minimizers. However,\nas we demonstrate, the stochastic mechanism fails to explain this phenomenon.\nHere, we propose an alternative deterministic mechanism that can accurately\nexplain why SGD prefers flatter minimizers to sharper minimizers. We derive\nthis mechanism based on a detailed analysis of a generic stochastic quadratic\nproblem, which generalizes known results for classical gradient descent.\nFinally, we verify the predictions of our deterministic mechanism on two\nnonconvex problems.\n",
"title": "The Impact of Local Geometry and Batch Size on Stochastic Gradient Descent for Nonconvex Problems"
}
| null | null | null | null | true | null |
13713
| null |
Default
| null | null |
null |
{
"abstract": " From a modern perspective cosmology is a historical science in so far that it\ndeals with the development of the universe since its origin some 14 billion\nyears ago. The origin itself may not be subject to scientific analysis and\nexplanation. Nonetheless, there are theories that claim to explain the ultimate\norigin or \"creation\" of the universe. As shown by the history of cosmological\nthought, the very concept of \"origin\" is problematic and can be understood in\ndifferent ways. While it is normally understood as a temporal concept, cosmic\norigin is not temporal by necessity. The universe can be assigned an origin\neven though it has no definite age. In order to clarify the question a view of\nearlier ideas will be helpful, these ideas coming not only from astronomy but\nalso from philosophy and theology.\n",
"title": "Cosmology and the Origin of the Universe: Historical and Conceptual Perspectives"
}
| null | null | null | null | true | null |
13714
| null |
Default
| null | null |
null |
{
"abstract": " We give a classification of semisimple and separable algebras in a\nmulti-fusion category over an arbitrary field in analogy to Wedderben-Artin\ntheorem in classical algebras. It turns out that, if the multi-fusion category\nadmits a semisimple Drinfeld center, the only obstruction to the separability\nof a semisimple algebra arises from inseparable field extensions as in\nclassical algebras. Among others, we show that a division algebra is separable\nif and only if it has a nonvanishing dimension.\n",
"title": "Semisimple and separable algebras in multi-fusion categories"
}
| null | null |
[
"Mathematics"
] | null | true | null |
13715
| null |
Validated
| null | null |
null |
{
"abstract": " We study the equivalence of microcanonical and canonical ensembles in\ncontinuous systems, in the sense of the convergence of the corresponding Gibbs\nmeasures. This is obtained by proving a local central limit theorem and a local\nlarge deviations principle. As an application we prove a formula due to\nLebowitz-Percus-Verlet. It gives mean square fluctuations of an extensive\nobservable, like the kinetic energy, in a classical micro canonical ensemble at\nfixed energy.\n",
"title": "Ensemble dependence of fluctuations and the canonical/micro-canonical equivalence of ensembles"
}
| null | null | null | null | true | null |
13716
| null |
Default
| null | null |
null |
{
"abstract": " This paper considers the problem of positive semidefinite factorization (PSD\nfactorization), a generalization of exact nonnegative matrix factorization.\nGiven an $m$-by-$n$ nonnegative matrix $X$ and an integer $k$, the PSD\nfactorization problem consists in finding, if possible, symmetric $k$-by-$k$\npositive semidefinite matrices $\\{A^1,...,A^m\\}$ and $\\{B^1,...,B^n\\}$ such\nthat $X_{i,j}=\\text{trace}(A^iB^j)$ for $i=1,...,m$, and $j=1,...n$. PSD\nfactorization is NP-hard. In this work, we introduce several local optimization\nschemes to tackle this problem: a fast projected gradient method and two\nalgorithms based on the coordinate descent framework. The main application of\nPSD factorization is the computation of semidefinite extensions, that is, the\nrepresentations of polyhedrons as projections of spectrahedra, for which the\nmatrix to be factorized is the slack matrix of the polyhedron. We compare the\nperformance of our algorithms on this class of problems. In particular, we\ncompute the PSD extensions of size $k=1+ \\lceil \\log_2(n) \\rceil$ for the\nregular $n$-gons when $n=5$, $8$ and $10$. We also show how to generalize our\nalgorithms to compute the square root rank (which is the size of the factors in\na PSD factorization where all factor matrices $A^i$ and $B^j$ have rank one)\nand completely PSD factorizations (which is the special case where the input\nmatrix is symmetric and equality $A^i=B^i$ is required for all $i$).\n",
"title": "Algorithms for Positive Semidefinite Factorization"
}
| null | null |
[
"Computer Science",
"Mathematics"
] | null | true | null |
13717
| null |
Validated
| null | null |
null |
{
"abstract": " We discuss various characterizations of synthetic upper Ricci bounds for\nmetric measure spaces in terms of heat flow, entropy and optimal transport. In\nparticular, we present a characterization in terms of semiconcavity of the\nentropy along certain Wasserstein geodesics which is stable under convergence\nof mm-spaces. And we prove that a related characterization is equivalent to an\nasymptotic lower bound on the growth of the Wasseretein distance between heat\nflows. For weighted Riemannian manifolds, the crucial result will be a precise\nuniform two-sided bound for \\begin{eqnarray*}\\frac{d}{dt}\\Big|_{t=0}W\\big(\\hat\nP_t\\delta_x,\\hat P_t\\delta_y\\big)\\end{eqnarray*} in terms of the mean value of\nthe Bakry-Emery Ricci tensor $\\mathrm{Ric}+\\mathrm{Hess}\\, f$ along the\nminimizing geodesic from $x$ to $y$ and an explicit correction term depending\non the bound for the curvature along this curve.\n",
"title": "Remarks about Synthetic Upper Ricci Bounds for Metric Measure Spaces"
}
| null | null | null | null | true | null |
13718
| null |
Default
| null | null |
null |
{
"abstract": " Given a pedestrian image as a query, the purpose of person re-identification\nis to identify the correct match from a large collection of gallery images\ndepicting the same person captured by disjoint camera views. The critical\nchallenge is how to construct a robust yet discriminative feature\nrepresentation to capture the compounded variations in pedestrian appearance.\nTo this end, deep learning methods have been proposed to extract hierarchical\nfeatures against extreme variability of appearance. However, existing methods\nin this category generally neglect the efficiency in the matching stage whereas\nthe searching speed of a re-identification system is crucial in real-world\napplications. In this paper, we present a novel deep hashing framework with\nConvolutional Neural Networks (CNNs) for fast person re-identification.\nTechnically, we simultaneously learn both CNN features and hash functions/codes\nto get robust yet discriminative features and similarity-preserving hash codes.\nThereby, person re-identification can be resolved by efficiently computing and\nranking the Hamming distances between images. A structured loss function\ndefined over positive pairs and hard negatives is proposed to formulate a novel\noptimization problem so that fast convergence and more stable optimized\nsolution can be obtained. Extensive experiments on two benchmarks CUHK03\n\\cite{FPNN} and Market-1501 \\cite{Market1501} show that the proposed deep\narchitecture is efficacy over state-of-the-arts.\n",
"title": "Structured Deep Hashing with Convolutional Neural Networks for Fast Person Re-identification"
}
| null | null | null | null | true | null |
13719
| null |
Default
| null | null |
null |
{
"abstract": " Recently, methods have been proposed that perform texture synthesis and style\ntransfer by using convolutional neural networks (e.g. Gatys et al.\n[2015,2016]). These methods are exciting because they can in some cases create\nresults with state-of-the-art quality. However, in this paper, we show these\nmethods also have limitations in texture quality, stability, requisite\nparameter tuning, and lack of user controls. This paper presents a multiscale\nsynthesis pipeline based on convolutional neural networks that ameliorates\nthese issues. We first give a mathematical explanation of the source of\ninstabilities in many previous approaches. We then improve these instabilities\nby using histogram losses to synthesize textures that better statistically\nmatch the exemplar. We also show how to integrate localized style losses in our\nmultiscale framework. These losses can improve the quality of large features,\nimprove the separation of content and style, and offer artistic controls such\nas paint by numbers. We demonstrate that our approach offers improved quality,\nconvergence in fewer iterations, and more stability over the optimization.\n",
"title": "Stable and Controllable Neural Texture Synthesis and Style Transfer Using Histogram Losses"
}
| null | null | null | null | true | null |
13720
| null |
Default
| null | null |
null |
{
"abstract": " The present work analyses a particular scenario of consensus formation, where\nthe individuals navigate across an underlying network defining the topology of\nthe walks. The consensus, associated to a given opinion coded as a simple\nmessages, is generated by interactions during the agent's walk and manifest\nitself in the collapse of the various opinions into a single one. We analyze\nhow the topology of the underlying networks and the rules of interaction\nbetween the agents promote or inhibit the emergence of this consensus. We find\nthat non-linear interaction rules are required to form consensus and that\nconsensus is more easily achieved in networks whose degree distribution is\nnarrower.\n",
"title": "Dynamical and Topological Aspects of Consensus Formation in Complex Networks"
}
| null | null | null | null | true | null |
13721
| null |
Default
| null | null |
null |
{
"abstract": " Administrators in all academic organizations across the world have to deal\nwith the unenviable task of comparing researchers on the basis of their\nacademic contributions. This job is further complicated by the need for\ncomparing single author publication with joint author publications.\nUnfortunately, however, there is no reasonably established consensus on the\nmethod of arriving at such comparisons, which typically involve trading off\naccomplishments in teaching, grant writing and academic publication. In this\npaper, we focus on the particular dimension of academic publication, and\nanalyze this issue from a more fundamental perspective than addressed by the\npopular $h$-index (which may lead to unfair and counter-intuitive comparisons\nin certain situations). In particular, we undertake an axiomatic analysis of\n{\\it all} possible ways to measure academic authorship for a given dataset of\nresearch articles and find that an egalitarian $e$-index is the \\textit{only}\nmethod which satisfies the axioms of anonymity, monotonicity, and efficiency.\nThis index divides authorship of joint projects equally and sums across all\npublications of an author. Thus, our index provides a method to prorate\nauthorship for multi-author projects, and thereby, delivers more balanced\nauthor comparisons.\n",
"title": "Measurement of authorship by publications: a normative approach"
}
| null | null | null | null | true | null |
13722
| null |
Default
| null | null |
null |
{
"abstract": " We introduce a condition on Garside groups that we call Dehornoy structure.\nAn iteration of such a structure leads to a left order on the group. We show\nconditions for a Garside group to admit a Dehornoy structure, and we apply\nthese criteria to prove that the Artin groups of type A and I 2 (m), m $\\ge$ 4,\nhave Dehornoy structures. We show that the left orders on the Artin groups of\ntype A obtained from their Dehornoy structures are the Dehornoy orders. In the\ncase of the Artin groups of type I 2 (m), m $\\ge$ 4, we show that the left\norders derived from their Dehornoy structures coincide with the orders obtained\nfrom embeddings of the groups into braid groups. 20F36\n",
"title": "Ordering Garside groups"
}
| null | null | null | null | true | null |
13723
| null |
Default
| null | null |
null |
{
"abstract": " In this paper, a new approach for classification of target task using limited\nlabeled target data as well as enormous unlabeled source data is proposed which\nis called self-taught learning. The target and source data can be drawn from\ndifferent distributions. In the previous approaches, covariate shift assumption\nis considered where the marginal distributions p(x) change over domains and the\nconditional distributions p(y|x) remain the same. In our approach, we propose a\nnew objective function which simultaneously learns a common space T(.) where\nthe conditional distributions over domains p(T(x)|y) remain the same and learns\nrobust SVM classifiers for target task using both source and target data in the\nnew representation. Hence, in the proposed objective function, the hidden label\nof the source data is also incorporated. We applied the proposed approach on\nCaltech-256, MSRC+LMO datasets and compared the performance of our algorithm to\nthe available competing methods. Our method has a superior performance to the\nsuccessful existing algorithms.\n",
"title": "Self-Taught Support Vector Machine"
}
| null | null | null | null | true | null |
13724
| null |
Default
| null | null |
null |
{
"abstract": " The canonical and grand-canonical ensembles are two usual marginal cases for\nultracold Bose gases, but real collections of experimental runs commonly have\nintermediate properties. Here we study the continuum of intermediate cases, and\nlook into the appearance of ensemble equivalence as interaction rises for\nmesoscopic 1d systems. We demonstrate how at sufficient interaction strength\nthe distributions of condensate and excited atoms become practically identical\nregardless of the ensemble used. Importantly, we find that features that are\nfragile in the ideal gas and appear only in a strict canonical ensemble can\nbecome robust in all ensembles when interactions become strong. As evidence,\nthe steep cliff in the distribution of the number of excited atoms is\npreserved. To make this study, a straightforward approach for generating\ncanonical and intermediate classical field ensembles using a modified\nstochastic Gross-Pitaevskii equation (SGPE) is developed.\n",
"title": "Continuum of classical-field ensembles from canonical to grand canonical and the onset of their equivalence"
}
| null | null | null | null | true | null |
13725
| null |
Default
| null | null |
null |
{
"abstract": " The partial representation extension problem, introduced by Klavík et al.\n(2011), generalizes the recognition problem. In this short note we show that\nthis problem is NP-complete for unit circular-arc graphs.\n",
"title": "Extending Partial Representations of Unit Circular-arc Graphs"
}
| null | null |
[
"Computer Science"
] | null | true | null |
13726
| null |
Validated
| null | null |
null |
{
"abstract": " The main aim of this paper is to study the Lipschitz continuity of certain\n$(K, K')$-quasiconformal mappings with respect to the distance ratio metric,\nand the Lipschitz continuity of the solution of a quasilinear differential\nequation with respect to the distance ratio metric.\n",
"title": "Lipschitz continuity of quasiconformal mappings and of the solutions to second order elliptic PDE with respect to the distance ratio metric"
}
| null | null | null | null | true | null |
13727
| null |
Default
| null | null |
null |
{
"abstract": " We propose an estimation method for the conditional mode when the\nconditioning variable is high-dimensional. In the proposed method, we first\nestimate the conditional density by solving quantile regressions multiple\ntimes. We then estimate the conditional mode by finding the maximum of the\nestimated conditional density. The proposed method has two advantages in that\nit is computationally stable because it has no initial parameter dependencies,\nand it is statistically efficient with a fast convergence rate. Synthetic and\nreal-world data experiments demonstrate the better performance of the proposed\nmethod compared to other existing ones.\n",
"title": "On Estimation of Conditional Modes Using Multiple Quantile Regressions"
}
| null | null |
[
"Mathematics",
"Statistics"
] | null | true | null |
13728
| null |
Validated
| null | null |
null |
{
"abstract": " We report results of isothermal magnetotransport and susceptibility\nmeasurements at elevated magnetic fields B down to very low temperatures T on\nhigh-quality single crystals of the frustrated Kondo-lattice system CePdAl.\nThey reveal a B*(T) line within the paramagnetic part of the phase diagram.\nThis line denotes a thermally broadened 'small'-to-'large' Fermi surface\ncrossover which substantially narrows upon cooling. At B_0* = B*(T=0) = (4.6\n+/- 0.1) T, this B*(T) line merges with two other crossover lines, viz. Tp(B)\nbelow and T_FL(B) above B_0*. Tp characterizes a frustration-dominated\nspin-liquid state, while T_FL is the Fermi-liquid temperature associated with\nthe lattice Kondo effect. Non-Fermi-liquid phenomena which are commonly\nobserved near a 'Kondo destruction' quantum critical point cannot be resolved\nin CePdAl. Our observations reveal a rare case where Kondo coupling,\nfrustration and quantum criticality are closely intertwined.\n",
"title": "Kondo destruction in a quantum paramagnet with magnetic frustration"
}
| null | null | null | null | true | null |
13729
| null |
Default
| null | null |
null |
{
"abstract": " Error backpropagation is a highly effective mechanism for learning\nhigh-quality hierarchical features in deep networks. Updating the features or\nweights in one layer, however, requires waiting for the propagation of error\nsignals from higher layers. Learning using delayed and non-local errors makes\nit hard to reconcile backpropagation with the learning mechanisms observed in\nbiological neural networks as it requires the neurons to maintain a memory of\nthe input long enough until the higher-layer errors arrive. In this paper, we\npropose an alternative learning mechanism where errors are generated locally in\neach layer using fixed, random auxiliary classifiers. Lower layers could thus\nbe trained independently of higher layers and training could either proceed\nlayer by layer, or simultaneously in all layers using local error information.\nWe address biological plausibility concerns such as weight symmetry\nrequirements and show that the proposed learning mechanism based on fixed,\nbroad, and random tuning of each neuron to the classification categories\noutperforms the biologically-motivated feedback alignment learning technique on\nthe MNIST, CIFAR10, and SVHN datasets, approaching the performance of standard\nbackpropagation. Our approach highlights a potential biological mechanism for\nthe supervised, or task-dependent, learning of feature hierarchies. In\naddition, we show that it is well suited for learning deep networks in custom\nhardware where it can drastically reduce memory traffic and data communication\noverheads.\n",
"title": "Deep supervised learning using local errors"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
13730
| null |
Validated
| null | null |
null |
{
"abstract": " The mean field variational Bayes method is becoming increasingly popular in\nstatistics and machine learning. Its iterative Coordinate Ascent Variational\nInference algorithm has been widely applied to large scale Bayesian inference.\nSee Blei et al. (2017) for a recent comprehensive review. Despite the\npopularity of the mean field method there exist remarkably little fundamental\ntheoretical justifications. To the best of our knowledge, the iterative\nalgorithm has never been investigated for any high dimensional and complex\nmodel. In this paper, we study the mean field method for community detection\nunder the Stochastic Block Model. For an iterative Batch Coordinate Ascent\nVariational Inference algorithm, we show that it has a linear convergence rate\nand converges to the minimax rate within $\\log n$ iterations. This complements\nthe results of Bickel et al. (2013) which studied the global minimum of the\nmean field variational Bayes and obtained asymptotic normal estimation of\nglobal model parameters. In addition, we obtain similar optimality results for\nGibbs sampling and an iterative procedure to calculate maximum likelihood\nestimation, which can be of independent interest.\n",
"title": "Theoretical and Computational Guarantees of Mean Field Variational Inference for Community Detection"
}
| null | null | null | null | true | null |
13731
| null |
Default
| null | null |
null |
{
"abstract": " The era of the next generation of giant telescopes requires not only the\nadvent of new technologies but also the development of novel methods, in order\nto exploit fully the extraordinary potential they are built for. Global Multi\nConjugate Adaptive Optics (GMCAO) pursues this approach, with the goal of\nachieving good performance over a field of view of a few arcmin and an increase\nin sky coverage. In this article, we show the gain offered by this technique to\nan astrophysical application, such as the photometric survey strategy applied\nto the Chandra Deep Field South as a case study. We simulated a close-to-real\nobservation of a 500 x 500 arcsec^2 extragalactic deep field with a 40-m class\ntelescope that implements GMCAO. We analysed mock K-band images of 6000\nhigh-redshift (up to z = 2.75) galaxies therein as if they were real to recover\nthe initial input parameters. We attained 94.5 per cent completeness for source\ndetection with SEXTRACTOR. We also measured the morphological parameters of all\nthe sources with the two-dimensional fitting tools GALFIT. The agreement we\nfound between recovered and intrinsic parameters demonstrates GMCAO as a\nreliable approach to assist extremely large telescope (ELT) observations of\nextragalactic interest.\n",
"title": "The Chandra Deep Field South as a test case for Global Multi Conjugate Adaptive Optics"
}
| null | null |
[
"Physics"
] | null | true | null |
13732
| null |
Validated
| null | null |
null |
{
"abstract": " Mastering the dynamics of social influence requires separating, in a database\nof information propagation traces, the genuine causal processes from temporal\ncorrelation, i.e., homophily and other spurious causes. However, most studies\nto characterize social influence, and, in general, most data-science analyses\nfocus on correlations, statistical independence, or conditional independence.\nOnly recently, there has been a resurgence of interest in \"causal data\nscience\", e.g., grounded on causality theories. In this paper we adopt a\nprincipled causal approach to the analysis of social influence from\ninformation-propagation data, rooted in the theory of probabilistic causation.\nOur approach consists of two phases. In the first one, in order to avoid the\npitfalls of misinterpreting causation when the data spans a mixture of several\nsubtypes (\"Simpson's paradox\"), we partition the set of propagation traces into\ngroups, in such a way that each group is as less contradictory as possible in\nterms of the hierarchical structure of information propagation. To achieve this\ngoal, we borrow the notion of \"agony\" and define the Agony-bounded Partitioning\nproblem, which we prove being hard, and for which we develop two efficient\nalgorithms with approximation guarantees. In the second phase, for each group\nfrom the first phase, we apply a constrained MLE approach to ultimately learn a\nminimal causal topology. Experiments on synthetic data show that our method is\nable to retrieve the genuine causal arcs w.r.t. a ground-truth generative\nmodel. Experiments on real data show that, by focusing only on the extracted\ncausal structures instead of the whole social graph, the effectiveness of\npredicting influence spread is significantly improved.\n",
"title": "Probabilistic Causal Analysis of Social Influence"
}
| null | null | null | null | true | null |
13733
| null |
Default
| null | null |
null |
{
"abstract": " It is of great concern to produce numerically efficient methods for moisture\ndiffusion through porous media, capable of accurately calculate moisture\ndistribution with a reduced computational effort. In this way, model reduction\nmethods are promising approaches to bring a solution to this issue since they\ndo not degrade the physical model and provide a significant reduction of\ncomputational cost. Therefore, this article explores in details the\ncapabilities of two model-reduction techniques - the Spectral Reduced-Order\nModel (Spectral-ROM) and the Proper Generalised Decomposition (PGD) - to\nnumerically solve moisture diffusive transfer through porous materials. Both\napproaches are applied to three different problems to provide clear examples of\nthe construction and use of these reduced-order models. The methodology of both\napproaches is explained extensively so that the article can be used as a\nnumerical benchmark by anyone interested in building a reduced-order model for\ndiffusion problems in porous materials. Linear and non-linear unsteady\nbehaviors of unidimensional moisture diffusion are investigated. The last case\nfocuses on solving a parametric problem in which the solution depends on space,\ntime and the diffusivity properties. Results have highlighted that both methods\nprovide accurate solutions and enable to reduce significantly the order of the\nmodel around ten times lower than the large original model. It also allows an\nefficient computation of the physical phenomena with an error lower than\n10^{-2} when compared to a reference solution.\n",
"title": "Advanced reduced-order models for moisture diffusion in porous media"
}
| null | null | null | null | true | null |
13734
| null |
Default
| null | null |
null |
{
"abstract": " We study gapless quantum spin chains with spin 1/2 and 1: the Fredkin and\nMotzkin models. Their entangled groundstates are known exactly but not their\nexcitation spectra. We first express the groundstates in the continuum which\nallows for the calculation of spin and entanglement properties in a unified\nfashion. Doing so, we uncover an emergent conformal-type symmetry, thus\nconsolidating the connection to a widely studied family of Lifshitz quantum\ncritical points in 2d. We then obtain the low lying excited states via\nlarge-scale DMRG simulations and find that the dynamical exponent is z = 3.2 in\nboth cases. Other excited states show a different z, indicating that these\nmodels have multiple dynamics. Moreover, we modify the spin-1/2 model by adding\na ferromagnetic Heisenberg term, which changes the entire spectrum. We track\nthe resulting non-trivial evolution of the dynamical exponents using DMRG.\nFinally, we exploit an exact map from the quantum Hamiltonian to the\nnon-equilibrium dynamics of a classical spin chain to shed light on the quantum\ndynamics.\n",
"title": "Gapless quantum spin chains: multiple dynamics and conformal wavefunctions"
}
| null | null |
[
"Physics"
] | null | true | null |
13735
| null |
Validated
| null | null |
null |
{
"abstract": " A new area in which passive WiFi analytics have promise for delivering value\nis the real-time monitoring of public transport systems. One example is\ndetermining the true (as opposed to the published) timetable of a public\ntransport system in real-time. In most cases, there are no other\npublicly-available sources for this information. Yet, it is indispensable for\nthe real-time monitoring of public transport service levels. Furthermore, this\ninformation, if accurate and temporally fine-grained, can be used for very\nlow-latency incident detection. In this work, we propose using spectral\nclustering based on trajectories derived from passive WiFi traces of users of a\npublic transport system to infer the true timetable and two key performance\nindicators of the transport service, namely public transport vehicle headway\nand in-station dwell time. By detecting anomalous dwell times or headways, we\ndemonstrate that a fast and accurate real-time incident-detection procedure can\nbe obtained. The method we introduce makes use of the advantages of the\nhigh-frequency WiFi data, which provides very low-latency,\nuniversally-accessible information, while minimizing the impact of the noise in\nthe data.\n",
"title": "Real-time public transport service-level monitoring using passive WiFi: a spectral clustering approach for train timetable estimation"
}
| null | null | null | null | true | null |
13736
| null |
Default
| null | null |
null |
{
"abstract": " We introduce a new method for finding network motifs: interesting or\ninformative subgraph patterns in a network. Current methods for finding motifs\nrely on the frequency of the motif: specifically, subgraphs are motifs when\ntheir frequency in the data is high compared to the expected frequency under a\nnull model. To compute this expectation, the search for motifs is normally\nrepeated on as many as 1000 random graphs sampled from the null model; a\nprohibitively expensive step. We use ideas from the Minimum Description Length\n(MDL) literature to define a new measure of motif relevance, and a new\nalgorithm for detecting motifs. Our method allows motif analysis to scale to\nnetworks with billions of links, while still resulting in informative motifs.\n",
"title": "Finding Network Motifs in Large Graphs using Compression as a Measure of Relevance"
}
| null | null | null | null | true | null |
13737
| null |
Default
| null | null |
null |
{
"abstract": " The traditional bag-of-words approach has found a wide range of applications\nin computer vision. The standard pipeline consists of a generation of a visual\nvocabulary, a quantization of the features into histograms of visual words, and\na classification step for which usually a support vector machine in combination\nwith a non-linear kernel is used. Given large amounts of data, however, the\nmodel suffers from a lack of discriminative power. This applies particularly\nfor action recognition, where the vast amount of video features needs to be\nsubsampled for unsupervised visual vocabulary generation. Moreover, the kernel\ncomputation can be very expensive on large datasets. In this work, we propose a\nrecurrent neural network that is equivalent to the traditional bag-of-words\napproach but enables for the application of discriminative training. The model\nfurther allows to incorporate the kernel computation into the neural network\ndirectly, solving the complexity issue and allowing to represent the complete\nclassification system within a single network. We evaluate our method on four\nrecent action recognition benchmarks and show that the conventional model as\nwell as sparse coding methods are outperformed.\n",
"title": "A Bag-of-Words Equivalent Recurrent Neural Network for Action Recognition"
}
| null | null | null | null | true | null |
13738
| null |
Default
| null | null |
null |
{
"abstract": " To date, the only limit on graviton mass using galaxy clusters was obtained\nby Goldhaber and Nieto in 1974, using the fact that the orbits of galaxy\nclusters are bound and closed, and extend up to 580 kpc. From positing that\nonly a Newtonian potential gives rise to such stable bound orbits, a limit on\nthe graviton mass $m_g<10^{-29}$ eV was obtained (PRD 9,1119, 1974). Recently,\nit has been shown that one can obtain closed bound orbits for Yukawa potential\n(arXiv:1705.02444), thus invalidating the main \\emph{ansatz} used in Goldhaber\nand Nieto to obtain the graviton mass bound. In order to obtain a revised\nestimate using galaxy clusters, we use dynamical mass models of the Abell 1689\n(A1689) galaxy cluster to check their compatibility with a Yukawa gravitational\npotential. We assume mass models for the gas, dark matter, and galaxies for\nA1689 from arXiv:1703.10219 and arXiv:1610.01543, who used this cluster to test\nvarious alternate gravity theories, which dispense with the need for dark\nmatter. We quantify the deviations in the acceleration profile using these mass\nmodels assuming a Yukawa potential and that obtained assuming a Newtonian\npotential by calculating the $\\chi^2$ residuals between the two profiles. Our\nestimated bound on the graviton mass ($m_g$) is thereby given by, $m_g < 1.37\n\\times 10^{-29}$ eV or in terms of the graviton Compton wavelength of,\n$\\lambda_g>9.1 \\times 10^{19}$ km at 90\\% confidence level.\n",
"title": "Limit on graviton mass from galaxy cluster Abell 1689"
}
| null | null |
[
"Physics"
] | null | true | null |
13739
| null |
Validated
| null | null |
null |
{
"abstract": " Artificial neural networks (ANNs) may not be worth their computational/memory\ncosts when used in mobile phones or embedded devices. Parameter-pruning\nalgorithms combat these costs, with some algorithms capable of removing over\n90% of an ANN's weights without harming the ANN's performance. Removing weights\nfrom an ANN is a form of regularization, but existing pruning algorithms do not\nsignificantly improve generalization error. We show that pruning ANNs can\nimprove generalization if pruning targets large weights instead of small\nweights. Applying our pruning algorithm to an ANN leads to a higher image\nclassification accuracy on CIFAR-10 data than applying the popular regularizer\ndropout. The pruning couples this higher accuracy with an 85% reduction of the\nANN's parameter count.\n",
"title": "Enhancing the Regularization Effect of Weight Pruning in Artificial Neural Networks"
}
| null | null | null | null | true | null |
13740
| null |
Default
| null | null |
null |
{
"abstract": " For any channel with a convex constraint set and finite Augustin capacity,\nexistence of a unique Augustin center and associated Erven-Harremoes bound are\nestablished. Augustin-Legendre capacity, center, and radius are introduced and\nproved to be equal to the corresponding Renyi-Gallager entities. Sphere packing\nbounds with polynomial prefactors are derived for codes on two families of\nchannels: (possibly non-stationary) memoryless channels with multiple additive\ncost constraints and stationary memoryless channels with convex constraints on\nthe empirical distribution of the input codewords.\n",
"title": "The Augustin Center and The Sphere Packing Bound For Memoryless Channels"
}
| null | null | null | null | true | null |
13741
| null |
Default
| null | null |
null |
{
"abstract": " This paper presents a new algorithm for calculating hash signatures of sets\nwhich can be directly used for Jaccard similarity estimation. The new approach\nis an improvement over the MinHash algorithm, because it has a better runtime\nbehavior and the resulting signatures allow a more precise estimation of the\nJaccard index.\n",
"title": "SuperMinHash - A New Minwise Hashing Algorithm for Jaccard Similarity Estimation"
}
| null | null | null | null | true | null |
13742
| null |
Default
| null | null |
null |
{
"abstract": " Generative Adversarial Networks (GAN) are one of the most prominent tools for\nlearning complicated distributions. However, problems such as mode collapse and\ncatastrophic forgetting, prevent GAN from learning the target distribution.\nThese problems are usually studied independently from each other. In this\npaper, we show that both problems are present in GAN and their combined effect\nmakes the training of GAN unstable. We also show that methods such as gradient\npenalties and momentum based optimizers can improve the stability of GAN by\neffectively preventing these problems from happening. Finally, we study a\nmechanism for mode collapse to occur and propagate in feedforward neural\nnetworks.\n",
"title": "On catastrophic forgetting and mode collapse in Generative Adversarial Networks"
}
| null | null | null | null | true | null |
13743
| null |
Default
| null | null |
null |
{
"abstract": " We propose to estimate a metamodel and the sensitivity indices of a complex\nmodel m in the Gaussian regression framework. Our approach combines methods for\nsensitivity analysis of complex models and statistical tools for sparse\nnon-parametric estimation in multivariate Gaussian regression model. It rests\non the construction of a metamodel for aproximating the Hoeffding-Sobol\ndecomposition of m. This metamodel belongs to a reproducing kernel Hilbert\nspace constructed as a direct sum of Hilbert spaces leading to a functional\nANOVA decomposition. The estimation of the metamodel is carried out via a\npenalized least-squares minimization allowing to select the subsets of\nvariables that contribute to predict the output. It allows to estimate the\nsensitivity indices of m. We establish an oracle-type inequality for the risk\nof the estimator, describe the procedure for estimating the metamodel and the\nsensitivity indices, and assess the performances of the procedure via a\nsimulation study.\n",
"title": "Metamodel Construction for Sensitivity Analysis"
}
| null | null |
[
"Mathematics",
"Statistics"
] | null | true | null |
13744
| null |
Validated
| null | null |
null |
{
"abstract": " Neuronal and glial cells release diverse proteoglycans and glycoproteins,\nwhich aggregate in the extracellular space and form the extracellular matrix\n(ECM) that may in turn regulate major cellular functions. Brain cells also\nrelease extracellular proteases that may degrade the ECM, and both synthesis\nand degradation of ECM are activity-dependent. In this study we introduce a\nmathematical model describing population dynamics of neurons interacting with\nECM molecules over extended timescales. It is demonstrated that depending on\nthe prevalent biophysical mechanism of ECM-neuronal interactions, different\ndynamical regimes of ECM activity can be observed, including bistable states\nwith stable stationary levels of ECM molecule concentration, spontaneous ECM\noscillations, and coexistence of ECM oscillations and a stationary state,\nallowing dynamical switches between activity regimes.\n",
"title": "Dynamics of the brain extracellular matrix governed by interactions with neural cells"
}
| null | null | null | null | true | null |
13745
| null |
Default
| null | null |
null |
{
"abstract": " We present AutoPerf, a generalized software performance regression diagnosis\nsystem. AutoPerf uses autoencoders, an unsupervised learning technique, and\nhardware performance counters to learn the performance signatures of a program.\nIt then uses this knowledge to identify when newer versions of the program\nsuffer from performance regressions, while simultaneously providing root cause\nanalysis to help programmers debug the program's performance.\nAutoPerf is the first zero-positive learning performance regression diagnosis\nsystem. It trains entirely in the negative (non-anomalous) space to learn\npositive (anomalous) behaviors. We demonstrate AutoPerf's generality against\nthree different types of performance regressions: (i) true sharing cache\ncontention, (ii) false sharing cache contention, and (iii) NUMA latencies\nacross 15 real world performance regressions and 7 open source programs. On\naverage, AutoPerf exhibits only 3.7% profiling overhead and diagnoses more\nregressions than prior state-of-the-art approaches.\n",
"title": "AutoPerf: A Generalized Zero-Positive Learning System to Detect Software Performance Anomalies"
}
| null | null | null | null | true | null |
13746
| null |
Default
| null | null |
null |
{
"abstract": " The way Quantum Mechanics (QM) is introduced to people used to Classical\nMechanics (CM) is by a complete change of the general methodology) despite QM\nhistorically stemming from CM as a means to explain experimental results.\nTherefore, it is desirable to build a bridge from CM to QM.\nThis paper presents a generalization of CM to QM. It starts from the\ngeneralization of a point-like object and naturally arrives at the quantum\nstate vector of quantum systems in the complex valued Hilbert space, its time\nevolution and quantum representation of a measurement apparatus of any size.\nEach time, when generalization is performed, there is a possibility to develop\nnew theory giving up most simple generalizations. It is shown that a\nmeasurement apparatus is a special case of a general quantum object. An example\nof a measurement apparatus of an intermediate size is considered in the end.\n",
"title": "Building a bridge between Classical and Quantum Mechanics"
}
| null | null | null | null | true | null |
13747
| null |
Default
| null | null |
null |
{
"abstract": " We study the sample complexity of learning neural networks, by providing new\nbounds on their Rademacher complexity assuming norm constraints on the\nparameter matrix of each layer. Compared to previous work, these complexity\nbounds have improved dependence on the network depth, and under some additional\nassumptions, are fully independent of the network size (both depth and width).\nThese results are derived using some novel techniques, which may be of\nindependent interest.\n",
"title": "Size-Independent Sample Complexity of Neural Networks"
}
| null | null | null | null | true | null |
13748
| null |
Default
| null | null |
null |
{
"abstract": " Variational Bayes (VB), also known as independent mean-field approximation,\nhas become a popular method for Bayesian network inference in recent years. Its\napplication is vast, e.g. in neural network, compressed sensing, clustering,\netc. to name just a few. In this paper, the independence constraint in VB will\nbe relaxed to a conditional constraint class, called copula in statistics.\nSince a joint probability distribution always belongs to a copula class, the\nnovel copula VB (CVB) approximation is a generalized form of VB. Via\ninformation geometry, we will see that CVB algorithm iteratively projects the\noriginal joint distribution to a copula constraint space until it reaches a\nlocal minimum Kullback-Leibler (KL) divergence. By this way, all mean-field\napproximations, e.g. iterative VB, Expectation-Maximization (EM), Iterated\nConditional Mode (ICM) and k-means algorithms, are special cases of CVB\napproximation.\nFor a generic Bayesian network, an augmented hierarchy form of CVB will also\nbe designed. While mean-field algorithms can only return a locally optimal\napproximation for a correlated network, the augmented CVB network, which is an\noptimally weighted average of a mixture of simpler network structures, can\npotentially achieve the globally optimal approximation for the first time. Via\nsimulations of Gaussian mixture clustering, the classification's accuracy of\nCVB will be shown to be far superior to that of state-of-the-art VB, EM and\nk-means algorithms.\n",
"title": "Copula Variational Bayes inference via information geometry"
}
| null | null | null | null | true | null |
13749
| null |
Default
| null | null |
null |
{
"abstract": " We have studied a two dimensional lattice model of Coulomb glass for a wide\nrange of disorders at $T\\sim 0$. The system was first annealed using Monte\nCarlo simulation. Further minimization of the total energy of the system was\ndone using Baranovskii et al algorithm followed by cluster flipping to obtain\nthe pseudo ground states. We have shown that the energy required to create a\ndomain of linear size L in d dimensions is proportional to $L^{d-1}$. Using\nImry-Ma arguments given for random field Ising model, one gets critical\ndimension $d_{c}\\geq 2$ for Coulomb glass. The investigations of domains in the\ntransition region shows a discontinuity in staggered magnetization which is an\nindication of a first-order type transition from charge-ordered phase to\ndisordered phase. The structure and nature of Random field fluctuations of the\nsecond largest domain in Coulomb glass are inconsistent with the assumptions of\nImry and Ma as was also reported for random field Ising model. The study of\ndomains showed that in the transition region there were mostly two large\ndomains and as disorder was increased, the two large domains remained but there\nwere a large number of small domains. We have also studied the properties of\nthe second largest domain as a function of disorder. We furthermore analysed\nthe effect of disorder on the density of states and showed a transition from\nhard gap at low disorders to a soft gap at higher disorders. At $W=2$, we have\nanalysed the soft gap in detail and found that the density of states deviates\nslightly ($\\delta\\approx 1.293 \\pm 0.027$) from the linear behaviour in two\ndimensions. Analysis of local minima show that the pseudo ground states have\nsimilar structure.\n",
"title": "Effect of increasing disorder on domains of the two-dimensional Coulomb glass"
}
| null | null |
[
"Physics"
] | null | true | null |
13750
| null |
Validated
| null | null |
null |
{
"abstract": " We propose a dynamic network model where two mechanisms control the\nprobability of a link between two nodes: (i) the existence or absence of this\nlink in the past, and (ii) node-specific latent variables (dynamic fitnesses)\ndescribing the propensity of each node to create links. Assuming a Markov\ndynamics for both mechanisms, we propose an Expectation-Maximization algorithm\nfor model estimation and inference of the latent variables. The estimated\nparameters and fitnesses can be used to forecast the presence of a link in the\nfuture. We apply our methodology to the e-MID interbank network for which the\ntwo linkage mechanisms are associated with two different trading behaviors in\nthe process of network formation, namely preferential trading and trading\ndriven by node-specific characteristics. The empirical results allow to\nrecognise preferential lending in the interbank market and indicate how a\nmethod that does not account for time-varying network topologies tends to\noverestimate preferential linkage.\n",
"title": "A dynamic network model with persistent links and node-specific latent variables, with an application to the interbank market"
}
| null | null | null | null | true | null |
13751
| null |
Default
| null | null |
null |
{
"abstract": " A stock market is considered as one of the highly complex systems, which\nconsists of many components whose prices move up and down without having a\nclear pattern. The complex nature of a stock market challenges us on making a\nreliable prediction of its future movements. In this paper, we aim at building\na new method to forecast the future movements of Standard & Poor's 500 Index\n(S&P 500) by constructing time-series complex networks of S&P 500 underlying\ncompanies by connecting them with links whose weights are given by the mutual\ninformation of 60-minute price movements of the pairs of the companies with the\nconsecutive 5,340 minutes price records. We showed that the changes in the\nstrength distributions of the networks provide an important information on the\nnetwork's future movements. We built several metrics using the strength\ndistributions and network measurements such as centrality, and we combined the\nbest two predictors by performing a linear combination. We found that the\ncombined predictor and the changes in S&P 500 show a quadratic relationship,\nand it allows us to predict the amplitude of the one step future change in S&P\n500. The result showed significant fluctuations in S&P 500 Index when the\ncombined predictor was high. In terms of making the actual index predictions,\nwe built ARIMA models. We found that adding the network measurements into the\nARIMA models improves the model accuracy. These findings are useful for\nfinancial market policy makers as an indicator based on which they can\ninterfere with the markets before the markets make a drastic change, and for\nquantitative investors to improve their forecasting models.\n",
"title": "Predicting stock market movements using network science: An information theoretic approach"
}
| null | null | null | null | true | null |
13752
| null |
Default
| null | null |
null |
{
"abstract": " In this article, the notion of bi-monotonic independence is introduced as an\nextension of monotonic independence to the two-faced framework for a family of\npairs of algebras in a non-commutative space. The associated cumulants are\ndefined and a moment-cumulant formula is derived in the bi-monotonic setting.\nIn general the bi-monotonic product of states is not a state and the\nbi-monotonic convolution of probability measures on the plane is not a\nprobability measure. This provides an additional example of how positivity need\nnot be preserved under conditional bi-free convolutions.\n",
"title": "Bi-monotonic independence for pairs of algebras"
}
| null | null | null | null | true | null |
13753
| null |
Default
| null | null |
null |
{
"abstract": " We present observations of the occulted active region AR12222 during the\nthird {\\em NuSTAR} solar campaign on 2014 December 11, with concurrent {\\em\nSDO/}AIA and {\\em FOXSI-2} sounding rocket observations. The active region\nproduced a medium size solar flare one day before the observations, at\n$\\sim18$UT on 2014 December 10, with the post-flare loops still visible at the\ntime of {\\em NuSTAR} observations. The time evolution of the source emission in\nthe {\\em SDO/}AIA $335\\textrm{\\AA}$ channel reveals the characteristics of an\nextreme-ultraviolet late phase event, caused by the continuous formation of new\npost-flare loops that arch higher and higher in the solar corona. The spectral\nfitting of {\\em NuSTAR} observations yields an isothermal source, with\ntemperature $3.8-4.6$ MK, emission measure $0.3-1.8 \\times 10^{46}\\textrm{\ncm}^{-3}$, and density estimated at $2.5-6.0 \\times 10^8 \\textrm{ cm}^{-3}$.\nThe observed AIA fluxes are consistent with the derived {\\em NuSTAR}\ntemperature range, favoring temperature values in the range $4.0-4.3$ MK. By\nexamining the post-flare loops' cooling times and energy content, we estimate\nthat at least 12 sets of post-flare loops were formed and subsequently cooled\nbetween the onset of the flare and {\\em NuSTAR} observations, with their total\nthermal energy content an order of magnitude larger than the energy content at\nflare peak time. This indicates that the standard approach of using only the\nflare peak time to derive the total thermal energy content of a flare can lead\nto a large underestimation of its value.\n",
"title": "Evidence of Significant Energy Input in the Late Phase of a Solar Flare from NuSTAR X-Ray Observations"
}
| null | null | null | null | true | null |
13754
| null |
Default
| null | null |
null |
{
"abstract": " Gender inequality starts before birth. Parents tend to prefer boys over\ngirls, which is manifested in reproductive behavior, marital life, and parents'\npastimes and investments in their children. While social media and sharing\ninformation about children (so-called \"sharenting\") have become an integral\npart of parenthood, it is not well-known if and how gender preference shapes\nonline behavior of users. In this paper, we investigate public mentions of\ndaughters and sons on social media. We use data from a popular social\nnetworking site on public posts from 635,665 users. We find that both men and\nwomen mention sons more often than daughters in their posts. We also find that\nposts featuring sons get more \"likes\" on average. Our results indicate that\ngirls are underrepresented in parents' digital narratives about their children.\nThis gender imbalance may send a message that girls are less important than\nboys, or that they deserve less attention, thus reinforcing gender inequality.\n",
"title": "Gender Bias in Sharenting: Both Men and Women Mention Sons More Often Than Daughters on Social Media"
}
| null | null | null | null | true | null |
13755
| null |
Default
| null | null |
null |
{
"abstract": " Hierarchical models for regionally aggregated disease incidence data commonly\ninvolve region specific latent random effects that are modelled jointly as\nhaving a multivariate Gaussian distribution. The covariance or precision matrix\nincorporates the spatial dependence between the regions. Common choices for the\nprecision matrix include the widely used intrinsic conditional autoregressive\nmodel, which is singular, and its nonsingular extension which lacks\ninterpretability. We propose a new parametric model for the precision matrix\nbased on a directed acyclic graph representation of the spatial dependence. Our\nmodel guarantees positive definiteness and, hence, in addition to being a valid\nprior for regional spatially correlated random effects, can also directly model\nthe outcome from dependent data like images and networks. Theoretical and\nempirical results demonstrate the interpretability of parameters in our model.\nOur precision matrix is sparse and the model is highly scalable for large\ndatasets. We also derive a novel order-free version which remedies the\ndependence of directed acyclic graphs on the ordering of the regions by\naveraging over all possible orderings. The resulting precision matrix is\navailable in closed form. We demonstrate the superior performance of our models\nover competing models using simulation experiments and a public health\napplication.\n",
"title": "Spatial disease mapping using Directed Acyclic Graph Auto-Regressive (DAGAR) models"
}
| null | null | null | null | true | null |
13756
| null |
Default
| null | null |
null |
{
"abstract": " Accelerometer measurements are the prime type of sensor information most\nthink of when seeking to measure physical activity. On the market, there are\nmany fitness measuring devices which aim to track calories burned and steps\ncounted through the use of accelerometers. These measurements, though good\nenough for the average consumer, are noisy and unreliable in terms of the\nprecision of measurement needed in a scientific setting. The contribution of\nthis paper is an innovative and highly accurate regression method which uses an\nintermediary two-stage classification step to better direct the regression of\nenergy expenditure values from accelerometer counts.\nWe show that through an additional unsupervised layer of intermediate feature\nconstruction, we can leverage latent patterns within accelerometer counts to\nprovide better grounds for activity classification than expert-constructed\ntimeseries features. For this, our approach utilizes a mathematical model\noriginating in natural language processing, the bag-of-words model, that has in\nthe past years been appearing in diverse disciplines outside of the natural\nlanguage processing field such as image processing. Further emphasizing the\nnatural language connection to stochastics, we use a gaussian mixture model to\nlearn the dictionary upon which the bag-of-words model is built. Moreover, we\nshow that with the addition of these features, we're able to improve regression\nroot mean-squared error of energy expenditure by approximately 1.4 units over\nexisting state-of-the-art methods.\n",
"title": "Bag-of-Words Method Applied to Accelerometer Measurements for the Purpose of Classification and Energy Estimation"
}
| null | null | null | null | true | null |
13757
| null |
Default
| null | null |
null |
{
"abstract": " We investigated the magnetic structure of the heavy fermion compound\nCePt$_2$In$_7$ below $T_N~=5.34(2)$ K using magnetic resonant X-ray diffraction\nat ambient pressure. The magnetic order is characterized by a commensurate\npropagation vector ${k}_{1/2}~=~\\left( \\frac{1}{2} , \\frac{1}{2},\n\\frac{1}{2}\\right)$ with spins lying in the basal plane. Our measurements did\nnot reveal the presence of an incommensurate order propagating along the high\nsymmetry directions in reciprocal space but cannot exclude other incommensurate\nmodulations or weak scattering intensities. The observed commensurate order can\nbe described equivalently by either a single-${k}$ structure or by a\nmulti-${k}$ structure. Furthermore we explain how a commensurate-only ordering\nmay explain the broad distribution of internal fields observed in nuclear\nquadrupolar resonance experiments (Sakai et al. 2011, Phys. Rev. B 83 140408)\nthat was previously attributed to an incommensurate order. We also report\npowder X-ray diffraction showing that the crystallographic structure of\nCePt$_2$In$_7$ changes monotonically with pressure up to $P~=~7.3$ GPa at room\ntemperature. The determined bulk modulus $B_0~=~81.1(3)$ GPa is similar to the\nones of the Ce-115 family. Broad diffraction peaks confirm the presence of\npronounced strain in polycrystalline samples of CePt$_2$In$_7$. We discuss how\nstrain effects can lead to different electronic and magnetic properties between\npolycrystalline and single crystal samples.\n",
"title": "Investigation of the commensurate magnetic structure in heavy fermion CePt2In7 using magnetic resonant X-ray diffraction"
}
| null | null |
[
"Physics"
] | null | true | null |
13758
| null |
Validated
| null | null |
null |
{
"abstract": " One of the most compelling features of Gaussian process (GP) regression is\nits ability to provide well-calibrated posterior distributions. Recent advances\nin inducing point methods have sped up GP marginal likelihood and posterior\nmean computations, leaving posterior covariance estimation and sampling as the\nremaining computational bottlenecks. In this paper we address these\nshortcomings by using the Lanczos algorithm to rapidly approximate the\npredictive covariance matrix. Our approach, which we refer to as LOVE (LanczOs\nVariance Estimates), substantially improves time and space complexity. In our\nexperiments, LOVE computes covariances up to 2,000 times faster and draws\nsamples 18,000 times faster than existing methods, all without sacrificing\naccuracy.\n",
"title": "Constant-Time Predictive Distributions for Gaussian Processes"
}
| null | null | null | null | true | null |
13759
| null |
Default
| null | null |
null |
{
"abstract": " We consider the global optimization of a function over a continuous domain.\nAt every evaluation attempt, we can observe the function at a chosen point in\nthe domain and we reap the reward of the value observed. We assume that drawing\nthese observations are expensive and noisy. We frame it as a continuum-armed\nbandit problem with a Gaussian Process prior on the function. In this regime,\nmost algorithms have been developed to minimize some form of regret. Contrary\nto this popular norm, in this paper, we study the convergence of the sequential\npoint $\\boldsymbol{x}^t$ to the global optimizer $\\boldsymbol{x}^*$ for the\nThompson Sampling approach. Under some assumptions and regularity conditions,\nwe show an exponential rate of convergence to the true optimal.\n",
"title": "Analysis of Thompson Sampling for Gaussian Process Optimization in the Bandit Setting"
}
| null | null | null | null | true | null |
13760
| null |
Default
| null | null |
null |
{
"abstract": " Over the years, many multiprocessor locking protocols have been designed and\nanalyzed. However, the performance of these protocols highly depends on how the\ntasks are partitioned and prioritized and how the resources are shared locally\nand globally. This paper answers a few fundamental questions when real-time\ntasks share resources in multiprocessor systems. We explore the fundamental\ndifficulty of the multiprocessor synchronization problem and show that a very\nsimplified version of this problem is ${\\mathcal NP}$-hard in the strong sense\nregardless of the number of processors and the underlying scheduling paradigm.\nTherefore, the allowance of preemption or migration does not reduce the\ncomputational complexity. For the positive side, we develop a dependency-graph\napproach, that is specifically useful for frame-based real-time tasks, in which\nall tasks have the same period and release their jobs always at the same time.\nWe present a series of algorithms with speedup factors between $2$ and $3$\nunder semi-partitioned scheduling. We further explore methodologies and\ntradeoffs of preemptive against non-preemptive scheduling algorithms and\npartitioned against semi-partitioned scheduling algorithms. The approach is\nextended to periodic tasks under certain conditions.\n",
"title": "Dependency Graph Approach for Multiprocessor Real-Time Synchronization"
}
| null | null | null | null | true | null |
13761
| null |
Default
| null | null |
null |
{
"abstract": " Topological cyclic homology is a refinement of Connes--Tsygan's cyclic\nhomology which was introduced by Bökstedt--Hsiang--Madsen in 1993 as an\napproximation to algebraic $K$-theory. There is a trace map from algebraic\n$K$-theory to topological cyclic homology, and a theorem of\nDundas--Goodwillie--McCarthy asserts that this induces an equivalence of\nrelative theories for nilpotent immersions, which gives a way for computing\n$K$-theory in various situations. The construction of topological cyclic\nhomology is based on genuine equivariant homotopy theory, the use of explicit\npoint-set models, and the elaborate notion of a cyclotomic spectrum.\nThe goal of this paper is to revisit this theory using only\nhomotopy-invariant notions. In particular, we give a new construction of\ntopological cyclic homology. This is based on a new definition of the\n$\\infty$-category of cyclotomic spectra: We define a cyclotomic spectrum to be\na spectrum $X$ with $S^1$-action (in the most naive sense) together with\n$S^1$-equivariant maps $\\varphi_p: X\\to X^{tC_p}$ for all primes $p$. Here\n$X^{tC_p}=\\mathrm{cofib}(\\mathrm{Nm}: X_{hC_p}\\to X^{hC_p})$ is the Tate\nconstruction. On bounded below spectra, we prove that this agrees with previous\ndefinitions. As a consequence, we obtain a new and simple formula for\ntopological cyclic homology.\nIn order to construct the maps $\\varphi_p: X\\to X^{tC_p}$ in the example of\ntopological Hochschild homology we introduce and study Tate diagonals for\nspectra and Frobenius homomorphisms of commutative ring spectra. In particular\nwe prove a version of the Segal conjecture for the Tate diagonals and relate\nthese Frobenius homomorphisms to power operations.\n",
"title": "On topological cyclic homology"
}
| null | null | null | null | true | null |
13762
| null |
Default
| null | null |
null |
{
"abstract": " Local graph partitioning is a key graph mining tool that allows researchers\nto identify small groups of interrelated nodes (e.g. people) and their\nconnective edges (e.g. interactions). Because local graph partitioning is\nprimarily focused on the network structure of the graph (vertices and edges),\nit often fails to consider the additional information contained in the\nattributes. In this paper we propose---(i) a scalable algorithm to improve\nlocal graph partitioning by taking into account both the network structure of\nthe graph and the attribute data and (ii) an application of the proposed local\ngraph partitioning algorithm (AttriPart) to predict the evolution of local\ncommunities (LocalForecasting). Experimental results show that our proposed\nAttriPart algorithm finds up to 1.6$\\times$ denser local partitions, while\nrunning approximately 43$\\times$ faster than traditional local partitioning\ntechniques (PageRank-Nibble). In addition, our LocalForecasting algorithm shows\na significant improvement in the number of nodes and edges correctly predicted\nover baseline methods.\n",
"title": "Local Partition in Rich Graphs"
}
| null | null | null | null | true | null |
13763
| null |
Default
| null | null |
null |
{
"abstract": " Understanding the detailed queueing behavior of a networking session is\ncritical in enabling low-latency services over the Internet. Especially when\nthe packet arrival and service rates at the queue of a link vary over time and\nmoreover when the session is short-lived, analyzing the corresponding queue\nbehavior as a function of time, which involves a transient analysis, becomes\nextremely challenging. In this paper, we propose and develop a new analytical\nframework that anatomizes the transient queue behavior under time-varying\narrival and service rates even under unstable conditions. Our framework is\ncapable of answering key questions in designing low-latency services such as\nthe time-dependent probability distribution of the queue length; the\ninstantaneous or time-averaged violation probability that the queue length\nexceeds a certain threshold; and the fraction of time during an interval $[0,\nt]$ at which the queue length exceeds a certain threshold. We validate our\nframework by comparing its prediction results over time with the statistical\nsimulation results and confirm that our analysis is accurate enough. Our\nextensive demonstrations on the efficacy of the analytical framework in\ndesigning low-latency services reveal that its prediction ability for the\ntransient queue behavior in diverse time-varying packet arrival and service\npatterns can be of a high practical value.\n",
"title": "A Transient Queueing Analysis under Time-varying Arrival and Service Rates for Enabling Low-Latency Services"
}
| null | null |
[
"Computer Science"
] | null | true | null |
13764
| null |
Validated
| null | null |
null |
{
"abstract": " Although cluttered indoor scenes have a lot of useful high-level semantic\ninformation which can be used for mapping and localization, most Visual\nOdometry (VO) algorithms rely on the usage of geometric features such as\npoints, lines and planes. Lately, driven by this idea, the joint optimization\nof semantic labels and obtaining odometry has gained popularity in the robotics\ncommunity. The joint optimization is good for accurate results but is generally\nvery slow. At the same time, in the vision community, direct and sparse\napproaches for VO have stricken the right balance between speed and accuracy.\nWe merge the successes of these two communities and present a way to\nincorporate semantic information in the form of visual saliency to Direct\nSparse Odometry - a highly successful direct sparse VO algorithm. We also\npresent a framework to filter the visual saliency based on scene parsing. Our\nframework, SalientDSO, relies on the widely successful deep learning based\napproaches for visual saliency and scene parsing which drives the feature\nselection for obtaining highly-accurate and robust VO even in the presence of\nas few as 40 point features per frame. We provide extensive quantitative\nevaluation of SalientDSO on the ICL-NUIM and TUM monoVO datasets and show that\nwe outperform DSO and ORB-SLAM - two very popular state-of-the-art approaches\nin the literature. We also collect and publicly release a CVL-UMD dataset which\ncontains two indoor cluttered sequences on which we show qualitative\nevaluations. To our knowledge this is the first paper to use visual saliency\nand scene parsing to drive the feature selection in direct VO.\n",
"title": "SalientDSO: Bringing Attention to Direct Sparse Odometry"
}
| null | null | null | null | true | null |
13765
| null |
Default
| null | null |
null |
{
"abstract": " Most commonly used distributed machine learning systems are either\nsynchronous or centralized asynchronous. Synchronous algorithms like\nAllReduce-SGD perform poorly in a heterogeneous environment, while asynchronous\nalgorithms using a parameter server suffer from 1) communication bottleneck at\nparameter servers when workers are many, and 2) significantly worse convergence\nwhen the traffic to parameter server is congested. Can we design an algorithm\nthat is robust in a heterogeneous environment, while being communication\nefficient and maintaining the best-possible convergence rate? In this paper, we\npropose an asynchronous decentralized stochastic gradient decent algorithm\n(AD-PSGD) satisfying all above expectations. Our theoretical analysis shows\nAD-PSGD converges at the optimal $O(1/\\sqrt{K})$ rate as SGD and has linear\nspeedup w.r.t. number of workers. Empirically, AD-PSGD outperforms the best of\ndecentralized parallel SGD (D-PSGD), asynchronous parallel SGD (A-PSGD), and\nstandard data parallel SGD (AllReduce-SGD), often by orders of magnitude in a\nheterogeneous environment. When training ResNet-50 on ImageNet with up to 128\nGPUs, AD-PSGD converges (w.r.t epochs) similarly to the AllReduce-SGD, but each\nepoch can be up to 4-8X faster than its synchronous counterparts in a\nnetwork-sharing HPC environment. To the best of our knowledge, AD-PSGD is the\nfirst asynchronous algorithm that achieves a similar epoch-wise convergence\nrate as AllReduce-SGD, at an over 100-GPU scale.\n",
"title": "Asynchronous Decentralized Parallel Stochastic Gradient Descent"
}
| null | null | null | null | true | null |
13766
| null |
Default
| null | null |
null |
{
"abstract": " The prevalence of smart wearable devices is increasing exponentially and we\nare witnessing a wide variety of fascinating new services that leverage the\ncapabilities of these wearables. Wearables are truly changing the way mobile\ncomputing is deployed and mobile applications are being developed. It is\npossible to leverage the capabilities such as connectivity, processing, and\nsensing of wearable devices in an adaptive manner for efficient resource usage\nand information accuracy within the personal area network. We show that\napplication developers are not yet taking advantage of these cross-device\ncapabilities, however, instead using wearables as passive sensors or simple end\ndisplays to provide notifications to the user. We thus design AFV (Application\nFunction Virtualization), an architecture enabling automated dynamic function\nvirtualization and scheduling across devices in a personal area network,\nsimplifying the development of the apps that are adaptive to context changes.\nAFV provides a simple set of APIs hiding complex architectural tasks from app\ndevelopers whilst continuously monitoring the user, device and network context,\nto enable the adaptive invocation of functions across devices. We show the\nfeasibility of our design by implementing AFV on Android, and the benefits for\nthe user in terms of resource efficiency, especially in saving energy\nconsumption, and quality of experience with multiple use cases.\n",
"title": "Seamless Resources Sharing in Wearable Networks by Application Function Virtualization"
}
| null | null | null | null | true | null |
13767
| null |
Default
| null | null |
null |
{
"abstract": " Q-Ensembles are a model-free approach where input images are fed into\ndifferent Q-networks and exploration is driven by the assumption that\nuncertainty is proportional to the variance of the output Q-values obtained.\nThey have been shown to perform relatively well compared to other exploration\nstrategies. Further, model-based approaches, such as encoder-decoder models\nhave been used successfully for next frame prediction given previous frames.\nThis paper proposes to integrate the model-free Q-ensembles and model-based\napproaches with the hope of compounding the benefits of both and achieving\nsuperior exploration as a result. Results show that a model-based trajectory\nmemory approach when combined with Q-ensembles produces superior performance\nwhen compared to only using Q-ensembles.\n",
"title": "Combining Model-Free Q-Ensembles and Model-Based Approaches for Informed Exploration"
}
| null | null |
[
"Statistics"
] | null | true | null |
13768
| null |
Validated
| null | null |
null |
{
"abstract": " We report on the status of our Cybersecurity Assessment Tools (CATS) project\nthat is creating and validating a concept inventory for cybersecurity, which\nassesses the quality of instruction of any first course in cybersecurity. In\nfall 2014, we carried out a Delphi process that identified core concepts of\ncybersecurity. In spring 2016, we interviewed twenty-six students to uncover\ntheir understandings and misconceptions about these concepts. In fall 2016, we\ngenerated our first assessment tool--a draft Cybersecurity Concept Inventory\n(CCI), comprising approximately thirty multiple-choice questions. Each question\ntargets a concept; incorrect answers are based on observed misconceptions from\nthe interviews. This year we are validating the draft CCI using cognitive\ninterviews, expert reviews, and psychometric testing. In this paper, we\nhighlight our progress to date in developing the CCI.\nThe CATS project provides infrastructure for a rigorous evidence-based\nimprovement of cybersecurity education. The CCI permits comparisons of\ndifferent instructional methods by assessing how well students learned the core\nconcepts of the field (especially adversarial thinking), where instructional\nmethods refer to how material is taught (e.g., lab-based, case-studies,\ncollaborative, competitions, gaming). Specifically, the CCI is a tool that will\nenable researchers to scientifically quantify and measure the effect of their\napproaches to, and interventions in, cybersecurity education.\n",
"title": "Creating a Cybersecurity Concept Inventory: A Status Report on the CATS Project"
}
| null | null | null | null | true | null |
13769
| null |
Default
| null | null |
null |
{
"abstract": " Critical periods are phases in the early development of humans and animals\nduring which experience can irreversibly affect the architecture of neuronal\nnetworks. In this work, we study the effects of visual stimulus deficits on the\ntraining of artificial neural networks (ANNs). Introducing well-characterized\nvisual deficits, such as cataract-like blurring, in the early training phase of\na standard deep neural network causes a permanent performance loss that closely\nmimics critical period behavior in humans and animal models. Deficits that do\nnot affect low-level image statistics, such as vertical flipping of the images,\nhave no lasting effect on the ANNs' performance and can be rapidly overcome\nwith further training. In addition, the deeper the ANN is, the more pronounced\nthe critical period. To better understand this phenomenon, we use Fisher\nInformation as a measure of the strength of the network's connections during\nthe training. Our information-theoretic analysis suggests that the first few\nepochs are critical for the creation of strong connections across different\nlayers, optimal for processing the input data distribution. Once such strong\nconnections are created, they do not appear to change during additional\ntraining. These findings suggest that the initial rapid learning phase of ANN\ntraining, under-scrutinized compared to its asymptotic behavior, plays a key\nrole in defining the final performance of networks. Our results also show how\ncritical periods are not restricted to biological systems, but can emerge\nnaturally in learning systems, whether biological or artificial, due to\nfundamental constrains arising from learning dynamics and information\nprocessing.\n",
"title": "Critical Learning Periods in Deep Neural Networks"
}
| null | null | null | null | true | null |
13770
| null |
Default
| null | null |
null |
{
"abstract": " In this work we present a whole-body Nonlinear Model Predictive Control\napproach for Rigid Body Systems subject to contacts. We use a full dynamic\nsystem model which also includes explicit contact dynamics. Therefore, contact\nlocations, sequences and timings are not prespecified but optimized by the\nsolver. Yet, thorough numerical and software engineering allows for running the\nnonlinear Optimal Control solver at rates up to 190 Hz on a quadruped for a\ntime horizon of half a second. This outperforms the state of the art by at\nleast one order of magnitude. Hardware experiments in form of periodic and\nnon-periodic tasks are applied to two quadrupeds with different actuation\nsystems. The obtained results underline the performance, transferability and\nrobustness of the approach.\n",
"title": "Whole-Body Nonlinear Model Predictive Control Through Contacts for Quadrupeds"
}
| null | null | null | null | true | null |
13771
| null |
Default
| null | null |
null |
{
"abstract": " Unmanned aerial vehicles (UAVs) represent a new frontier in a wide range of\nmonitoring and research applications. To fully leverage their potential, a key\nchallenge is planning missions for efficient data acquisition in complex\nenvironments. To address this issue, this article introduces a general\ninformative path planning (IPP) framework for monitoring scenarios using an\naerial robot. The approach is capable of mapping either discrete or continuous\ntarget variables on a terrain using variable-resolution data received from\nprobabilistic sensors. During a mission, the terrain maps built online are used\nto plan information-rich trajectories in continuous 3-D space by optimizing\ninitial solutions obtained by a course grid search. Extensive simulations show\nthat our approach is more efficient than existing methods. We also demonstrate\nits real-time application on a photorealistic mapping scenario using a publicly\navailable dataset.\n",
"title": "An informative path planning framework for UAV-based terrain monitoring"
}
| null | null | null | null | true | null |
13772
| null |
Default
| null | null |
null |
{
"abstract": " Recurrent Neural Networks (RNN) are a type of statistical model designed to\nhandle sequential data. The model reads a sequence one symbol at a time. Each\nsymbol is processed based on information collected from the previous symbols.\nWith existing RNN architectures, each symbol is processed using only\ninformation from the previous processing step. To overcome this limitation, we\npropose a new kind of RNN model that computes a recurrent weighted average\n(RWA) over every past processing step. Because the RWA can be computed as a\nrunning average, the computational overhead scales like that of any other RNN\narchitecture. The approach essentially reformulates the attention mechanism\ninto a stand-alone model. The performance of the RWA model is assessed on the\nvariable copy problem, the adding problem, classification of artificial\ngrammar, classification of sequences by length, and classification of the MNIST\nimages (where the pixels are read sequentially one at a time). On almost every\ntask, the RWA model is found to outperform a standard LSTM model.\n",
"title": "Machine Learning on Sequential Data Using a Recurrent Weighted Average"
}
| null | null | null | null | true | null |
13773
| null |
Default
| null | null |
null |
{
"abstract": " Functionals of a stochastic process Y(t) model many physical time-extensive\nobservables, e.g. particle positions, local and occupation times or accumulated\nmechanical work. When Y(t) is a normal diffusive process, their statistics are\nobtained as the solution of the Feynman-Kac equation. This equation provides\nthe crucial link between the expected values of diffusion processes and the\nsolutions of deterministic second-order partial differential equations. When\nY(t) is an anomalous diffusive process, generalizations of the Feynman-Kac\nequation that incorporate power-law or more general waiting time distributions\nof the underlying random walk have recently been derived. A general\nrepresentation of such waiting times is provided in terms of a Lévy process\nwhose Laplace exponent is related to the memory kernel appearing in the\ngeneralized Feynman-Kac equation. The corresponding anomalous processes have\nbeen shown to capture nonlinear mean square displacements exhibiting crossovers\nbetween different scaling regimes, which have been observed in biological\nsystems like migrating cells or diffusing macromolecules in intracellular\nenvironments. However, the case where both space- and time-dependent forces\ndrive the dynamics of the generalized anomalous process has not been solved\nyet. Here, we present the missing derivation of the Feynman-Kac equation in\nsuch general case by using the subordination technique. Furthermore, we discuss\nits extension to functionals explicitly depending on time, which are relevant\nfor the stochastic thermodynamics of anomalous diffusive systems. Exact results\non the work fluctuations of a simple non-equilibrium model are obtained. In\nthis paper we also provide a pedagogical introduction to Lévy processes,\nsemimartingales and their associated stochastic calculus, which underlie the\nmathematical formulation of anomalous diffusion as a subordinated process.\n",
"title": "Feynman-Kac equation for anomalous processes with space- and time-dependent forces"
}
| null | null | null | null | true | null |
13774
| null |
Default
| null | null |
null |
{
"abstract": " As efficient traffic-management platforms, public vehicle (PV) systems are\nenvisioned to be a promising approach to solving traffic congestions and\npollutions for future smart cities. PV systems provide online/dynamic\npeer-to-peer ride-sharing services with the goal of serving sufficient number\nof customers with minimum number of vehicles and lowest possible cost. A key\ncomponent of the PV system is the online ride-sharing scheduling strategy. In\nthis paper, we propose an efficient path planning strategy that focuses on a\nlimited potential search area for each vehicle by filtering out the requests\nthat violate passenger service quality level, so that the global search is\nreduced to local search. We analyze the performance of the proposed solution\nsuch as reduction ratio of computational complexity. Simulations based on the\nManhattan taxi data set show that, the computing time is reduced by 22%\ncompared with the exhaustive search method under the same service quality\nperformance.\n",
"title": "An Online Ride-Sharing Path Planning Strategy for Public Vehicle Systems"
}
| null | null | null | null | true | null |
13775
| null |
Default
| null | null |
null |
{
"abstract": " We frame Question Answering (QA) as a Reinforcement Learning task, an\napproach that we call Active Question Answering. We propose an agent that sits\nbetween the user and a black box QA system and learns to reformulate questions\nto elicit the best possible answers. The agent probes the system with,\npotentially many, natural language reformulations of an initial question and\naggregates the returned evidence to yield the best answer. The reformulation\nsystem is trained end-to-end to maximize answer quality using policy gradient.\nWe evaluate on SearchQA, a dataset of complex questions extracted from\nJeopardy!. The agent outperforms a state-of-the-art base model, playing the\nrole of the environment, and other benchmarks. We also analyze the language\nthat the agent has learned while interacting with the question answering\nsystem. We find that successful question reformulations look quite different\nfrom natural language paraphrases. The agent is able to discover non-trivial\nreformulation strategies that resemble classic information retrieval techniques\nsuch as term re-weighting (tf-idf) and stemming.\n",
"title": "Ask the Right Questions: Active Question Reformulation with Reinforcement Learning"
}
| null | null | null | null | true | null |
13776
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we present results on dynamic multivariate scalar risk\nmeasures, which arise in markets with transaction costs and systemic risk. Dual\nrepresentations of such risk measures are presented. These are then used to\nobtain the main results of this paper on time consistency; namely, an\nequivalent recursive formulation of multivariate scalar risk measures to\nmultiportfolio time consistency. We are motivated to study time consistency of\nmultivariate scalar risk measures as the superhedging risk measure in markets\nwith transaction costs (with a single eligible asset) (Jouini and Kallal\n(1995), Roux and Zastawniak (2016), Loehne and Rudloff (2014)) does not satisfy\nthe usual scalar concept of time consistency. In fact, as demonstrated in\n(Feinstein and Rudloff (2018)), scalar risk measures with the same\nscalarization weight at all times would not be time consistent in general. The\ndeduced recursive relation for the scalarizations of multiportfolio time\nconsistent set-valued risk measures provided in this paper requires\nconsideration of the entire family of scalarizations. In this way we develop a\ndirect notion of a \"moving scalarization\" for scalar time consistency that\ncorroborates recent research on scalarizations of dynamic multi-objective\nproblems (Karnam, Ma, and Zhang (2017), Kovacova and Rudloff (2018)).\n",
"title": "Time consistency for scalar multivariate risk measures"
}
| null | null | null | null | true | null |
13777
| null |
Default
| null | null |
null |
{
"abstract": " The design, simulation and measurement of a beam steerable slotted waveguide\nantenna operating in X band are presented. The proposed beam steerable antenna\nconsists of a standard rectangular waveguide (RWG) section with longitudinal\nslots in the broad wall. The beam steering in this configuration is achieved by\nrotating two dielectric slabs inside the waveguide and consequently changing\nthe phase of the slots excitations. In order to confirm the usefulness of this\nconcept, a non-resonant 20-slot waveguide array antenna with an element spacing\nof d = 0.58{\\lambda}0 has been designed, built and measured. A 14 deg beam\nscanning from near broadside ({\\theta} = 4 deg) toward end-fire ({\\theta} = 18\ndeg) direction is observed. The gain varies from 18.33 dB to 19.11 dB which\ncorresponds to the radiation efficiencies between 95% and 79%. The side-lobe\nlevel is -14 dB at the design frequency of 9.35 GHz. The simulated co-polarized\nrealized gain closely matches the fabricated prototype patterns.\n",
"title": "A Continuous Beam Steering Slotted Waveguide Antenna Using Rotating Dielectric Slabs"
}
| null | null | null | null | true | null |
13778
| null |
Default
| null | null |
null |
{
"abstract": " The Internet of Mobile Things encompasses stream data being generated by\nsensors, network communications that pull and push these data streams, as well\nas running processing and analytics that can effectively leverage actionable\ninformation for transportation planning, management, and business advantage.\nEdge computing emerges as a new paradigm that decentralizes the communication,\ncomputation, control and storage resources from the cloud to the edge of the\nnetwork. This paper proposes an edge computing platform where mobile edge nodes\nare physical devices deployed on a transit bus where descriptive analytics is\nused to uncover meaningful patterns from real-time transit data streams. An\napplication experiment is used to evaluate the advantages and disadvantages of\nour proposed platform to support descriptive analytics at a mobile edge node\nand generate actionable information to transit managers.\n",
"title": "Developing an edge computing platform for real-time descriptive analytics"
}
| null | null | null | null | true | null |
13779
| null |
Default
| null | null |
null |
{
"abstract": " During the Space Telescope and Optical Reverberation Mapping Project (STORM)\nobservations of NGC 5548, the continuum and emission-line variability became\nde-correlated during the second half of the 6-month long observing campaign.\nHere we present Swift and Chandra X-ray spectra of NGC 5548 obtained as a part\nof the campaign. The Swift spectra show that excess flux (relative to a\npower-law continuum) in the soft X-ray band appears before the start of the\nanomalous emission-line behavior, peaks during the period of the anomaly, and\nthen declines. This is a model-independent result suggesting that the soft\nexcess is related to the anomaly. We divide the Swift data into on- and\noff-anomaly spectra to characterize the soft excess via spectral fitting. The\ncause of the spectral differences is likely due to a change in the intrinsic\nspectrum rather than being due to variable obscuration or partial covering. The\nChandra spectra have lower signal-to-noise ratios, but are consistent with\nSwift data. Our preferred model of the soft excess is emission from an\noptically thick, warm Comptonizing corona, the effective optical depth of which\nincreases during the anomaly. This model simultaneously explains all the three\nobservations: the UV emission line flux decrease, the soft-excess increase, and\nthe emission line anomaly.\n",
"title": "Space Telescope and Optical Reverberation Mapping Project. VII. Understanding the UV anomaly in NGC 5548 with X-Ray Spectroscopy"
}
| null | null | null | null | true | null |
13780
| null |
Default
| null | null |
null |
{
"abstract": " We present the grasping system and design approach behind Cartman, the\nwinning entrant in the 2017 Amazon Robotics Challenge. We investigate the\ndesign processes leading up to the final iteration of the system and describe\nthe emergent solution by comparing it with key robotics design aspects.\nFollowing our experience, we propose a new design aspect, precision vs.\nredundancy, that should be considered alongside the previously proposed design\naspects of modularity vs. integration, generality vs. assumptions, computation\nvs. embodiment and planning vs. feedback. We present the grasping system behind\nCartman, the winning robot in the 2017 Amazon Robotics Challenge. The system\nmakes strong use of redundancy in design by implementing complimentary tools, a\nsuction gripper and a parallel gripper. This multi-modal end-effector is\ncombined with three grasp synthesis algorithms to accommodate the range of\nobjects provided by Amazon during the challenge. We provide a detailed system\ndescription and an evaluation of its performance before discussing the broader\nnature of the system with respect to the key aspects of robotic design as\ninitially proposed by the winners of the first Amazon Picking Challenge. To\naddress the principal nature of our grasping system and the reason for its\nsuccess, we propose an additional robotic design aspect `precision vs.\nredundancy'. The full design of our robotic system, including the end-effector,\nis open sourced and available at\nthis http URL\n",
"title": "Design of a Multi-Modal End-Effector and Grasping System: How Integrated Design helped win the Amazon Robotics Challenge"
}
| null | null |
[
"Computer Science"
] | null | true | null |
13781
| null |
Validated
| null | null |
null |
{
"abstract": " We consider the semantics of prepositions, revisiting a broad-coverage\nannotation scheme used for annotating all 4,250 preposition tokens in a 55,000\nword corpus of English. Attempts to apply the scheme to adpositions and case\nmarkers in other languages, as well as some problematic cases in English, have\nled us to reconsider the assumption that a preposition's lexical contribution\nis equivalent to the role/relation that it mediates. Our proposal is to embrace\nthe potential for construal in adposition use, expressing such phenomena\ndirectly at the token level to manage complexity and avoid sense proliferation.\nWe suggest a framework to represent both the scene role and the adposition's\nlexical function so they can be annotated at scale---supporting automatic,\nstatistical processing of domain-general language---and sketch how this\nrepresentation would inform a constructional analysis.\n",
"title": "Coping with Construals in Broad-Coverage Semantic Annotation of Adpositions"
}
| null | null | null | null | true | null |
13782
| null |
Default
| null | null |
null |
{
"abstract": " Random walks are at the heart of many existing deep learning algorithms for\ngraph data. However, such algorithms have many limitations that arise from the\nuse of random walks, e.g., the features resulting from these methods are unable\nto transfer to new nodes and graphs as they are tied to node identity. In this\nwork, we introduce the notion of attributed random walks which serves as a\nbasis for generalizing existing methods such as DeepWalk, node2vec, and many\nothers that leverage random walks. Our proposed framework enables these methods\nto be more widely applicable for both transductive and inductive learning as\nwell as for use on graphs with attributes (if available). This is achieved by\nlearning functions that generalize to new nodes and graphs. We show that our\nproposed framework is effective with an average AUC improvement of 16.1% while\nrequiring on average 853 times less space than existing methods on a variety of\ngraphs from several domains.\n",
"title": "A Framework for Generalizing Graph-based Representation Learning Methods"
}
| null | null | null | null | true | null |
13783
| null |
Default
| null | null |
null |
{
"abstract": " Gated Recurrent Unit (GRU) is a recently-developed variation of the long\nshort-term memory (LSTM) unit, both of which are types of recurrent neural\nnetwork (RNN). Through empirical evidence, both models have been proven to be\neffective in a wide variety of machine learning tasks such as natural language\nprocessing (Wen et al., 2015), speech recognition (Chorowski et al., 2015), and\ntext classification (Yang et al., 2016). Conventionally, like most neural\nnetworks, both of the aforementioned RNN variants employ the Softmax function\nas its final output layer for its prediction, and the cross-entropy function\nfor computing its loss. In this paper, we present an amendment to this norm by\nintroducing linear support vector machine (SVM) as the replacement for Softmax\nin the final output layer of a GRU model. Furthermore, the cross-entropy\nfunction shall be replaced with a margin-based function. While there have been\nsimilar studies (Alalshekmubarak & Smith, 2013; Tang, 2013), this proposal is\nprimarily intended for binary classification on intrusion detection using the\n2013 network traffic data from the honeypot systems of Kyoto University.\nResults show that the GRU-SVM model performs relatively higher than the\nconventional GRU-Softmax model. The proposed model reached a training accuracy\nof ~81.54% and a testing accuracy of ~84.15%, while the latter was able to\nreach a training accuracy of ~63.07% and a testing accuracy of ~70.75%. In\naddition, the juxtaposition of these two final output layers indicate that the\nSVM would outperform Softmax in prediction time - a theoretical implication\nwhich was supported by the actual training and testing time in the study.\n",
"title": "A Neural Network Architecture Combining Gated Recurrent Unit (GRU) and Support Vector Machine (SVM) for Intrusion Detection in Network Traffic Data"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
13784
| null |
Validated
| null | null |
null |
{
"abstract": " In this paper, we reformulated the spell correction problem as a machine\ntranslation task under the encoder-decoder framework. This reformulation\nenabled us to use a single model for solving the problem that is traditionally\nformulated as learning a language model and an error model. This model employs\nmulti-layer recurrent neural networks as an encoder and a decoder. We\ndemonstrate the effectiveness of this model using an internal dataset, where\nthe training data is automatically obtained from user logs. The model offers\ncompetitive performance as compared to the state of the art methods but does\nnot require any feature engineering nor hand tuning between models.\n",
"title": "Spelling Correction as a Foreign Language"
}
| null | null | null | null | true | null |
13785
| null |
Default
| null | null |
null |
{
"abstract": " Two-dimensional (2D) transition metal dichalcogenides (TMDs) have recently\nemerged as promising candidates for future electronics and optoelectronics.\nWhile most of TMDs are intrinsic n-type semiconductors due to electron donating\nwhich originates from chalcogen vacancies, obtaining intrinsic high-quality\np-type semiconducting TMDs has been challenging. Here, we report an\nexperimental approach to obtain intrinsic p-type Tungsten (W)-based TMDs by\nsubstitutional Ta-doping. The obtained few-layer Ta-doped WSe2 (Ta0.01W0.99Se2)\nfield-effect transistor (FET) devices exhibit competitive p-type performances,\nincluding ~10^6 current on/off at room temperature. We also demonstrate high\nquality van der Waals (vdW) p-n heterojunctions based on Ta0.01W0.99Se2/MoS2\nstructure, which exhibit nearly ideal diode characteristics (with an ideality\nfactor approaching 1 and a rectification ratio up to 10^5) and excellent\nphotodetecting performance. Our study suggests that substitutional Ta-doping\nholds great promise to realize intrinsic p-type W-based TMDs for future\nelectronic and photonic applications.\n",
"title": "Intrinsic p-type W-based transition metal dichalcogenide by substitutional Ta-doping"
}
| null | null | null | null | true | null |
13786
| null |
Default
| null | null |
null |
{
"abstract": " In this work we study permutation synchronisation for the challenging case of\npartial permutations, which plays an important role for the problem of matching\nmultiple objects (e.g. images or shapes). The term synchronisation refers to\nthe property that the set of pairwise matchings is cycle-consistent, i.e. in\nthe full matching case all compositions of pairwise matchings over cycles must\nbe equal to the identity. Motivated by clustering and matrix factorisation\nperspectives of cycle-consistency, we derive an algorithm to tackle the\npermutation synchronisation problem based on non-negative factorisations. In\norder to deal with the inherent non-convexity of the permutation\nsynchronisation problem, we use an initialisation procedure based on a novel\nrotation scheme applied to the solution of the spectral relaxation. Moreover,\nthis rotation scheme facilitates a convenient Euclidean projection to obtain a\nbinary solution after solving our relaxed problem. In contrast to\nstate-of-the-art methods, our approach is guaranteed to produce\ncycle-consistent results. We experimentally demonstrate the efficacy of our\nmethod and show that it achieves better results compared to existing methods.\n",
"title": "Synchronisation of Partial Multi-Matchings via Non-negative Factorisations"
}
| null | null | null | null | true | null |
13787
| null |
Default
| null | null |
null |
{
"abstract": " Associated varieties of vertex algebras are analogue of the associated\nvarieties of primitive ideals of the universal enveloping algebras of\nsemisimple Lie algebras. They not only capture some of the important properties\nof vertex algebras but also have interesting relationship with the Higgs\nbranches of four-dimensional $N=2$ superconformal field theories (SCFTs). As a\nconsequence, one can deduce the modular invariance of Schur indices of 4d $N=2$\nSCFTs from the theory of vertex algebras.\n",
"title": "Associated varieties and Higgs branches (a survey)"
}
| null | null |
[
"Mathematics"
] | null | true | null |
13788
| null |
Validated
| null | null |
null |
{
"abstract": " A new method is developed to deal with the problem that a complex\ndecentralized control system needs to keep centralized control performance. The\nsystematic procedure emphasizes quickly finding the decentralized\nsubcontrollers that matching the closed-loop performance and robustness\ncharacteristics of the centralized controller, which is featured by the fact\nthat GA is used to optimize the design of centralized H-infinity controller\nK(s) and decentralized engine subcontroller KT(s), and that only one interface\nvariable needs to satisfy decentralized control system requirement according to\nthe proposed selection principle. The optimization design is motivated by the\nimplementation issues where it is desirable to reduce the time in trial and\nerror process and accurately find the best decentralized subcontrollers. The\nmethod is applied to decentralized control system design for a short takeoff\nand landing fighter. By comparing the simulation results of the decentralized\ncontrol system with those of the centralized control system, the target of the\ndecentralized control attains the performance and robustness of centralized\ncontrol is validated.\n",
"title": "Optimization Design of Decentralized Control for Complex Decentralized Systems"
}
| null | null |
[
"Computer Science"
] | null | true | null |
13789
| null |
Validated
| null | null |
null |
{
"abstract": " This paper presents a Semantic Attribute Modulation (SAM) for language\nmodeling and style variation. The semantic attribute modulation includes\nvarious document attributes, such as titles, authors, and document categories.\nWe consider two types of attributes, (title attributes and category\nattributes), and a flexible attribute selection scheme by automatically scoring\nthem via an attribute attention mechanism. The semantic attributes are embedded\ninto the hidden semantic space as the generation inputs. With the attributes\nproperly harnessed, our proposed SAM can generate interpretable texts with\nregard to the input attributes. Qualitative analysis, including word semantic\nanalysis and attention values, shows the interpretability of SAM. On several\ntypical text datasets, we empirically demonstrate the superiority of the\nSemantic Attribute Modulated language model with different combinations of\ndocument attributes. Moreover, we present a style variation for the lyric\ngeneration using SAM, which shows a strong connection between the style\nvariation and the semantic attributes.\n",
"title": "SAM: Semantic Attribute Modulation for Language Modeling and Style Variation"
}
| null | null | null | null | true | null |
13790
| null |
Default
| null | null |
null |
{
"abstract": " Positive--unlabeled (PU) learning considers two samples, a positive set P\nwith observations from only one class and an unlabeled set U with observations\nfrom two classes. The goal is to classify observations in U. Class mixture\nproportion estimation (MPE) in U is a key step in PU learning. Blanchard et al.\n[2010] showed that MPE in PU learning is a generalization of the problem of\nestimating the proportion of true null hypotheses in multiple testing problems.\nMotivated by this idea, we propose reducing the problem to one dimension via\nconstruction of a probabilistic classifier trained on the P and U data sets\nfollowed by application of a one--dimensional mixture proportion method from\nthe multiple testing literature to the observation class probabilities. The\nflexibility of this framework lies in the freedom to choose the classifier and\nthe one--dimensional MPE method. We prove consistency of two mixture proportion\nestimators using bounds from empirical process theory, develop tuning parameter\nfree implementations, and demonstrate that they have competitive performance on\nsimulated waveform data and a protein signaling problem.\n",
"title": "A Flexible Procedure for Mixture Proportion Estimation in Positive--Unlabeled Learning"
}
| null | null | null | null | true | null |
13791
| null |
Default
| null | null |
null |
{
"abstract": " JUNO is a multipurpose neutrino experiment which is designed to determine\nneutrino mass hierarchy and precisely measure oscillation parameters. As one of\nthe important systems, the JUNO offline software is being developed using the\nSNiPER software. In this proceeding, we focus on the requirements of JUNO\nsimulation and present the working solution based on the SNiPER.\nThe JUNO simulation framework is in charge of managing event data, detector\ngeometries and materials, physics processes, simulation truth information etc.\nIt glues physics generator, detector simulation and electronics simulation\nmodules together to achieve a full simulation chain. In the implementation of\nthe framework, many attractive characteristics of the SNiPER have been used,\nsuch as dynamic loading, flexible flow control, multiple event management and\nPython binding. Furthermore, additional efforts have been made to make both\ndetector and electronics simulation flexible enough to accommodate and optimize\ndifferent detector designs.\nFor the Geant4-based detector simulation, each sub-detector component is\nimplemented as a SNiPER tool which is a dynamically loadable and configurable\nplugin. So it is possible to select the detector configuration at runtime. The\nframework provides the event loop to drive the detector simulation and\ninteracts with the Geant4 which is implemented as a passive service. All levels\nof user actions are wrapped into different customizable tools, so that user\nfunctions can be easily extended by just adding new tools. The electronics\nsimulation has been implemented by following an event driven scheme. The SNiPER\ntask component is used to simulate data processing steps in the electronics\nmodules. The electronics and trigger are synchronized by triggered events\ncontaining possible physics signals.\n",
"title": "The Application of SNiPER to the JUNO Simulation"
}
| null | null |
[
"Physics"
] | null | true | null |
13792
| null |
Validated
| null | null |
null |
{
"abstract": " Sigma-Pi-Sigma neural networks (SPSNNs) as a kind of high-order neural\nnetworks can provide more powerful mapping capability than the traditional\nfeedforward neural networks (Sigma-Sigma neural networks). In the existing\nliterature, in order to reduce the number of the Pi nodes in the Pi layer, a\nspecial multinomial P_s is used in SPSNNs. Each monomial in P_s is linear with\nrespect to each particular variable sigma_i when the other variables are taken\nas constants. Therefore, the monomials like sigma_i^n or sigma_i^n sigma_j with\nn>1 are not included. This choice may be somehow intuitive, but is not\nnecessarily the best. We propose in this paper a modified Sigma-Pi-Sigma neural\nnetwork (MSPSNN) with an adaptive approach to find a better multinomial for a\ngiven problem. To elaborate, we start from a complete multinomial with a given\norder. Then we employ a regularization technique in the learning process for\nthe given problem to reduce the number of monomials used in the multinomial,\nand end up with a new SPSNN involving the same number of monomials (= the\nnumber of nodes in the Pi-layer) as in P_s. Numerical experiments on some\nbenchmark problems show that our MSPSNN behaves better than the traditional\nSPSNN with P_s.\n",
"title": "A Modified Sigma-Pi-Sigma Neural Network with Adaptive Choice of Multinomials"
}
| null | null |
[
"Statistics"
] | null | true | null |
13793
| null |
Validated
| null | null |
null |
{
"abstract": " Proceedings of the 2017 AdKDD and TargetAd Workshop held in conjunction with\nthe 23rd ACM SIGKDD Conference on Knowledge Discovery and Data Mining Halifax,\nNova Scotia, Canada.\n",
"title": "Proceedings of the 2017 AdKDD & TargetAd Workshop"
}
| null | null | null | null | true | null |
13794
| null |
Default
| null | null |
null |
{
"abstract": " Due to the possible lack of primal-dual-type error bounds, the superlinear\nconvergence for the Karush-Kuhn-Tucker (KKT) residues of the sequence generated\nby augmented Lagrangian method (ALM) for solving convex composite conic\nprogramming (CCCP) has long been an outstanding open question. In this paper,\nwe aim to resolve this issue by first conducting convergence rate analysis for\nthe ALM with Rockafellar's stopping criteria under only a mild quadratic growth\ncondition on the dual of CCCP. More importantly, by further assuming that the\nRobinson constraint qualification holds, we establish the R-superlinear\nconvergence of the KKT residues of the iterative sequence under\neasy-to-implement stopping criteria {for} the augmented Lagrangian subproblems.\nEquipped with this discovery, we gain insightful interpretations on the\nimpressive numerical performance of several recently developed semismooth\nNewton-CG based ALM solvers for solving linear and convex quadratic\nsemidefinite programming.\n",
"title": "On the R-superlinear convergence of the KKT residues generated by the augmented Lagrangian method for convex composite conic programming"
}
| null | null | null | null | true | null |
13795
| null |
Default
| null | null |
null |
{
"abstract": " Expertise in programming traditionally assumes a binary novice-expert divide.\nLearning resources typically target programmers who are learning programming\nfor the first time, or expert programmers for that language. An\nunderrepresented, yet important group of programmers are those that are\nexperienced in one programming language, but desire to author code in a\ndifferent language. For this scenario, we postulate that an effective form of\nfeedback is presented as a transfer from concepts in the first language to the\nsecond. Current programming environments do not support this form of feedback.\nIn this study, we apply the theory of learning transfer to teach a language\nthat programmers are less familiar with--such as R--in terms of a programming\nlanguage they already know--such as Python. We investigate learning transfer\nusing a new tool called Transfer Tutor that presents explanations for R code in\nterms of the equivalent Python code. Our study found that participants\nleveraged learning transfer as a cognitive strategy, even when unprompted.\nParticipants found Transfer Tutor to be useful across a number of affordances\nlike stepping through and highlighting facts that may have been missed or\nmisunderstood. However, participants were reluctant to accept facts without\ncode execution or sometimes had difficulty reading explanations that are\nverbose or complex. These results provide guidance for future designs and\nresearch directions that can support learning transfer when learning new\nprogramming languages.\n",
"title": "It's Like Python But: Towards Supporting Transfer of Programming Language Knowledge"
}
| null | null | null | null | true | null |
13796
| null |
Default
| null | null |
null |
{
"abstract": " Gaining a better understanding of how and what machine learning systems learn\nis important to increase confidence in their decisions and catalyze further\nresearch. In this paper, we analyze the predictions made by a specific type of\nrecurrent neural network, mixture density RNNs (MD-RNNs). These networks learn\nto model predictions as a combination of multiple Gaussian distributions,\nmaking them particularly interesting for problems where a sequence of inputs\nmay lead to several distinct future possibilities. An example is learning\ninternal models of an environment, where different events may or may not occur,\nbut where the average over different events is not meaningful. By analyzing the\npredictions made by trained MD-RNNs, we find that their different Gaussian\ncomponents have two complementary roles: 1) Separately modeling different\nstochastic events and 2) Separately modeling scenarios governed by different\nrules. These findings increase our understanding of what is learned by\npredictive MD-RNNs, and open up new research directions for further\nunderstanding how we can benefit from their self-organizing model\ndecomposition.\n",
"title": "How do Mixture Density RNNs Predict the Future?"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
13797
| null |
Validated
| null | null |
null |
{
"abstract": " The advanced operation of future electricity distribution systems is likely\nto require significant observability of the different parameters of interest\n(e.g., demand, voltages, currents, etc.). Ensuring completeness of data is,\ntherefore, paramount. In this context, an algorithm for recovering missing\nstate variable observations in electricity distribution systems is presented.\nThe proposed method exploits the low rank structure of the state variables via\na matrix completion approach while incorporating prior knowledge in the form of\nsecond order statistics. Specifically, the recovery method combines nuclear\nnorm minimization with Bayesian estimation. The performance of the new\nalgorithm is compared to the information-theoretic limits and tested trough\nsimulations using real data of an urban low voltage distribution system. The\nimpact of the prior knowledge is analyzed when a mismatched covariance is used\nand for a Markovian sampling that introduces structure in the observation\npattern. Numerical results demonstrate that the proposed algorithm is robust\nand outperforms existing state of the art algorithms.\n",
"title": "Robust Recovery of Missing Data in Electricity Distribution Systems"
}
| null | null | null | null | true | null |
13798
| null |
Default
| null | null |
null |
{
"abstract": " We propose a nonlinear Discrete Duality Finite Volume scheme to approximate\nthe solutions of drift diffusion equations. The scheme is built to preserve at\nthe discrete level even on severely distorted meshes the energy / energy\ndissipation relation. This relation is of paramount importance to capture the\nlong-time behavior of the problem in an accurate way. To enforce it, the linear\nconvection diffusion equation is rewritten in a nonlinear form before being\ndiscretized. We establish the existence of positive solutions to the scheme.\nBased on compactness arguments, the convergence of the approximate solution\ntowards a weak solution is established. Finally, we provide numerical evidences\nof the good behavior of the scheme when the discretization parameters tend to 0\nand when time goes to infinity.\n",
"title": "Numerical analysis of a nonlinear free-energy diminishing Discrete Duality Finite Volume scheme for convection diffusion equations"
}
| null | null | null | null | true | null |
13799
| null |
Default
| null | null |
null |
{
"abstract": " $N$-body simulations study the dynamics of $N$ particles under the influence\nof mutual long-distant forces such as gravity. In practice, $N$-body codes will\nviolate Newton's third law if they use either an approximate Poisson solver or\nindividual timesteps. In this study, we construct a novel $N$-body scheme by\ncombining a fast multipole method (FMM) based Poisson solver and a time\nintegrator using a hierarchical Hamiltonian splitting (HHS) technique. We test\nour implementation for collision-less systems using several problems in\ngalactic dynamics. As a result of the momentum conserving nature of these two\nkey components, the new $N$-body scheme is also momentum conserving. Moreover,\nwe can fully utilize the $\\mathcal O(\\textit N)$ complexity of FMM with the\nintegrator. With the restored force symmetry, we can improve both angular\nmomentum conservation and energy conservation substantially. The new scheme\nwill be suitable for many applications in galactic dynamics and structure\nformation. Our implementation, in the code Taichi, is publicly available at\nthis https URL.\n",
"title": "A momentum conserving $N$-body scheme with individual timesteps"
}
| null | null | null | null | true | null |
13800
| null |
Default
| null | null |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.