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": " This paper presents a generic Bayesian framework that enables any deep\nlearning model to actively learn from targeted crowds. Our framework inherits\nfrom recent advances in Bayesian deep learning, and extends existing work by\nconsidering the targeted crowdsourcing approach, where multiple annotators with\nunknown expertise contribute an uncontrolled amount (often limited) of\nannotations. Our framework leverages the low-rank structure in annotations to\nlearn individual annotator expertise, which then helps to infer the true labels\nfrom noisy and sparse annotations. It provides a unified Bayesian model to\nsimultaneously infer the true labels and train the deep learning model in order\nto reach an optimal learning efficacy. Finally, our framework exploits the\nuncertainty of the deep learning model during prediction as well as the\nannotators' estimated expertise to minimize the number of required annotations\nand annotators for optimally training the deep learning model.\nWe evaluate the effectiveness of our framework for intent classification in\nAlexa (Amazon's personal assistant), using both synthetic and real-world\ndatasets. Experiments show that our framework can accurately learn annotator\nexpertise, infer true labels, and effectively reduce the amount of annotations\nin model training as compared to state-of-the-art approaches. We further\ndiscuss the potential of our proposed framework in bridging machine learning\nand crowdsourcing towards improved human-in-the-loop systems.\n",
"title": "Leveraging Crowdsourcing Data For Deep Active Learning - An Application: Learning Intents in Alexa"
}
| null | null | null | null | true | null |
11301
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we propose a supervised learning system for counting and\nlocalizing palm trees in high-resolution, panchromatic satellite imagery\n(40cm/pixel to 1.5m/pixel). A convolutional neural network classifier trained\non a set of palm and no-palm images is applied across a satellite image scene\nin a sliding window fashion. The resultant confidence map is smoothed with a\nuniform filter. A non-maximal suppression is applied onto the smoothed\nconfidence map to obtain peaks. Trained with a small dataset of 500 images of\nsize 40x40 cropped from satellite images, the system manages to achieve a tree\ncount accuracy of over 99%.\n",
"title": "Using Convolutional Neural Networks to Count Palm Trees in Satellite Images"
}
| null | null | null | null | true | null |
11302
| null |
Default
| null | null |
null |
{
"abstract": " Woodin has shown that if there is a measurable Woodin cardinal then there is,\nin an appropriate sense, a sharp for the Chang model. We produce, in a weaker\nsense, a sharp for the Chang model using only the existence of a cardinal\n$\\kappa$ having an extender of length $\\kappa^{+\\omega_1}$.\n",
"title": "The sharp for the Chang model is small"
}
| null | null | null | null | true | null |
11303
| null |
Default
| null | null |
null |
{
"abstract": " Collective effects in deformed atomic nuclei present possible avenues of\nstudy on the non-spherical distribution of neutrons and the violation of the\nlocal Lorentz invariance. We introduce the weak quadrupole moment of nuclei,\nrelated to the quadrupole distribution of the weak charge in the nucleus. The\nweak quadrupole moment produces tensor weak interaction between the nucleus and\nelectrons and can be observed in atomic and molecular experiments measuring\nparity nonconservation. The dominating contribution to the weak quadrupole is\ngiven by the quadrupole moment of the neutron distribution, therefore,\ncorresponding experiments should allow one to measure the neutron quadrupoles.\nUsing the deformed oscillator model and the Schmidt model we calculate the\nquadrupole distributions of neutrons, $Q_{n}$, the weak quadrupole moments\n,$Q_{W}^{(2)}$, and the Lorentz Innvariance violating energy shifts in\n$^{9}$Be, $^{21}$Ne , $^{27}$Al, $^{131}$Xe, $^{133}$Cs, $^{151}$Eu,\n$^{153}$Eu, $^{163}$Dy, $^{167}$Er, $^{173}$Yb, $^{177}$Hf, $^{179}$Hf,\n$^{181}$Ta, $^{201}$Hg and $^{229}$Th.\n",
"title": "Weak quadrupole moments"
}
| null | null |
[
"Physics"
] | null | true | null |
11304
| null |
Validated
| null | null |
null |
{
"abstract": " We propose an efficient and scalable method for incrementally building a\ndense, semantically annotated 3D map in real-time. The proposed method assigns\nclass probabilities to each region, not each element (e.g., surfel and voxel),\nof the 3D map which is built up through a robust SLAM framework and\nincrementally segmented with a geometric-based segmentation method. Differently\nfrom all other approaches, our method has a capability of running at over 30Hz\nwhile performing all processing components, including SLAM, segmentation, 2D\nrecognition, and updating class probabilities of each segmentation label at\nevery incoming frame, thanks to the high efficiency that characterizes the\ncomputationally intensive stages of our framework. By utilizing a specifically\ndesigned CNN to improve the frame-wise segmentation result, we can also achieve\nhigh accuracy. We validate our method on the NYUv2 dataset by comparing with\nthe state of the art in terms of accuracy and computational efficiency, and by\nmeans of an analysis in terms of time and space complexity.\n",
"title": "Fast and Accurate Semantic Mapping through Geometric-based Incremental Segmentation"
}
| null | null | null | null | true | null |
11305
| null |
Default
| null | null |
null |
{
"abstract": " We study a pumping lemma for the word/tree languages generated by\nhigher-order grammars. Pumping lemmas are known up to order-2 word languages\n(i.e., for regular/context-free/indexed languages), and have been used to show\nthat a given language does not belong to the classes of\nregular/context-free/indexed languages. We prove a pumping lemma for word/tree\nlanguages of arbitrary orders, modulo a conjecture that a higher-order version\nof Kruskal's tree theorem holds. We also show that the conjecture indeed holds\nfor the order-2 case, which yields a pumping lemma for order-2 tree languages\nand order-3 word languages.\n",
"title": "Pumping Lemma for Higher-order Languages"
}
| null | null | null | null | true | null |
11306
| null |
Default
| null | null |
null |
{
"abstract": " In order to alleviate data sparsity and overfitting problems in maximum\nlikelihood estimation (MLE) for sequence prediction tasks, we propose the\nGenerative Bridging Network (GBN), in which a novel bridge module is introduced\nto assist the training of the sequence prediction model (the generator\nnetwork). Unlike MLE directly maximizing the conditional likelihood, the bridge\nextends the point-wise ground truth to a bridge distribution conditioned on it,\nand the generator is optimized to minimize their KL-divergence. Three different\nGBNs, namely uniform GBN, language-model GBN and coaching GBN, are proposed to\npenalize confidence, enhance language smoothness and relieve learning burden.\nExperiments conducted on two recognized sequence prediction tasks (machine\ntranslation and abstractive text summarization) show that our proposed GBNs can\nyield significant improvements over strong baselines. Furthermore, by analyzing\nsamples drawn from different bridges, expected influences on the generator are\nverified.\n",
"title": "Generative Bridging Network in Neural Sequence Prediction"
}
| null | null | null | null | true | null |
11307
| null |
Default
| null | null |
null |
{
"abstract": " Cognitive arithmetic studies the mental processes used in solving math\nproblems. This area of research explores the retrieval mechanisms and\nstrategies used by people during a common cognitive task. Past research has\nshown that human performance in arithmetic operations is correlated to the\nnumerical size of the problem. Past research on cognitive arithmetic has\npinpointed this trend to either retrieval strength, error checking, or\nstrategy-based approaches when solving equations. This paper describes a\nrule-based computational model that performs the four major arithmetic\noperations (addition, subtraction, multiplication and division) on two\noperands. We then evaluated our model to probe its validity in representing the\nprevailing concepts observed in psychology experiments from the related works.\nThe experiments specifically explore the problem size effect, an\nactivation-based model for fact retrieval, backup strategies when retrieval\nfails, and finally optimization strategies when faced with large operands. From\nour experimental results, we concluded that our model's response times were\ncomparable to results observed when people performed similar tasks during\npsychology experiments. The fit of our model in reproducing these results and\nincorporating accuracy into our model are discussed.\n",
"title": "A Rule-Based Computational Model of Cognitive Arithmetic"
}
| null | null | null | null | true | null |
11308
| null |
Default
| null | null |
null |
{
"abstract": " Arbitrarily many pairwise inequivalent modular categories can share the same\nmodular data. We exhibit a family of examples that are module categories over\ntwisted Drinfeld doubles of finite groups, and thus in particular integral\nmodular categories.\n",
"title": "Modular categories are not determined by their modular data"
}
| null | null | null | null | true | null |
11309
| null |
Default
| null | null |
null |
{
"abstract": " While anomaly detection in static networks has been extensively studied, only\nrecently, researchers have focused on dynamic networks. This trend is mainly\ndue to the capacity of dynamic networks in representing complex physical,\nbiological, cyber, and social systems. This paper proposes a new methodology\nfor modeling and monitoring of dynamic attributed networks for quick detection\nof temporal changes in network structures. In this methodology, the generalized\nlinear model (GLM) is used to model static attributed networks. This model is\nthen combined with a state transition equation to capture the dynamic behavior\nof the system. Extended Kalman filter (EKF) is used as an online, recursive\ninference procedure to predict and update network parameters over time. In\norder to detect changes in the underlying mechanism of edge formation,\nprediction residuals are monitored through an Exponentially Weighted Moving\nAverage (EWMA) control chart. The proposed modeling and monitoring procedure is\nexamined through simulations for attributed binary and weighted networks. The\nemail communication data from the Enron corporation is used as a case study to\nshow how the method can be applied in real-world problems.\n",
"title": "Change Detection in a Dynamic Stream of Attributed Networks"
}
| null | null | null | null | true | null |
11310
| null |
Default
| null | null |
null |
{
"abstract": " In this work we explore the utility of locally differentially private thermal\nsensor data. We design a locally differentially private recovery algorithm for\nthe 1-dimensional, discrete heat source location problem and analyse its\nperformance in terms of the Earth Mover Distance error. Our work indicates that\nit is possible to produce locally private sensor measurements that both keep\nthe exact locations of the heat sources private and permit recovery of the\n\"general geographic vicinity\" of the sources. We also discuss the relationship\nbetween the property of an inverse problem being ill-conditioned and the amount\nof noise needed to maintain privacy.\n",
"title": "Local Differential Privacy for Physical Sensor Data and Sparse Recovery"
}
| null | null | null | null | true | null |
11311
| null |
Default
| null | null |
null |
{
"abstract": " We propose a method for feature selection that employs kernel-based measures\nof independence to find a subset of covariates that is maximally predictive of\nthe response. Building on past work in kernel dimension reduction, we show how\nto perform feature selection via a constrained optimization problem involving\nthe trace of the conditional covariance operator. We prove various consistency\nresults for this procedure, and also demonstrate that our method compares\nfavorably with other state-of-the-art algorithms on a variety of synthetic and\nreal data sets.\n",
"title": "Kernel Feature Selection via Conditional Covariance Minimization"
}
| null | null | null | null | true | null |
11312
| null |
Default
| null | null |
null |
{
"abstract": " We demonstrate the existence of the excited state of an exciton-polariton in\na semiconductor microcavity. The strong coupling of the quantum well heavy-hole\nexciton in an excited 2s state to the cavity photon is observed in non-zero\nmagnetic field due to surprisingly fast increase of Rabi energy of the 2s\nexciton-polariton in magnetic field. This effect is explained by a strong\nmodification of the wave-function of the relative electron-hole motion for the\n2s exciton state.\n",
"title": "2s exciton-polariton revealed in an external magnetic field"
}
| null | null | null | null | true | null |
11313
| null |
Default
| null | null |
null |
{
"abstract": " In this paper, we discuss the generalized Hamming weights of a class of\nlinear codes associated with non-degenerate quadratic forms. In order to do so,\nwe study the quadratic forms over subspaces of finite field and obtain some\ninteresting results about subspaces and their dual spaces. On this basis, we\nsolve all the generalized Hamming weights of these linear codes.\n",
"title": "Weight hierarchy of a class of linear codes relating to non-degenerate quadratic forms"
}
| null | null | null | null | true | null |
11314
| null |
Default
| null | null |
null |
{
"abstract": " We examine systematically the (in)consistency between cosmological\nconstraints as obtained from various current data sets of the expansion\nhistory, Large Scale Structure (LSS), and Cosmic Microwave Background (CMB)\nfrom Planck. We run (dis)concordance tests within each set and across the sets\nusing a recently introduced index of inconsistency (IOI) capable of dissecting\ninconsistencies between two or more data sets. First, we compare the\nconstraints on $H_0$ from five different methods and find that the IOI drops\nfrom 2.85 to 0.88 (on Jeffreys' scales) when the local $H_0$ measurements is\nremoved. This seems to indicate that the local measurement is an outlier, thus\nfavoring a systematics-based explanation. We find a moderate inconsistency\n(IOI=2.61) between Planck temperature and polarization. We find that current\nLSS data sets including WiggleZ, SDSS RSD, CFHTLenS, CMB lensing and SZ cluster\ncount, are consistent one with another and when all combined. However, we find\na persistent moderate inconsistency between Planck and individual or combined\nLSS probes. For Planck TT+lowTEB versus individual LSS probes, the IOI spans\nthe range 2.92--3.72 and increases to 3.44--4.20 when the polarization data is\nadded in. The joint LSS versus the combined Planck temperature and polarization\nhas an IOI of 2.83 in the most conservative case. But if Planck lowTEB is added\nto the joint LSS to constrain $\\tau$ and break degeneracies, the inconsistency\nbetween Planck and joint LSS data increases to the high-end of the moderate\nrange with IOI=4.81. Whether due to systematic effects in the data or to the\nunderlying model, these inconsistencies need to be resolved. Finally, we\nperform forecast calculations using LSST and find that the discordance between\nPlanck and future LSS data, if it persists as present, can rise up to a high\nIOI of 17, thus falling in the strong range of inconsistency. (Abridged).\n",
"title": "Cosmological discordances II: Hubble constant, Planck and large-scale-structure data sets"
}
| null | null |
[
"Physics"
] | null | true | null |
11315
| null |
Validated
| null | null |
null |
{
"abstract": " We theoretically investigate the spin injection from a ferromagnetic silicene\nto a normal silicene (FS/NS), where the magnetization in the FS is assumed from\nthe magnetic proximity effect. Based on a silicene lattice model, we\ndemonstrated that the pure spin injection could be obtained by tuning the Fermi\nenergy of two spin species, where one is in the spin orbit coupling gap and the\nother one is outside the gap. Moreover, the valley polarity of the spin species\ncan be controlled by a perpendicular electric field in the FS region. Our\nfindings may shed light on making silicene-based spin and valley devices in the\nspintronics and valleytronics field.\n",
"title": "The perfect spin injection in silicene FS/NS junction"
}
| null | null | null | null | true | null |
11316
| null |
Default
| null | null |
null |
{
"abstract": " Contact-assisted protein folding has made very good progress, but two\nchallenges remain. One is accurate contact prediction for proteins lack of many\nsequence homologs and the other is that time-consuming folding simulation is\noften needed to predict good 3D models from predicted contacts. We show that\nprotein distance matrix can be predicted well by deep learning and then\ndirectly used to construct 3D models without folding simulation at all. Using\ndistance geometry to construct 3D models from our predicted distance matrices,\nwe successfully folded 21 of the 37 CASP12 hard targets with a median family\nsize of 58 effective sequence homologs within 4 hours on a Linux computer of 20\nCPUs. In contrast, contacts predicted by direct coupling analysis (DCA) cannot\nfold any of them in the absence of folding simulation and the best CASP12 group\nfolded 11 of them by integrating predicted contacts into complex,\nfragment-based folding simulation. The rigorous experimental validation on 15\nCASP13 targets show that among the 3 hardest targets of new fold our\ndistance-based folding servers successfully folded 2 large ones with <150\nsequence homologs while the other servers failed on all three, and that our ab\ninitio folding server also predicted the best, high-quality 3D model for a\nlarge homology modeling target. Further experimental validation in CAMEO shows\nthat our ab initio folding server predicted correct fold for a membrane protein\nof new fold with 200 residues and 229 sequence homologs while all the other\nservers failed. These results imply that deep learning offers an efficient and\naccurate solution for ab initio folding on a personal computer.\n",
"title": "Distance-based Protein Folding Powered by Deep Learning"
}
| null | null | null | null | true | null |
11317
| null |
Default
| null | null |
null |
{
"abstract": " A semiorder is a model of preference relations where each element $x$ is\nassociated with a utility value $\\alpha(x)$, and there is a threshold $t$ such\nthat $y$ is preferred to $x$ iff $\\alpha(y) > \\alpha(x)+t$. These are motivated\nby the notion that there is some uncertainty in the utility values we assign an\nobject or that a subject may be unable to distinguish a preference between\nobjects whose values are close. However, they fail to model the well-known\nphenomenon that preferences are not always transitive. Also, if we are\nuncertain of the utility values, it is not logical that preference is\ndetermined absolutely by a comparison of them with an exact threshold. We\npropose a new model in which there are two thresholds, $t_1$ and $t_2$; if the\ndifference $\\alpha(y) - \\alpha(x)$ less than $t_1$, then $y$ is not preferred\nto $x$; if the difference is greater than $t_2$ then $y$ is preferred to $x$;\nif it is between $t_1$ and $t_2$, then then $y$ may or may not be preferred to\n$x$. We call such a relation a double-threshold semiorder, and the\ncorresponding directed graph $G = (V,E)$ a double threshold digraph. Every\ndirected acyclic graph is a double threshold graph; bounds on $t_2/t_1$ give a\nnested hierarchy of subclasses of the directed acyclic graphs. In this paper we\ncharacterize the subclasses in terms of forbidden subgraphs, and give\nalgorithms for finding an assignment of of utility values that explains the\nrelation in terms of a given $(t_1,t_2)$ or else produces a forbidden subgraph,\nand finding the minimum value $\\lambda$ of $t_2/t_1$ that is satisfiable for a\ngiven directed acyclic graph. We show that $\\lambda$ gives a measure of the\ncomplexity of a directed acyclic graph with respect to several optimization\nproblems that are NP-hard on arbitrary directed acyclic graphs.\n",
"title": "Double Threshold Digraphs"
}
| null | null | null | null | true | null |
11318
| null |
Default
| null | null |
null |
{
"abstract": " Let $\\{ R_n, {\\mathfrak m}_n \\}_{n \\ge 0}$ be an infinite sequence of regular\nlocal rings with $R_{n+1}$ birationally dominating $R_n$ and ${\\mathfrak\nm}_nR_{n+1}$ a principal ideal of $R_{n+1}$ for each $n$. We examine properties\nof the integrally closed local domain $S = \\bigcup_{n \\ge 0}R_n$.\n",
"title": "Directed unions of local quadratic transforms of regular local rings and pullbacks"
}
| null | null | null | null | true | null |
11319
| null |
Default
| null | null |
null |
{
"abstract": " Deep neural networks are notorious for being sensitive to small well-chosen\nperturbations, and estimating the regularity of such architectures is of utmost\nimportance for safe and robust practical applications. In this paper, we\ninvestigate one of the key characteristics to assess the regularity of such\nmethods: the Lipschitz constant of deep learning architectures. First, we show\nthat, even for two layer neural networks, the exact computation of this\nquantity is NP-hard and state-of-art methods may significantly overestimate it.\nThen, we both extend and improve previous estimation methods by providing\nAutoLip, the first generic algorithm for upper bounding the Lipschitz constant\nof any automatically differentiable function. We provide a power method\nalgorithm working with automatic differentiation, allowing efficient\ncomputations even on large convolutions. Second, for sequential neural\nnetworks, we propose an improved algorithm named SeqLip that takes advantage of\nthe linear computation graph to split the computation per pair of consecutive\nlayers. Third we propose heuristics on SeqLip in order to tackle very large\nnetworks. Our experiments show that SeqLip can significantly improve on the\nexisting upper bounds.\n",
"title": "Lipschitz regularity of deep neural networks: analysis and efficient estimation"
}
| null | null |
[
"Statistics"
] | null | true | null |
11320
| null |
Validated
| null | null |
null |
{
"abstract": " We introduce a new model of teaching named \"preference-based teaching\" and a\ncorresponding complexity parameter---the preference-based teaching dimension\n(PBTD)---representing the worst-case number of examples needed to teach any\nconcept in a given concept class. Although the PBTD coincides with the\nwell-known recursive teaching dimension (RTD) on finite classes, it is\nradically different on infinite ones: the RTD becomes infinite already for\ntrivial infinite classes (such as half-intervals) whereas the PBTD evaluates to\nreasonably small values for a wide collection of infinite classes including\nclasses consisting of so-called closed sets w.r.t. a given closure operator,\nincluding various classes related to linear sets over $\\mathbb{N}_0$ (whose RTD\nhad been studied quite recently) and including the class of Euclidean\nhalf-spaces. On top of presenting these concrete results, we provide the reader\nwith a theoretical framework (of a combinatorial flavor) which helps to derive\nbounds on the PBTD.\n",
"title": "Preference-based Teaching"
}
| null | null | null | null | true | null |
11321
| null |
Default
| null | null |
null |
{
"abstract": " We study a binary spin-mixture of a zero-temperature repulsively interacting\n$^6$Li atoms using both the atomic-orbital and the density functional\napproaches. The gas is initially prepared in a configuration of two magnetic\ndomains and we determine the frequency of the spin-dipole oscillations which\nare emerging after the repulsive barrier, initially separating the domains, is\nremoved. We find, in agreement with recent experiment (G. Valtolina et al.,\narXiv:1605.07850 (2016)), the occurrence of a ferromagnetic instability in an\natomic gas while the interaction strength between different spin states is\nincreased, after which the system becomes ferromagnetic. The ferromagnetic\ninstability is preceded by the softening of the spin-dipole mode.\n",
"title": "Unified description of dynamics of a repulsive two-component Fermi gas"
}
| null | null | null | null | true | null |
11322
| null |
Default
| null | null |
null |
{
"abstract": " In an ideal test of the equivalence principle, the test masses fall in a\ncommon inertial frame. A real experiment is affected by gravity gradients,\nwhich introduce systematic errors by coupling to initial kinematic differences\nbetween the test masses. We demonstrate a method that reduces the sensitivity\nof a dual-species atom interferometer to initial kinematics by using a\nfrequency shift of the mirror pulse to create an effective inertial frame for\nboth atomic species. This suppresses the gravity-gradient-induced dependence of\nthe differential phase on initial kinematic differences by a factor of 100 and\nenables a precise measurement of these differences. We realize a relative\nprecision of $\\Delta g / g \\approx 6 \\times 10^{-11}$ per shot, which improves\non the best previous result for a dual-species atom interferometer by more than\nthree orders of magnitude. By suppressing gravity gradient systematic errors to\nbelow one part in $10^{13}$, these results pave the way for an atomic test of\nthe equivalence principle at an accuracy comparable with state-of-the-art\nclassical tests.\n",
"title": "Effective inertial frame in an atom interferometric test of the equivalence principle"
}
| null | null | null | null | true | null |
11323
| null |
Default
| null | null |
null |
{
"abstract": " To understand emergent magnetic monopole dynamics in the spin ices\nHo$_2$Ti$_2$O$_7$ and Dy$_2$Ti$_2$O$_7$, it is necessary to investigate the\nmechanisms by which spins flip in these materials. Presently there are thought\nto be two processes: quantum tunneling at low and intermediate temperatures and\nthermally activated at high temperatures. We identify possible couplings\nbetween crystal field and optical phonon excitations and construct a strictly\nconstrained model of phonon-mediated spin flipping that quantitatively\ndescribes the high-temperature processes in both compounds, as measured by\nquasielastic neutron scattering. We support the model with direct experimental\nevidence of the coupling between crystal field states and optical phonons in\nHo$_2$Ti$_2$O$_7$.\n",
"title": "Phonon-mediated spin-flipping mechanism in the spin ices Dy$_2$Ti$_2$O$_7$ and Ho$_2$Ti$_2$O$_7$"
}
| null | null | null | null | true | null |
11324
| null |
Default
| null | null |
null |
{
"abstract": " The hexatic phase predicted by the theories of two-dimensional melting is\ncharacterised by the power law decay of the orientational correlations whereas\nthe in-layer bond orientational order in all the hexatic smectic phases\nobserved so far was found to be long-range. We report a hexatic smectic phase\nwhere the in-layer bond orientational correlations decay as $\\propto r^{-1/4}$,\nin quantitative agreement with the hexatic ordering predicted by the theory for\ntwo dimensions. The phase was formed in a molecular dynamics simulation of a\none-component system of particles interacting via a spherically symmetric\npotential. This is the first observation of the theoretically predicted\ntwo-dimensional hexatic order in a three-dimensional system.\n",
"title": "A hexatic smectic phase with algebraically decaying bond-orientational order"
}
| null | null | null | null | true | null |
11325
| null |
Default
| null | null |
null |
{
"abstract": " Despite the fact that the observed gradient in water content among the\nGalilean satellites is globally consistent with a formation in a circum-Jovian\ndisk on both sides of the snowline, the mechanisms that led to a low water mass\nfraction in Europa ($\\sim$$8\\%$) are not yet understood. Here, we present new\nmodeling results of solids transport in the circum-Jovian disk accounting for\naerodynamic drag, turbulent diffusion, surface temperature evolution and\nsublimation of water ice. We find that the water mass fraction of pebbles\n(e.g., solids with sizes of 10$^{-2}$ -- 1 m) as they drift inward is globally\nconsistent with the current water content of the Galilean system. This opens\nthe possibility that each satellite could have formed through pebble accretion\nwithin a delimited region whose boundaries were defined by the position of the\nsnowline. This further implies that the migration of the forming satellites was\ntied to the evolution of the snowline so that Europa fully accreted from\npartially dehydrated material in the region just inside of the snowline.\n",
"title": "Pebble accretion at the origin of water in Europa"
}
| null | null | null | null | true | null |
11326
| null |
Default
| null | null |
null |
{
"abstract": " Traffic forecasting is a particularly challenging application of\nspatiotemporal forecasting, due to the complicated spatial dependencies on\nroadway networks and the time-varying traffic patterns. To address this\nchallenge, we learn the traffic network as a graph and propose a novel deep\nlearning framework, Traffic Graph Convolutional Long Short-Term Memory Neural\nNetwork (TGC-LSTM), to learn the interactions between roadways in the traffic\nnetwork and forecast the network-wide traffic state. We define the traffic\ngraph convolution based on the physical network topology. The relationship\nbetween traffic graph convolution and the spectral graph convolution is also\ndiscussed. The proposed model employs L1-norms on the graph convolution weights\nand L2-norms on the extracted features to identify the most influential\nroadways in the traffic network. Experiments show that our TGC-LSTM network is\nable to capture the complex spatial-temporal dependencies efficiently present\nin a vehicle traffic network and consistently outperforms state-of-the-art\nbaseline methods on two heterogeneous real-world traffic datasets. The\nvisualization of graph convolution weights shows that the proposed framework\ncan accurately recognize the most influential roadway segments in real-world\ntraffic networks.\n",
"title": "Traffic Graph Convolutional Recurrent Neural Network: A Deep Learning Framework for Network-Scale Traffic Learning and Forecasting"
}
| null | null |
[
"Statistics"
] | null | true | null |
11327
| null |
Validated
| null | null |
null |
{
"abstract": " We consider two types of averaging of complex covariance matrices, a sample\nmean (average) and the sample Fréchet mean. We analyse the performance of\nthese quantities as estimators for the true covariance matrix via `intrinsic'\nversions of bias and mean square error, a methodology which takes account of\ngeometric structure. We derive simple expressions for the intrinsic bias in\nboth cases, and the simple average is seen to be preferable. The same is true\nfor the asymptotic Riemannian risk, and for the Riemannian risk itself in the\nscalar case. Combined with a similar preference for the simple average using\nnon-intrinsic analysis, we conclude that the simple average is preferred\noverall to the sample Fréchet mean in this context.\n",
"title": "Intrinsic Analysis of the Sample Fréchet Mean and Sample Mean of Complex Wishart Matrices"
}
| null | null | null | null | true | null |
11328
| null |
Default
| null | null |
null |
{
"abstract": " This paper characterizes the capacity region of Gaussian MIMO broadcast\nchannels (BCs) with per-antenna power constraint (PAPC). While the capacity\nregion of MIMO BCs with a sum power constraint (SPC) was extensively studied,\nthat under PAPC has received less attention. A reason is that efficient\nsolutions for this problem are hard to find. The goal of this paper is to\ndevise an efficient algorithm for determining the capacity region of Gaussian\nMIMO BCs subject to PAPC, which is scalable to the problem size. To this end,\nwe first transform the weighted sum capacity maximization problem, which is\ninherently nonconvex with the input covariance matrices, into a convex\nformulation in the dual multiple access channel by minimax duality. Then we\nderive a computationally efficient algorithm combining the concept of\nalternating optimization and successive convex approximation. The proposed\nalgorithm achieves much lower complexity compared to an existing interiorpoint\nmethod. Moreover, numerical results demonstrate that the proposed algorithm\nconverges very fast under various scenarios.\n",
"title": "Alternating Optimization for Capacity Region of Gaussian MIMO Broadcast Channels with Per-antenna Power Constraint"
}
| null | null | null | null | true | null |
11329
| null |
Default
| null | null |
null |
{
"abstract": " Lifestyles are a valuable model for understanding individuals' physical and\nmental lives, comparing social groups, and making recommendations for improving\npeople's lives. In this paper, we examine and compare lifestyle behaviors of\npeople living in cities of different sizes, utilizing freely available social\nmedia data as a large-scale, low-cost alternative to traditional survey\nmethods. We use the Greater New York City area as a representative for large\ncities, and the Greater Rochester area as a representative for smaller cities\nin the United States. We employed matrix factor analysis as an unsupervised\nmethod to extract salient mobility and work-rest patterns for a large\npopulation of users within each metropolitan area. We discovered interesting\nhuman behavior patterns at both a larger scale and a finer granularity than is\npresent in previous literature, some of which allow us to quantitatively\ncompare the behaviors of individuals of living in big cities to those living in\nsmall cities. We believe that our social media-based approach to lifestyle\nanalysis represents a powerful tool for social computing in the big data age.\n",
"title": "Tales of Two Cities: Using Social Media to Understand Idiosyncratic Lifestyles in Distinctive Metropolitan Areas"
}
| null | null |
[
"Computer Science"
] | null | true | null |
11330
| null |
Validated
| null | null |
null |
{
"abstract": " With the availability of more powerful computers, iterative reconstruction\nalgorithms are the subject of an ongoing work in the design of more efficient\nreconstruction algorithms for X-ray computed tomography. In this work, we show\nhow two analytical reconstruction algorithms can be improved by correcting the\ncorresponding reconstructions using a randomized iterative reconstruction\nalgorithm. The combined analytical reconstruction followed by randomized\niterative reconstruction can also be viewed as a reconstruction algorithm\nwhich, in the experiments we have conducted, uses up to $35\\%$ less projection\nangles as compared to the analytical reconstruction algorithms and produces the\nsame results in terms of quality of reconstruction, without increasing the\nexecution time significantly.\n",
"title": "Randomized Iterative Reconstruction for Sparse View X-ray Computed Tomography"
}
| null | null | null | null | true | null |
11331
| null |
Default
| null | null |
null |
{
"abstract": " We propose two algorithms that can find local minima faster than the\nstate-of-the-art algorithms in both finite-sum and general stochastic nonconvex\noptimization. At the core of the proposed algorithms is\n$\\text{One-epoch-SNVRG}^+$ using stochastic nested variance reduction (Zhou et\nal., 2018a), which outperforms the state-of-the-art variance reduction\nalgorithms such as SCSG (Lei et al., 2017). In particular, for finite-sum\noptimization problems, the proposed\n$\\text{SNVRG}^{+}+\\text{Neon2}^{\\text{finite}}$ algorithm achieves\n$\\tilde{O}(n^{1/2}\\epsilon^{-2}+n\\epsilon_H^{-3}+n^{3/4}\\epsilon_H^{-7/2})$\ngradient complexity to converge to an $(\\epsilon, \\epsilon_H)$-second-order\nstationary point, which outperforms $\\text{SVRG}+\\text{Neon2}^{\\text{finite}}$\n(Allen-Zhu and Li, 2017) , the best existing algorithm, in a wide regime. For\ngeneral stochastic optimization problems, the proposed\n$\\text{SNVRG}^{+}+\\text{Neon2}^{\\text{online}}$ achieves\n$\\tilde{O}(\\epsilon^{-3}+\\epsilon_H^{-5}+\\epsilon^{-2}\\epsilon_H^{-3})$\ngradient complexity, which is better than both\n$\\text{SVRG}+\\text{Neon2}^{\\text{online}}$ (Allen-Zhu and Li, 2017) and\nNatasha2 (Allen-Zhu, 2017) in certain regimes. Furthermore, we explore the\nacceleration brought by third-order smoothness of the objective function.\n",
"title": "Finding Local Minima via Stochastic Nested Variance Reduction"
}
| null | null | null | null | true | null |
11332
| null |
Default
| null | null |
null |
{
"abstract": " The mean growth rate of the state vector is evaluated for a generalized\nlinear stochastic second-order system with a symmetric matrix. Diagonal entries\nof the matrix are assumed to be independent and exponentially distributed with\ndifferent means, while the off-diagonal entries are equal to zero.\n",
"title": "Growth rate of the state vector in a generalized linear stochastic system with symmetric matrix"
}
| null | null | null | null | true | null |
11333
| null |
Default
| null | null |
null |
{
"abstract": " Doctors often rely on their past experience in order to diagnose patients.\nFor a doctor with enough experience, almost every patient would have\nsimilarities to key cases seen in the past, and each new patient could be\nviewed as a mixture of these key past cases. Because doctors often tend to\nreason this way, an efficient computationally aided diagnostic tool that thinks\nin the same way might be helpful in locating key past cases of interest that\ncould assist with diagnosis. This article develops a novel mathematical model\nto mimic the type of logical thinking that physicians use when considering past\ncases. The proposed model can also provide physicians with explanations that\nwould be similar to the way they would naturally reason about cases. The\nproposed method is designed to yield predictive accuracy, computational\nefficiency, and insight into medical data; the key element is the insight into\nmedical data, in some sense we are automating a complicated process that\nphysicians might perform manually. We finally implemented the result of this\nwork on two publicly available healthcare datasets, for heart disease\nprediction and breast cancer prediction.\n",
"title": "Bayesian Patchworks: An Approach to Case-Based Reasoning"
}
| null | null | null | null | true | null |
11334
| null |
Default
| null | null |
null |
{
"abstract": " Machine Learning (ML) and Deep Learning (DL) models have achieved\nstate-of-the-art performance on multiple learning tasks, from vision to natural\nlanguage modelling. With the growing adoption of ML and DL to many areas of\ncomputer science, recent research has also started focusing on the security\nproperties of these models. There has been a lot of work undertaken to\nunderstand if (deep) neural network architectures are resilient to black-box\nadversarial attacks which craft perturbed input samples that fool the\nclassifier without knowing the architecture used. Recent work has also focused\non the transferability of adversarial attacks and found that adversarial\nattacks are generally easily transferable between models, datasets, and\ntechniques. However, such attacks and their analysis have not been covered from\nthe perspective of unsupervised machine learning algorithms. In this paper, we\nseek to bridge this gap through multiple contributions. We first provide a\nstrong (iterative) black-box adversarial attack that can craft adversarial\nsamples which will be incorrectly clustered irrespective of the choice of\nclustering algorithm. We choose 4 prominent clustering algorithms, and a\nreal-world dataset to show the working of the proposed adversarial algorithm.\nUsing these clustering algorithms we also carry out a simple study of\ncross-technique adversarial attack transferability.\n",
"title": "Strong Black-box Adversarial Attacks on Unsupervised Machine Learning Models"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
11335
| null |
Validated
| null | null |
null |
{
"abstract": " We define the formal affine Demazure algebra and formal affine Hecke algebra\nassociated to a Kac-Moody root system. We prove the structure theorems of these\nalgebras, hence, extending several result and construction (presentation in\nterms of generators and relations, coproduct and product structures, filtration\nby codimension of Bott-Samelson classes, root polynomials and multiplication\nformulas) that were previously known for finite root system.\n",
"title": "Formal affine Demazure and Hecke algebras of Kac-Moody root systems"
}
| null | null | null | null | true | null |
11336
| null |
Default
| null | null |
null |
{
"abstract": " Homographs, words with different meanings but the same surface form, have\nlong caused difficulty for machine translation systems, as it is difficult to\nselect the correct translation based on the context. However, with the advent\nof neural machine translation (NMT) systems, which can theoretically take into\naccount global sentential context, one may hypothesize that this problem has\nbeen alleviated. In this paper, we first provide empirical evidence that\nexisting NMT systems in fact still have significant problems in properly\ntranslating ambiguous words. We then proceed to describe methods, inspired by\nthe word sense disambiguation literature, that model the context of the input\nword with context-aware word embeddings that help to differentiate the word\nsense be- fore feeding it into the encoder. Experiments on three language pairs\ndemonstrate that such models improve the performance of NMT systems both in\nterms of BLEU score and in the accuracy of translating homographs.\n",
"title": "Handling Homographs in Neural Machine Translation"
}
| null | null |
[
"Computer Science"
] | null | true | null |
11337
| null |
Validated
| null | null |
null |
{
"abstract": " We show that a Hitchin representation is determined by the spectral radii of\nthe images of simple, non-separating closed curves. As a consequence, we\nclassify isometries of the intersection function on Hitchin components of\ndimension 3 and on the self-dual Hitchin components in all dimensions. As an\nimportant tool in the proof, we establish a transversality result for positive\nquadruples of flags.\n",
"title": "Simple Length Rigidity for Hitchin Representations"
}
| null | null | null | null | true | null |
11338
| null |
Default
| null | null |
null |
{
"abstract": " Digital pathology is not only one of the most promising fields of diagnostic\nmedicine, but at the same time a hot topic for fundamental research. Digital\npathology is not just the transfer of histopathological slides into digital\nrepresentations. The combination of different data sources (images, patient\nrecords, and *omics data) together with current advances in artificial\nintelligence/machine learning enable to make novel information accessible and\nquantifiable to a human expert, which is not yet available and not exploited in\ncurrent medical settings. The grand goal is to reach a level of usable\nintelligence to understand the data in the context of an application task,\nthereby making machine decisions transparent, interpretable and explainable.\nThe foundation of such an \"augmented pathologist\" needs an integrated approach:\nWhile machine learning algorithms require many thousands of training examples,\na human expert is often confronted with only a few data points. Interestingly,\nhumans can learn from such few examples and are able to instantly interpret\ncomplex patterns. Consequently, the grand goal is to combine the possibilities\nof artificial intelligence with human intelligence and to find a well-suited\nbalance between them to enable what neither of them could do on their own. This\ncan raise the quality of education, diagnosis, prognosis and prediction of\ncancer and other diseases. In this paper we describe some (incomplete) research\nissues which we believe should be addressed in an integrated and concerted\neffort for paving the way towards the augmented pathologist.\n",
"title": "Towards the Augmented Pathologist: Challenges of Explainable-AI in Digital Pathology"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
11339
| null |
Validated
| null | null |
null |
{
"abstract": " We present an algorithm to generate synthetic datasets of tunable difficulty\non classification of Morse code symbols for supervised machine learning\nproblems, in particular, neural networks. The datasets are spatially\none-dimensional and have a small number of input features, leading to high\ndensity of input information content. This makes them particularly challenging\nwhen implementing network complexity reduction methods. We explore how network\nperformance is affected by deliberately adding various forms of noise and\nexpanding the feature set and dataset size. Finally, we establish several\nmetrics to indicate the difficulty of a dataset, and evaluate their merits. The\nalgorithm and datasets are open-source.\n",
"title": "Morse Code Datasets for Machine Learning"
}
| null | null | null | null | true | null |
11340
| null |
Default
| null | null |
null |
{
"abstract": " Given the widespread popularity of spectral clustering (SC) for partitioning\ngraph data, we study a version of constrained SC in which we try to incorporate\nthe fairness notion proposed by Chierichetti et al. (2017). According to this\nnotion, a clustering is fair if every demographic group is approximately\nproportionally represented in each cluster. To this end, we develop variants of\nboth normalized and unnormalized constrained SC and show that they help find\nfairer clusterings on both synthetic and real data. We also provide a rigorous\ntheoretical analysis of our algorithms. While there have been efforts to\nincorporate various constraints into the SC framework, theoretically analyzing\nthem is a challenging problem. We overcome this by proposing a natural variant\nof the stochastic block model where h groups have strong inter-group\nconnectivity, but also exhibit a \"natural\" clustering structure which is fair.\nWe prove that our algorithms can recover this fair clustering with high\nprobability.\n",
"title": "Guarantees for Spectral Clustering with Fairness Constraints"
}
| null | null | null | null | true | null |
11341
| null |
Default
| null | null |
null |
{
"abstract": " The stochastic block model is widely used for detecting community structures\nin network data. How to test the goodness-of-fit of the model is one of the\nfundamental problems and has gained growing interests in recent years. In this\npaper, we propose a novel goodness-of-fit test based on the maximum entry of\nthe centered and re-scaled adjacency matrix for the stochastic block model. One\nnoticeable advantage of the proposed test is that the number of communities can\nbe allowed to grow linearly with the number of nodes ignoring a logarithmic\nfactor. We prove that the null distribution of the test statistic converges in\ndistribution to a Gumbel distribution, and we show that both the number of\ncommunities and the membership vector can be tested via the proposed method.\nFurther, we show that the proposed test has asymptotic power guarantee against\na class of alternatives. We also demonstrate that the proposed method can be\nextended to the degree-corrected stochastic block model. Both simulation\nstudies and real-world data examples indicate that the proposed method works\nwell.\n",
"title": "Using Maximum Entry-Wise Deviation to Test the Goodness-of-Fit for Stochastic Block Models"
}
| null | null |
[
"Statistics"
] | null | true | null |
11342
| null |
Validated
| null | null |
null |
{
"abstract": " Ego networks have proved to be a valuable tool for understanding the\nrelationships that individuals establish with their peers, both in offline and\nonline social networks. Particularly interesting are the cognitive constraints\nassociated with the interactions between the ego and the members of their ego\nnetwork, whereby individuals cannot maintain meaningful interactions with more\nthan 150 people, on average. In this work, we focus on the ego networks of\njournalists on Twitter, and we investigate whether they feature the same\ncharacteristics observed for other relevant classes of Twitter users, like\npoliticians and generic users. Our findings are that journalists are generally\nmore active and interact with more people than generic users. Their ego network\nstructure is very aligned with reference models derived from the social brain\nhypothesis and observed in general human ego networks. Remarkably, the\nsimilarity is even higher than the one of politicians and generic users ego\nnetworks. This may imply a greater cognitive involvement with Twitter than with\nother social interaction means. Moreover, the ego networks of journalists are\nmuch stabler than those of politicians and generic users, and the ego-alter\nties are often information-driven.\n",
"title": "Twitter and the Press: an Ego-Centred Analysis"
}
| null | null |
[
"Computer Science"
] | null | true | null |
11343
| null |
Validated
| null | null |
null |
{
"abstract": " In the space of less than one decade, the search for Majorana quasiparticles\nin condensed matter has become one of the hottest topics in physics. The aim of\nthis review is to provide a brief perspective of where we are with strong focus\non artificial implementations of one-dimensional topological superconductivity.\nAfter a self-contained introduction and some technical parts, an overview of\nthe current experimental status is given and some of the most successful\nexperiments of the last few years are discussed in detail. These include the\nnovel generation of ballistic InSb nanowire devices, epitaxial Al-InAs\nnanowires and Majorana boxes, high frequency experiments with proximitized\nquantum spin Hall insulators realised in HgTe quantum wells and recent\nexperiments on ferromagnetic atomic chains on top of superconducting surfaces.\n",
"title": "Majorana quasiparticles in condensed matter"
}
| null | null | null | null | true | null |
11344
| null |
Default
| null | null |
null |
{
"abstract": " We performed a comparative study of extraction of large-scale flow structures\nin Rayleigh Bénard convection using proper orthogonal decomposition (POD) and\n{\\em Fourier analysis}. We show that the free-slip basis functions capture the\nflow profiles successfully for the no-slip boundary conditions. We observe that\nthe large-scale POD modes capture a larger fraction of total energy than the\nFourier modes. However, the Fourier modes capture the rarer flow structures\nlike flow reversals better. The flow profiles of the dominant POD and Fourier\nmodes are quite similar. Our results show that the Fourier analysis provides an\nattractive alternative to POD analysis for capturing large-scale flow\nstructures.\n",
"title": "Proper orthogonal decomposition vs. Fourier analysis for extraction of large-scale structures of thermal convection"
}
| null | null | null | null | true | null |
11345
| null |
Default
| null | null |
null |
{
"abstract": " We propose a conjectural explicit formula of generating series of a new type\nfor Gromov--Witten invariants of $\\mathbb{P}^1$ of all degrees in full genera.\n",
"title": "On Gromov--Witten invariants of $\\mathbb{P}^1$"
}
| null | null | null | null | true | null |
11346
| null |
Default
| null | null |
null |
{
"abstract": " We describe a method for formation-change trajectory planning for large\nquadrotor teams in obstacle-rich environments. Our method decomposes the\nplanning problem into two stages: a discrete planner operating on a graph\nrepresentation of the workspace, and a continuous refinement that converts the\nnon-smooth graph plan into a set of C^k-continuous trajectories, locally\noptimizing an integral-squared-derivative cost. We account for the downwash\neffect, allowing safe flight in dense formations. We demonstrate the\ncomputational efficiency in simulation with up to 200 robots and the physical\nplausibility with an experiment with 32 nano-quadrotors. Our approach can\ncompute safe and smooth trajectories for hundreds of quadrotors in dense\nenvironments with obstacles in a few minutes.\n",
"title": "Downwash-Aware Trajectory Planning for Large Quadrotor Teams"
}
| null | null | null | null | true | null |
11347
| null |
Default
| null | null |
null |
{
"abstract": " Bubbly flows, as present in bubble column reactors, can be simulated using a\nvariety of simulation techniques. In order to gain high resolution CFD methods\nare used to simulate a pseudo 2D bubble column using EL and EE techniques. The\nforces on bubble dynamics are solved within open access software OpenFOAM with\nbubble interactions computed via Monte Carlo methods. The estimated bubble size\ndistribution and the predicted hold-up are compared to experimental data and\nother simulative work using EE approach and show reasonable consensus for both.\nBenchmarks with state of the art EE simulations shows that the EL approach is\nadvantageous if the bubble number stays at a certain level, as the EL approach\nscales linearly with the number of bubbles simulated. Therefore, different\ncomputational meshes have been used to also account for influence of the\nresolution quality. The EL approach indicated faster solution for all realistic\ncases, only deliberate decrease of coalescence rates could push CPU time to the\nlimits. Critical bubble number - when EE becomes advantageous over the EL\napproach - was estimated to be 40.000 in this particular case.\n",
"title": "Flow simulation in a 2D bubble column with the Euler-Lagrange and Euler-Euler method"
}
| null | null | null | null | true | null |
11348
| null |
Default
| null | null |
null |
{
"abstract": " In this paper, we study the problem of sampling from a given probability\ndensity function that is known to be smooth and strongly log-concave. We\nanalyze several methods of approximate sampling based on discretizations of the\n(highly overdamped) Langevin diffusion and establish guarantees on its error\nmeasured in the Wasserstein-2 distance. Our guarantees improve or extend the\nstate-of-the-art results in three directions. First, we provide an upper bound\non the error of the first-order Langevin Monte Carlo (LMC) algorithm with\noptimized varying step-size. This result has the advantage of being horizon\nfree (we do not need to know in advance the target precision) and to improve by\na logarithmic factor the corresponding result for the constant step-size.\nSecond, we study the case where accurate evaluations of the gradient of the\nlog-density are unavailable, but one can have access to approximations of the\naforementioned gradient. In such a situation, we consider both deterministic\nand stochastic approximations of the gradient and provide an upper bound on the\nsampling error of the first-order LMC that quantifies the impact of the\ngradient evaluation inaccuracies. Third, we establish upper bounds for two\nversions of the second-order LMC, which leverage the Hessian of the\nlog-density. We nonasymptotic guarantees on the sampling error of these\nsecond-order LMCs. These guarantees reveal that the second-order LMC algorithms\nimprove on the first-order LMC in ill-conditioned settings.\n",
"title": "User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient"
}
| null | null | null | null | true | null |
11349
| null |
Default
| null | null |
null |
{
"abstract": " The Madry Lab recently hosted a competition designed to test the robustness\nof their adversarially trained MNIST model. Attacks were constrained to perturb\neach pixel of the input image by a scaled maximal $L_\\infty$ distortion\n$\\epsilon$ = 0.3. This discourages the use of attacks which are not optimized\non the $L_\\infty$ distortion metric. Our experimental results demonstrate that\nby relaxing the $L_\\infty$ constraint of the competition, the elastic-net\nattack to deep neural networks (EAD) can generate transferable adversarial\nexamples which, despite their high average $L_\\infty$ distortion, have minimal\nvisual distortion. These results call into question the use of $L_\\infty$ as a\nsole measure for visual distortion, and further demonstrate the power of EAD at\ngenerating robust adversarial examples.\n",
"title": "Attacking the Madry Defense Model with $L_1$-based Adversarial Examples"
}
| null | null | null | null | true | null |
11350
| null |
Default
| null | null |
null |
{
"abstract": " Difficult problems described in terms of interacting quantum fields evolving\nin real time or out of equilibrium are abound in condensed-matter and\nhigh-energy physics. Addressing such problems via controlled experiments in\natomic, molecular, and optical physics would be a breakthrough in the field of\nquantum simulations. In this work, we present a quantum-sensing protocol to\nmeasure the generating functional of an interacting quantum field theory and,\nwith it, all the relevant information about its in or out of equilibrium\nphenomena. Our protocol can be understood as a collective interferometric\nscheme based on a generalization of the notion of Schwinger sources in quantum\nfield theories, which make it possible to probe the generating functional. We\nshow that our scheme can be realized in crystals of trapped ions acting as\nanalog quantum simulators of self-interacting scalar quantum field theories.\n",
"title": "Quantum sensors for the generating functional of interacting quantum field theories"
}
| null | null | null | null | true | null |
11351
| null |
Default
| null | null |
null |
{
"abstract": " We describe Sockeye (version 1.12), an open-source sequence-to-sequence\ntoolkit for Neural Machine Translation (NMT). Sockeye is a production-ready\nframework for training and applying models as well as an experimental platform\nfor researchers. Written in Python and built on MXNet, the toolkit offers\nscalable training and inference for the three most prominent encoder-decoder\narchitectures: attentional recurrent neural networks, self-attentional\ntransformers, and fully convolutional networks. Sockeye also supports a wide\nrange of optimizers, normalization and regularization techniques, and inference\nimprovements from current NMT literature. Users can easily run standard\ntraining recipes, explore different model settings, and incorporate new ideas.\nIn this paper, we highlight Sockeye's features and benchmark it against other\nNMT toolkits on two language arcs from the 2017 Conference on Machine\nTranslation (WMT): English-German and Latvian-English. We report competitive\nBLEU scores across all three architectures, including an overall best score for\nSockeye's transformer implementation. To facilitate further comparison, we\nrelease all system outputs and training scripts used in our experiments. The\nSockeye toolkit is free software released under the Apache 2.0 license.\n",
"title": "Sockeye: A Toolkit for Neural Machine Translation"
}
| null | null | null | null | true | null |
11352
| null |
Default
| null | null |
null |
{
"abstract": " Shape information is of great importance in many applications. For example,\nthe oil-bearing capacity of sand bodies, the subterranean remnants of ancient\nrivers, is related to their cross-sectional shapes. The analysis of these\nshapes is therefore of some interest, but current classifications are\nsimplistic and ad hoc. In this paper, we describe the first steps towards a\ncoherent statistical analysis of these shapes by deriving the integrated\nlikelihood for data shapes given class parameters. The result is of interest\nbeyond this particular application.\n",
"title": "Bayesian shape modelling of cross-sectional geological data"
}
| null | null | null | null | true | null |
11353
| null |
Default
| null | null |
null |
{
"abstract": " In the present day, AES is one the most widely used and most secure\nEncryption Systems prevailing. So, naturally lots of research work is going on\nto mount a significant attack on AES. Many different forms of Linear and\ndifferential cryptanalysis have been performed on AES. Of late, an active area\nof research has been Algebraic Cryptanalysis of AES, where although fast\nprogress is being made, there are still numerous scopes for research and\nimprovement. One of the major reasons behind this being that algebraic\ncryptanalysis mainly depends on I/O relations of the AES S- Box (a major\ncomponent of the AES). As, already known, that the key recovery algorithm of\nAES can be broken down as an MQ problem which is itself considered hard.\nSolving these equations depends on our ability reduce them into linear forms\nwhich are easily solvable under our current computational prowess. The lower\nthe degree of these equations, the easier it is for us to linearlize hence the\nattack complexity reduces. The aim of this paper is to analyze the various\nrelations involving small number of monomials of the AES S- Box and to answer\nthe question whether it is actually possible to have such monomial equations\nfor the S- Box if we restrict the degree of the monomials. In other words this\npaper aims to study such equations and see if they can be applicable for AES.\n",
"title": "Analysing Relations involving small number of Monomials in AES S- Box"
}
| null | null | null | null | true | null |
11354
| null |
Default
| null | null |
null |
{
"abstract": " We establish a natural connection of the $q$-Virasoro algebra $D_{q}$\nintroduced by Belov and Chaltikian with affine Kac-Moody Lie algebras. More\nspecifically, for each abelian group $S$ together with a one-to-one linear\ncharacter $\\chi$, we define an infinite-dimensional Lie algebra $D_{S}$ which\nreduces to $D_{q}$ when $S=\\mathbb{Z}$. Guided by the theory of equivariant\nquasi modules for vertex algebras, we introduce another Lie algebra\n${\\mathfrak{g}}_{S}$ with $S$ as an automorphism group and we prove that\n$D_{S}$ is isomorphic to the $S$-covariant algebra of the affine Lie algebra\n$\\widehat{\\mathfrak{g}_{S}}$. We then relate restricted $D_{S}$-modules of\nlevel $\\ell\\in \\mathbb{C}$ to equivariant quasi modules for the vertex algebra\n$V_{\\widehat{\\mathfrak{g}_{S}}}(\\ell,0)$ associated to\n$\\widehat{\\mathfrak{g}_{S}}$ with level $\\ell$. Furthermore, we show that if\n$S$ is a finite abelian group of order $2l+1$, $D_{S}$ is isomorphic to the\naffine Kac-Moody algebra of type $B^{(1)}_{l}$.\n",
"title": "q-Virasoro algebra and affine Kac-Moody Lie algebras"
}
| null | null |
[
"Mathematics"
] | null | true | null |
11355
| null |
Validated
| null | null |
null |
{
"abstract": " This paper examines the noise handling properties of three of the most widely\nused algorithms for numerically inverting the Laplace Transform. After\nexamining the genesis of the algorithms, the regularization properties are\nevaluated through a series of standard test functions in which noise is added\nto the inverse transform. Comparisons are then made with the exact data. Our\nmain finding is that the Talbot inversion algorithm is very good at handling\nnoisy data and performs much better than the Fourier Series and Stehfest\nnumerical inversion schemes as outlined in this paper. This offers a\nconsiderable advantage for it's use in inverting the Laplace Transform when\nseeking numerical solutions to time dependent differential equations.\n",
"title": "The Noise Handling Properties of the Talbot Algorithm for Numerically Inverting the Laplace Transform"
}
| null | null | null | null | true | null |
11356
| null |
Default
| null | null |
null |
{
"abstract": " Despite its ubiquity in our daily lives, AI is only just starting to make\nadvances in what may arguably have the largest societal impact thus far, the\nnascent field of autonomous driving. In this work we discuss this important\ntopic and address one of crucial aspects of the emerging area, the problem of\npredicting future state of autonomous vehicle's surrounding necessary for safe\nand efficient operations. We introduce a deep learning-based approach that\ntakes into account current world state and produces rasterized representations\nof each actor's vicinity. The raster images are then used by deep convolutional\nmodels to infer future movement of actors while accounting for inherent\nuncertainty of the prediction task. Extensive experiments on real-world data\nstrongly suggest benefits of the proposed approach. Moreover, following\nsuccessful tests the system was deployed to a fleet of autonomous vehicles.\n",
"title": "Short-term Motion Prediction of Traffic Actors for Autonomous Driving using Deep Convolutional Networks"
}
| null | null | null | null | true | null |
11357
| null |
Default
| null | null |
null |
{
"abstract": " D. Grigoriev-G. Koshevoy recently proved that tropical Schur polynomials have\n(at worst) polynomial tropical semiring complexity. They also conjectured\ntropical skew Schur polynomials have at least exponential complexity; we\nestablish a polynomial complexity upper bound. Our proof uses results about\n(stable) Schubert polynomials, due to R. P. Stanley and S. Billey-W.\nJockusch-R. P. Stanley, together with a sufficient condition for polynomial\ncomplexity that is connected to the saturated Newton polytope property.\n",
"title": "Tropicalization, symmetric polynomials, and complexity"
}
| null | null | null | null | true | null |
11358
| null |
Default
| null | null |
null |
{
"abstract": " We study the normal closure of a big power of one or several Dehn twists in a\nMapping Class Group. We prove that it has a presentation whose relators\nconsists only of commutators between twists of disjoint support, thus answering\na question of Ivanov. Our method is to use the theory of projection complexes\nof Bestvina Bromberg and Fujiwara, together with the theory of rotating\nfamilies, simultaneously on several spaces.\n",
"title": "The normal closure of big Dehn twists, and plate spinning with rotating families"
}
| null | null | null | null | true | null |
11359
| null |
Default
| null | null |
null |
{
"abstract": " Cyber Physical Systems (CPS) are becoming ubiquitous and affect the physical\nworld, yet security is seldom at the forefront of their design. This is\nespecially true of robotic control algorithms which seldom consider the effect\nof a cyber attack on mission objectives and success. This work presents a\nsecure optimal control algorithm in the face of a cyber attack on a robot's\nknowledge of the environment. This work focuses on cyber attack, but the\nresults generalize to incomplete or outdated information of an environment.\nThis work fuses ideas from robust control, optimal control, and sensor based\nplanning to provide a generalization of stopping distance in 3D. The planner is\nimplemented in simulation and its properties are analyzed.\n",
"title": "Secure Minimum Time Planning Under Environmental Uncertainty: an Extended Treatment"
}
| null | null | null | null | true | null |
11360
| null |
Default
| null | null |
null |
{
"abstract": " Treatment effects can be estimated from observational data as the difference\nin potential outcomes. In this paper, we address the challenge of estimating\nthe potential outcome when treatment-dose levels can vary continuously over\ntime. Further, the outcome variable may not be measured at a regular frequency.\nOur proposed solution represents the treatment response curves using linear\ntime-invariant dynamical systems---this provides a flexible means for modeling\nresponse over time to highly variable dose curves. Moreover, for multivariate\ndata, the proposed method: uncovers shared structure in treatment response and\nthe baseline across multiple markers; and, flexibly models challenging\ncorrelation structure both across and within signals over time. For this, we\nbuild upon the framework of multiple-output Gaussian Processes. On simulated\nand a challenging clinical dataset, we show significant gains in accuracy over\nstate-of-the-art models.\n",
"title": "Treatment-Response Models for Counterfactual Reasoning with Continuous-time, Continuous-valued Interventions"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
11361
| null |
Validated
| null | null |
null |
{
"abstract": " Analytical electron microscopy and spectroscopy of biological specimens,\npolymers, and other beam sensitive materials has been a challenging area due to\nirradiation damage. There is a pressing need to develop novel imaging and\nspectroscopic imaging methods that will minimize such sample damage as well as\nreduce the data acquisition time. The latter is useful for high-throughput\nanalysis of materials structure and chemistry. In this work, we present a novel\nmachine learning based method for dynamic sparse sampling of EDS data using a\nscanning electron microscope. Our method, based on the supervised learning\napproach for dynamic sampling algorithm and neural networks based\nclassification of EDS data, allows a dramatic reduction in the total sampling\nof up to 90%, while maintaining the fidelity of the reconstructed elemental\nmaps and spectroscopic data. We believe this approach will enable imaging and\nelemental mapping of materials that would otherwise be inaccessible to these\nanalysis techniques.\n",
"title": "Reduced Electron Exposure for Energy-Dispersive Spectroscopy using Dynamic Sampling"
}
| null | null | null | null | true | null |
11362
| null |
Default
| null | null |
null |
{
"abstract": " We consider the problem of global optimization of an unknown non-convex\nsmooth function with zeroth-order feedback. In this setup, an algorithm is\nallowed to adaptively query the underlying function at different locations and\nreceives noisy evaluations of function values at the queried points (i.e. the\nalgorithm has access to zeroth-order information). Optimization performance is\nevaluated by the expected difference of function values at the estimated\noptimum and the true optimum. In contrast to the classical optimization setup,\nfirst-order information like gradients are not directly accessible to the\noptimization algorithm. We show that the classical minimax framework of\nanalysis, which roughly characterizes the worst-case query complexity of an\noptimization algorithm in this setting, leads to excessively pessimistic\nresults. We propose a local minimax framework to study the fundamental\ndifficulty of optimizing smooth functions with adaptive function evaluations,\nwhich provides a refined picture of the intrinsic difficulty of zeroth-order\noptimization. We show that for functions with fast level set growth around the\nglobal minimum, carefully designed optimization algorithms can identify a near\nglobal minimizer with many fewer queries. For the special case of strongly\nconvex and smooth functions, our implied convergence rates match the ones\ndeveloped for zeroth-order convex optimization problems. At the other end of\nthe spectrum, for worst-case smooth functions no algorithm can converge faster\nthan the minimax rate of estimating the entire unknown function in the\n$\\ell_\\infty$-norm. We provide an intuitive and efficient algorithm that\nattains the derived upper error bounds.\n",
"title": "Optimization of Smooth Functions with Noisy Observations: Local Minimax Rates"
}
| null | null | null | null | true | null |
11363
| null |
Default
| null | null |
null |
{
"abstract": " This study proposes a fully convolutional network (FCN) model for raw\nwaveform-based speech enhancement. The proposed system performs speech\nenhancement in an end-to-end (i.e., waveform-in and waveform-out) manner, which\ndif-fers from most existing denoising methods that process the magnitude\nspectrum (e.g., log power spectrum (LPS)) only. Because the fully connected\nlayers, which are involved in deep neural networks (DNN) and convolutional\nneural networks (CNN), may not accurately characterize the local information of\nspeech signals, particularly with high frequency components, we employed fully\nconvolutional layers to model the waveform. More specifically, FCN consists of\nonly convolutional layers and thus the local temporal structures of speech\nsignals can be efficiently and effectively preserved with relatively few\nweights. Experimental results show that DNN- and CNN-based models have limited\ncapability to restore high frequency components of waveforms, thus leading to\ndecreased intelligibility of enhanced speech. By contrast, the proposed FCN\nmodel can not only effectively recover the waveforms but also outperform the\nLPS-based DNN baseline in terms of short-time objective intelligibility (STOI)\nand perceptual evaluation of speech quality (PESQ). In addition, the number of\nmodel parameters in FCN is approximately only 0.2% compared with that in both\nDNN and CNN.\n",
"title": "Raw Waveform-based Speech Enhancement by Fully Convolutional Networks"
}
| null | null |
[
"Computer Science",
"Statistics"
] | null | true | null |
11364
| null |
Validated
| null | null |
null |
{
"abstract": " Recent technological development has enabled researchers to study social\nphenomena scientifically in detail and financial markets has particularly\nattracted physicists since the Brownian motion has played the key role as in\nphysics. In our previous report (arXiv:1703.06739; to appear in Phys. Rev.\nLett.), we have presented a microscopic model of trend-following high-frequency\ntraders (HFTs) and its theoretical relation to the dynamics of financial\nBrownian motion, directly supported by a data analysis of tracking trajectories\nof individual HFTs in a financial market. Here we show the mathematical\nfoundation for the HFT model paralleling to the traditional kinetic theory in\nstatistical physics. We first derive the time-evolution equation for the\nphase-space distribution for the HFT model exactly, which corresponds to the\nLiouville equation in conventional analytical mechanics. By a systematic\nreduction of the Liouville equation for the HFT model, the\nBogoliubov-Born-Green-Kirkwood-Yvon hierarchal equations are derived for\nfinancial Brownian motion. We then derive the Boltzmann-like and Langevin-like\nequations for the order-book and the price dynamics by making the assumption of\nmolecular chaos. The qualitative behavior of the model is asymptotically\nstudied by solving the Boltzmann-like and Langevin-like equations for the large\nnumber of HFTs, which is numerically validated through the Monte-Carlo\nsimulation. Our kinetic description highlights the parallel mathematical\nstructure between the financial Brownian motion and the physical Brownian\nmotion.\n",
"title": "Kinetic Theory for Finance Brownian Motion from Microscopic Dynamics"
}
| null | null |
[
"Quantitative Finance"
] | null | true | null |
11365
| null |
Validated
| null | null |
null |
{
"abstract": " Modern technology for producing extremely bright and coherent X-ray laser\npulses provides the possibility to acquire a large number of diffraction\npatterns from individual biological nanoparticles, including proteins, viruses,\nand DNA. These two-dimensional diffraction patterns can be practically\nreconstructed and retrieved down to a resolution of a few \\angstrom. In\nprinciple, a sufficiently large collection of diffraction patterns will contain\nthe required information for a full three-dimensional reconstruction of the\nbiomolecule. The computational methodology for this reconstruction task is\nstill under development and highly resolved reconstructions have not yet been\nproduced.\nWe analyze the Expansion-Maximization-Compression scheme, the current state\nof the art approach for this very challenging application, by isolating\ndifferent sources of uncertainty. Through numerical experiments on synthetic\ndata we evaluate their respective impact. We reach conclusions of relevance for\nhandling actual experimental data, as well as pointing out certain improvements\nto the underlying estimation algorithm.\nWe also introduce a practically applicable computational methodology in the\nform of bootstrap procedures for assessing reconstruction uncertainty in the\nreal data case. We evaluate the sharpness of this approach and argue that this\ntype of procedure will be critical in the near future when handling the\nincreasing amount of data.\n",
"title": "Assessing Uncertainties in X-ray Single-particle Three-dimensional reconstructions"
}
| null | null | null | null | true | null |
11366
| null |
Default
| null | null |
null |
{
"abstract": " Many real-world applications require robust algorithms to learn point\nprocesses based on a type of incomplete data --- the so-called short\ndoubly-censored (SDC) event sequences. We study this critical problem of\nquantitative asynchronous event sequence analysis under the framework of Hawkes\nprocesses by leveraging the idea of data synthesis. Given SDC event sequences\nobserved in a variety of time intervals, we propose a sampling-stitching data\nsynthesis method --- sampling predecessors and successors for each SDC event\nsequence from potential candidates and stitching them together to synthesize\nlong training sequences. The rationality and the feasibility of our method are\ndiscussed in terms of arguments based on likelihood. Experiments on both\nsynthetic and real-world data demonstrate that the proposed data synthesis\nmethod improves learning results indeed for both time-invariant and\ntime-varying Hawkes processes.\n",
"title": "Learning Hawkes Processes from Short Doubly-Censored Event Sequences"
}
| null | null | null | null | true | null |
11367
| null |
Default
| null | null |
null |
{
"abstract": " The recombination of charges is an important process in organic photonic\ndevices because the process influences the device characteristics such as the\ndriving voltage, efficiency and lifetime. By combining the dipole trap theory\nwith the drift-diffusion model, we report that the stationary dipole moment\n({\\mu}0) of the dopant is a major factor determining the recombination\nmechanism in the dye-doped organic light emitting diodes when the trap depth\n({\\Delta}Et) is larger than 0.3 eV where any de-trapping effect becomes\nnegligible. Dopants with large {\\mu}0 (e.g., homoleptic Ir(III) dyes) induce\nlarge charge trapping on them, resulting in high driving voltage and\ntrap-assisted-recombination dominated emission. On the other hand, dyes with\nsmall {\\mu}0 (e.g., heteroleptic Ir(III) dyes) show much less trapping on them\nno matter what {\\Delta}Et is, leading to lower driving voltage, higher\nefficiencies and Langevin recombination dominated emission characteristics.\nThis finding will be useful in any organic photonic devices where trapping and\nrecombination sites play key roles.\n",
"title": "Unveiling the Role of Dopant Polarity on the Recombination, and Performance of Organic Light-Emitting Diodes"
}
| null | null |
[
"Physics"
] | null | true | null |
11368
| null |
Validated
| null | null |
null |
{
"abstract": " Gaussian mixture models (GMM) are powerful parametric tools with many\napplications in machine learning and computer vision. Expectation maximization\n(EM) is the most popular algorithm for estimating the GMM parameters. However,\nEM guarantees only convergence to a stationary point of the log-likelihood\nfunction, which could be arbitrarily worse than the optimal solution. Inspired\nby the relationship between the negative log-likelihood function and the\nKullback-Leibler (KL) divergence, we propose an alternative formulation for\nestimating the GMM parameters using the sliced Wasserstein distance, which\ngives rise to a new algorithm. Specifically, we propose minimizing the\nsliced-Wasserstein distance between the mixture model and the data distribution\nwith respect to the GMM parameters. In contrast to the KL-divergence, the\nenergy landscape for the sliced-Wasserstein distance is more well-behaved and\ntherefore more suitable for a stochastic gradient descent scheme to obtain the\noptimal GMM parameters. We show that our formulation results in parameter\nestimates that are more robust to random initializations and demonstrate that\nit can estimate high-dimensional data distributions more faithfully than the EM\nalgorithm.\n",
"title": "Sliced Wasserstein Distance for Learning Gaussian Mixture Models"
}
| null | null | null | null | true | null |
11369
| null |
Default
| null | null |
null |
{
"abstract": " We study the effect of constant shifts on the zeros of rational harmomic\nfunctions $f(z) = r(z) - \\conj{z}$. In particular, we characterize how shifting\nthrough the caustics of $f$ changes the number of zeros and their respective\norientations. This also yields insight into the nature of the singular zeros of\n$f$. Our results have applications in gravitational lensing theory, where\ncertain such functions $f$ represent gravitational point-mass lenses, and a\nconstant shift can be interpreted as the position of the light source of the\nlens.\n",
"title": "How constant shifts affect the zeros of certain rational harmonic functions"
}
| null | null | null | null | true | null |
11370
| null |
Default
| null | null |
null |
{
"abstract": " Joint visual attention is characterized by two or more individuals looking at\na common target at the same time. The ability to identify joint attention in\nscenes, the people involved, and their common target, is fundamental to the\nunderstanding of social interactions, including others' intentions and goals.\nIn this work we deal with the extraction of joint attention events, and the use\nof such events for image descriptions. The work makes two novel contributions.\nFirst, our extraction algorithm is the first which identifies joint visual\nattention in single static images. It computes 3D gaze direction, identifies\nthe gaze target by combining gaze direction with a 3D depth map computed for\nthe image, and identifies the common gaze target. Second, we use a human study\nto demonstrate the sensitivity of humans to joint attention, suggesting that\nthe detection of such a configuration in an image can be useful for\nunderstanding the image, including the goals of the agents and their joint\nactivity, and therefore can contribute to image captioning and related tasks.\n",
"title": "Discovery and usage of joint attention in images"
}
| null | null | null | null | true | null |
11371
| null |
Default
| null | null |
null |
{
"abstract": " We consider the IBVP in exterior domains for the p-Laplacian parabolic\nsystem. We prove regularity up to the boundary, extinction properties for p \\in\n( 2n/(n+2) , 2n/(n+1) ) and exponential decay for p= 2n/(n+1) .\n",
"title": "Singular p-Laplacian parabolic system in exterior domains: higher regularity of solutions and related properties of extinction and asymptotic behavior in time"
}
| null | null | null | null | true | null |
11372
| null |
Default
| null | null |
null |
{
"abstract": " Using a sample of galaxies selected from the Sloan Digital Sky Survey Data\nRelease 7 (SDSS DR7) and a catalog of bulge-disk decompositions, we study how\nthe size distribution of galaxies depends on the intrinsic properties of\ngalaxies, such as concentration, morphology, specific star formation rate\n(sSFR), and bulge fraction, and on the large-scale environments in the context\nof central/satellite decomposition, halo environment, the cosmic web: \\cluster,\n\\filament, \\sheet ~and \\void, as well as galaxy number density. We find that\nthere is a strong dependence of the luminosity- or mass-size relation on the\ngalaxy concentration, morphology, sSFR, and bulge fraction. Compared with\nlate-type (spiral) galaxies, there is a clear trend of smaller sizes and\nsteeper slope for early-type (elliptical) galaxies. Similarly, galaxies with\nhigh bulge fraction have smaller sizes and steeper slope than those with low\nbulge fraction. Fitting formula of the average luminosity- and mass-size\nrelations are provided for galaxies of these different intrinsic properties.\nExamining galaxies in terms of their large scale environments, we find that the\nmass-size relation has some weak dependence on the halo mass and\ncentral/satellite segregation for galaxies within mass range $9.0\\le \\log\nM_{\\ast} \\le 10.5$, where satellites or galaxies in more massive halos have\nslightly smaller sizes than their counterparts. While the cosmic web and local\nnumber density dependence of the mass-size relation is almost negligible.\n",
"title": "Size distribution of galaxies in SDSS DR7: weak dependence on halo environment"
}
| null | null | null | null | true | null |
11373
| null |
Default
| null | null |
null |
{
"abstract": " Deep network pruning is an effective method to reduce the storage and\ncomputation cost of deep neural networks when applying them to resource-limited\ndevices. Among many pruning granularities, neuron level pruning will remove\nredundant neurons and filters in the model and result in thinner networks. In\nthis paper, we propose a gradually global pruning scheme for neuron level\npruning. In each pruning step, a small percent of neurons were selected and\ndropped across all layers in the model. We also propose a simple method to\neliminate the biases in evaluating the importance of neurons to make the scheme\nfeasible. Compared with layer-wise pruning scheme, our scheme avoid the\ndifficulty in determining the redundancy in each layer and is more effective\nfor deep networks. Our scheme would automatically find a thinner sub-network in\noriginal network under a given performance.\n",
"title": "Towards thinner convolutional neural networks through Gradually Global Pruning"
}
| null | null | null | null | true | null |
11374
| null |
Default
| null | null |
null |
{
"abstract": " We propose a systematic learning-based approach to the generation of massive\nquantities of synthetic 3D scenes and arbitrary numbers of photorealistic 2D\nimages thereof, with associated ground truth information, for the purposes of\ntraining, benchmarking, and diagnosing learning-based computer vision and\nrobotics algorithms. In particular, we devise a learning-based pipeline of\nalgorithms capable of automatically generating and rendering a potentially\ninfinite variety of indoor scenes by using a stochastic grammar, represented as\nan attributed Spatial And-Or Graph, in conjunction with state-of-the-art\nphysics-based rendering. Our pipeline is capable of synthesizing scene layouts\nwith high diversity, and it is configurable inasmuch as it enables the precise\ncustomization and control of important attributes of the generated scenes. It\nrenders photorealistic RGB images of the generated scenes while automatically\nsynthesizing detailed, per-pixel ground truth data, including visible surface\ndepth and normal, object identity, and material information (detailed to object\nparts), as well as environments (e.g., illuminations and camera viewpoints). We\ndemonstrate the value of our synthesized dataset, by improving performance in\ncertain machine-learning-based scene understanding tasks--depth and surface\nnormal prediction, semantic segmentation, reconstruction, etc.--and by\nproviding benchmarks for and diagnostics of trained models by modifying object\nattributes and scene properties in a controllable manner.\n",
"title": "Configurable 3D Scene Synthesis and 2D Image Rendering with Per-Pixel Ground Truth using Stochastic Grammars"
}
| null | null | null | null | true | null |
11375
| null |
Default
| null | null |
null |
{
"abstract": " Vision sensors lie in the heart of computer vision. In many computer vision\napplications, such as AR/VR, non-contacting near-field communication (NFC) with\nhigh throughput is required to transfer information to algorithms. In this\nwork, we proposed a novel NFC system which utilizes multiple frequency bands to\nachieve high throughput.\n",
"title": "Multiband NFC for High-Throughput Wireless Computer Vision Sensor Network"
}
| null | null | null | null | true | null |
11376
| null |
Default
| null | null |
null |
{
"abstract": " We propose a methodology that adapts graph embedding techniques (DeepWalk\n(Perozzi et al., 2014) and node2vec (Grover and Leskovec, 2016)) as well as\ncross-lingual vector space mapping approaches (Least Squares and Canonical\nCorrelation Analysis) in order to merge the corpus and ontological sources of\nlexical knowledge. We also perform comparative analysis of the used algorithms\nin order to identify the best combination for the proposed system. We then\napply this to the task of enhancing the coverage of an existing word\nembedding's vocabulary with rare and unseen words. We show that our technique\ncan provide considerable extra coverage (over 99%), leading to consistent\nperformance gain (around 10% absolute gain is achieved with w2v-gn-500K cf.§\n3.3) on the Rare Word Similarity dataset.\n",
"title": "Learning Rare Word Representations using Semantic Bridging"
}
| null | null | null | null | true | null |
11377
| null |
Default
| null | null |
null |
{
"abstract": " The electrical conductivity and dielectric properties of Ni1.5Fe1.5O4 ferrite\nhas been controlled by varying the annealing temperature of the chemical routed\nsamples. The frequency activated conductivity obeyed Jonschers power law and\nuniversal scaling suggested semiconductor nature. An unusual metal like state\nhas been revealed in the measurement temperature scale in between two\nsemiconductor states with different activation energy. The metal like state has\nbeen affected by thermal annealing of the material. The analysis of electrical\nimpedance and modulus spectra has confirmed non-Debye dielectric relaxation\nwith contributions from grains and grain boundaries. The dielectric relaxation\nprocess is thermally activated in terms of measurement temperature and\nannealing temperature of the samples. The hole hopping process, due to presence\nof Ni3+ ions in the present Ni rich ferrite, played a significant role in\ndetermining the thermal activated conduction mechanism. This work has\nsuccessfully applied the technique of a combined variation of annealing\ntemperature and pH value during chemical reaction for tuning electrical\nparameters in a wide range; for example dc limit of conductivity 10power(-4)\n-10power(-12) S/cm, and unusually high activation energy 0.17-1.36 eV.\n",
"title": "Effect of annealing temperatures on the electrical conductivity and dielectric properties of Ni1.5Fe1.5O4 spinel ferrite prepared by chemical reaction at different pH values"
}
| null | null | null | null | true | null |
11378
| null |
Default
| null | null |
null |
{
"abstract": " One of the varieties of pores, often found in natural or artificial building\nmaterials, are the so-called blind pores of dead-end or saccate type.\nThree-dimensional model of such kind of pore has been developed in this work.\nThis model has been used for simulation of water vapor interaction with\nindividual pore by molecular dynamics in combination with the diffusion\nequation method. Special investigations have been done to find dependencies\nbetween thermostats implementations and conservation of thermodynamic and\nstatistical values of water vapor - pore system. The two types of evolution of\nwater-pore system have been investigated: drying and wetting of the pore. Full\nresearch of diffusion coefficient, diffusion velocity and other diffusion\nparameters has been made.\n",
"title": "Molecular dynamic simulation of water vapor interaction with blind pore of dead-end and saccate type"
}
| null | null | null | null | true | null |
11379
| null |
Default
| null | null |
null |
{
"abstract": " Successful programs are written to be maintained. One aspect to this is that\nprogrammers order the components in the code files in a particular way. This is\npart of programming style. While the conventions for ordering are sometimes\ngiven as part of a style guideline, such guidelines are often incomplete and\nprogrammers tend to have their own more comprehensive orderings in mind. This\npaper defines a model for ordering program components and shows how this model\ncan be learned from sample code. Such a model is a useful tool for a\nprogramming environment in that it can be used to find the proper location for\ninserting new components or for reordering files to better meet the needs of\nthe programmer. The model is designed so that it can be fine- tuned by the\nprogrammer. The learning framework is evaluated both by looking at code with\nknown style guidelines and by testing whether it inserts existing components\ninto a file correctly.\n",
"title": "Learning Program Component Order"
}
| null | null | null | null | true | null |
11380
| null |
Default
| null | null |
null |
{
"abstract": " Over the last decade, both the neural network and kernel adaptive filter have\nsuccessfully been used for nonlinear signal processing. However, they suffer\nfrom high computational cost caused by their complex/growing network\nstructures. In this paper, we propose two random Euler filters for\ncomplex-valued nonlinear filtering problem, i.e., linear random Euler\ncomplex-valued filter (LRECF) and its widely-linear version (WLRECF), which\npossess a simple and fixed network structure. The transient and steady-state\nperformances are studied in a non-stationary environment. The analytical\nminimum mean square error (MSE) and optimum step-size are derived. Finally,\nnumerical simulations on complex-valued nonlinear system identification and\nnonlinear channel equalization are presented to show the effectiveness of the\nproposed methods.\n",
"title": "Random Euler Complex-Valued Nonlinear Filters"
}
| null | null | null | null | true | null |
11381
| null |
Default
| null | null |
null |
{
"abstract": " Memory has a great impact on the evolution of every process related to human\nsocieties. Among them, the evolution of an epidemic is directly related to the\nindividuals' experiences. Indeed, any real epidemic process is clearly\nsustained by a non-Markovian dynamics: memory effects play an essential role in\nthe spreading of diseases. Including memory effects in the\nsusceptible-infected-recovered (SIR) epidemic model seems very appropriate for\nsuch an investigation. Thus, the memory prone SIR model dynamics is\ninvestigated using fractional derivatives. The decay of long-range memory,\ntaken as a power-law function, is directly controlled by the order of the\nfractional derivatives in the corresponding nonlinear fractional differential\nevolution equations. Here we assume \"fully mixed\" approximation and show that\nthe epidemic threshold is shifted to higher values than those for the\nmemoryless system, depending on this memory \"length\" decay exponent. We also\nconsider the SIR model on structured networks and study the effect of topology\non threshold points in a non- Markovian dynamics. Furthermore, the lack of\naccess to the precise information about the initial conditions or the past\nevents plays a very relevant role in the correct estimation or prediction of\nthe epidemic evolution. Such a \"constraint\" is analyzed and discussed.\n",
"title": "Memory effects on epidemic evolution: The susceptible-infected-recovered epidemic model"
}
| null | null | null | null | true | null |
11382
| null |
Default
| null | null |
null |
{
"abstract": " The Camassa-Holm equation and its two-component Camassa-Holm system\ngeneralization both experience wave breaking in finite time. To analyze this,\nand to obtain solutions past wave breaking, it is common to reformulate the\noriginal equation given in Eulerian coordinates, into a system of ordinary\ndifferential equations in Lagrangian coordinates. It is of considerable\ninterest to study the stability of solutions and how this is manifested in\nEulerian and Lagrangian variables. We identify criteria of convergence, such\nthat convergence in Eulerian coordinates is equivalent to convergence in\nLagrangian coordinates. In addition, we show how one can approximate global\nconservative solutions of the scalar Camassa-Holm equation by smooth solutions\nof the two-component Camassa-Holm system that do not experience wave breaking.\n",
"title": "On the equivalence of Eulerian and Lagrangian variables for the two-component Camassa-Holm system"
}
| null | null | null | null | true | null |
11383
| null |
Default
| null | null |
null |
{
"abstract": " In the hydrodynamic regime, the evolution of a stochastic lattice gas with\nsymmetric hopping rules is described by a diffusion equation with\ndensity-dependent diffusion coefficient encapsulating all microscopic details\nof the dynamics. This diffusion coefficient is, in principle, determined by a\nGreen-Kubo formula. In practice, even when the equilibrium properties of a\nlattice gas are analytically known, the diffusion coefficient cannot be\ncomputed except when a lattice gas additionally satisfies the gradient\ncondition. We develop a procedure to systematically obtain analytical\napproximations for the diffusion coefficient for non-gradient lattice gases\nwith known equilibrium. The method relies on a variational formula found by\nVaradhan and Spohn which is a version of the Green-Kubo formula particularly\nsuitable for diffusive lattice gases. Restricting the variational formula to\nfinite-dimensional sub-spaces allows one to perform the minimization and gives\nupper bounds for the diffusion coefficient. We apply this approach to a\nkinetically constrained non-gradient lattice gas, viz. to the Kob-Andersen\nmodel on the square lattice.\n",
"title": "Bulk diffusion in a kinetically constrained lattice gas"
}
| null | null | null | null | true | null |
11384
| null |
Default
| null | null |
null |
{
"abstract": " Max-mixture processes are defined as Z = max(aX, (1 -- a)Y) with X an\nasymptotic dependent (AD) process, Y an asymptotic independent (AI) process and\na $\\in$ [0, 1]. So that, the mixing coefficient a may reveal the strength of\nthe AD part present in the max-mixture process. In this paper we focus on two\ntests based on censored pairwise likelihood estimates. We compare their\nperformance through an extensive simulation study. Monte Carlo simulation plays\na fundamental tool for asymptotic variance calculations. We apply our tests to\ndaily precipitations from the East of Australia. Drawbacks and possible\ndevelopments are discussed.\n",
"title": "Censored pairwise likelihood-based tests for mixing coefficient of spatial max-mixture models"
}
| null | null |
[
"Mathematics",
"Statistics"
] | null | true | null |
11385
| null |
Validated
| null | null |
null |
{
"abstract": " The purpose of this paper is to point out a new connection between\ninformation theory and dynamical systems. In the information theory side, we\nconsider rate distortion theory, which studies lossy data compression of\nstochastic processes under distortion constraints. In the dynamical systems\nside, we consider mean dimension theory, which studies how many parameters per\nsecond we need to describe a dynamical system. The main results are new\nvariational principles connecting rate distortion function to metric mean\ndimension.\n",
"title": "From rate distortion theory to metric mean dimension: variational principle"
}
| null | null |
[
"Computer Science",
"Mathematics"
] | null | true | null |
11386
| null |
Validated
| null | null |
null |
{
"abstract": " Quantized Neural Networks (QNNs), which use low bitwidth numbers for\nrepresenting parameters and performing computations, have been proposed to\nreduce the computation complexity, storage size and memory usage. In QNNs,\nparameters and activations are uniformly quantized, such that the\nmultiplications and additions can be accelerated by bitwise operations.\nHowever, distributions of parameters in Neural Networks are often imbalanced,\nsuch that the uniform quantization determined from extremal values may under\nutilize available bitwidth. In this paper, we propose a novel quantization\nmethod that can ensure the balance of distributions of quantized values. Our\nmethod first recursively partitions the parameters by percentiles into balanced\nbins, and then applies uniform quantization. We also introduce computationally\ncheaper approximations of percentiles to reduce the computation overhead\nintroduced. Overall, our method improves the prediction accuracies of QNNs\nwithout introducing extra computation during inference, has negligible impact\non training speed, and is applicable to both Convolutional Neural Networks and\nRecurrent Neural Networks. Experiments on standard datasets including ImageNet\nand Penn Treebank confirm the effectiveness of our method. On ImageNet, the\ntop-5 error rate of our 4-bit quantized GoogLeNet model is 12.7\\%, which is\nsuperior to the state-of-the-arts of QNNs.\n",
"title": "Balanced Quantization: An Effective and Efficient Approach to Quantized Neural Networks"
}
| null | null |
[
"Computer Science"
] | null | true | null |
11387
| null |
Validated
| null | null |
null |
{
"abstract": " The Frame Problem (FP) is a puzzle in philosophy of mind and epistemology,\narticulated by the Stanford Encyclopedia of Philosophy as follows: \"How do we\naccount for our apparent ability to make decisions on the basis only of what is\nrelevant to an ongoing situation without having explicitly to consider all that\nis not relevant?\" In this work, we focus on the causal variant of the FP, the\nCausal Frame Problem (CFP). Assuming that a reasoner's mental causal model can\nbe (implicitly) represented by a causal Bayes net, we first introduce a notion\ncalled Potential Level (PL). PL, in essence, encodes the relative position of a\nnode with respect to its neighbors in a causal Bayes net. Drawing on the\npsychological literature on causal judgment, we substantiate the claim that PL\nmay bear on how time is encoded in the mind. Using PL, we propose an inference\nframework, called the PL-based Inference Framework (PLIF), which permits a\nboundedly-rational approach to the CFP to be formally articulated at Marr's\nalgorithmic level of analysis. We show that our proposed framework, PLIF, is\nconsistent with a wide range of findings in causal judgment literature, and\nthat PL and PLIF make a number of predictions, some of which are already\nsupported by existing findings.\n",
"title": "The Causal Frame Problem: An Algorithmic Perspective"
}
| null | null | null | null | true | null |
11388
| null |
Default
| null | null |
null |
{
"abstract": " In this paper we present a data visualization method together with its\npotential usefulness in digital humanities and philosophy of language. We\ncompile a multilingual parallel corpus from different versions of\nWittgenstein's Tractatus Logico-Philosophicus, including the original in German\nand translations into English, Spanish, French, and Russian. Using this corpus,\nwe compute a similarity measure between propositions and render a visual\nnetwork of relations for different languages.\n",
"title": "A Visual Representation of Wittgenstein's Tractatus Logico-Philosophicus"
}
| null | null | null | null | true | null |
11389
| null |
Default
| null | null |
null |
{
"abstract": " If the very early Universe is dominated by the non-minimally coupled Higgs\nfield and Starobinsky's curvature-squared term together, the potential diagram\nwould mimic the landscape of a valley, serving as a cosmological attractor. The\ninflationary dynamics along this valley is studied, model parameters are\nconstrained against observational data, and the isocurvature perturbation is\nevaluated.\n",
"title": "Primordial perturbations generated by Higgs field and $R^2$ operator"
}
| null | null | null | null | true | null |
11390
| null |
Default
| null | null |
null |
{
"abstract": " We examine the relationship between the (double) Schubert polynomials of\nBilley-Haiman and Ikeda-Mihalcea-Naruse and the (double) theta and eta\npolynomials of Buch-Kresch-Tamvakis and Wilson from the perspective of Weyl\ngroup invariants. We obtain generators for the kernel of the natural map from\nthe corresponding ring of Schubert polynomials to the (equivariant) cohomology\nring of symplectic and orthogonal flag manifolds.\n",
"title": "Schubert polynomials, theta and eta polynomials, and Weyl group invariants"
}
| null | null | null | null | true | null |
11391
| null |
Default
| null | null |
null |
{
"abstract": " During inflation, massive fields can contribute to the power spectrum of\ncurvature perturbation via a dimension-5 operator. This contribution can be\nconsidered as a bias for the program of using $n_s$ and $r$ to select inflation\nmodels. Even the dimension-5 operator is suppressed by $\\Lambda = M_p$, there\nis still a significant shift on the $n_s$-$r$ diagram if the massive fields\nhave $m\\sim H$. On the other hand, if the heavy degree of freedom appears only\nat the same energy scale as the suppression scale of the dimension-5 operator,\nthen significant shift on the $n_s$-$r$ diagram takes place at $m=\\Lambda \\sim\n70H$, which is around the inflationary time-translation symmetry breaking\nscale. Hence, the systematics from massive fields pose a greater challenge for\nfuture high precision experiments for inflationary model selection. This result\ncan be thought of as the impact of UV sensitivity to inflationary observables.\n",
"title": "Massive Fields as Systematics for Single Field Inflation"
}
| null | null | null | null | true | null |
11392
| null |
Default
| null | null |
null |
{
"abstract": " These lecture notes are concerned with the solvability of the second boundary\nvalue problem of the prescribed affine mean curvature equation and related\nregularity theory of the Monge-Ampère and linearized Monge-Ampère\nequations. The prescribed affine mean curvature equation is a fully nonlinear,\nfourth order, geometric partial differential equation of the following form\n$$\\sum_{i, j=1}^n U^{ij}\\frac{\\partial^2}{\\partial\n{x_i}\\partial{x_j}}\\left[(\\det D^2 u)^{-\\frac{n+1}{n+2}}\\right]=f$$ where\n$(U^{ij})$ is the cofactor matrix of the Hessian matrix $D^2 u$ of a locally\nuniformly convex function $u$. Its variant is related to the problem of finding\nKähler metrics of constant scalar curvature in complex geometry. We first\nintroduce the background of the prescribed affine mean curvature equation which\ncan be viewed as a coupled system of Monge-Ampère and linearized\nMonge-Ampère equations. Then we state key open problems and present the\nsolution of the second boundary value problem that prescribes the boundary\nvalues of the solution $u$ and its Hessian determinant $\\det D^2 u$. Its proof\nuses important tools from the boundary regularity theory of the Monge-Ampère\nand linearized Monge-Ampère equations that we will present in the lecture\nnotes.\n",
"title": "The second boundary value problem of the prescribed affine mean curvature equation and related linearized Monge-Ampère equation"
}
| null | null | null | null | true | null |
11393
| null |
Default
| null | null |
null |
{
"abstract": " This text contains over three hundred specific open questions on various\ntopics in additive combinatorics, each placed in context by reviewing all\nrelevant results. While the primary purpose is to provide an ample supply of\nproblems for student research, it is hopefully also useful for a wider\naudience. It is the author's intention to keep the material current, thus all\nfeedback and updates are greatly appreciated.\n",
"title": "Additive Combinatorics: A Menu of Research Problems"
}
| null | null | null | null | true | null |
11394
| null |
Default
| null | null |
null |
{
"abstract": " We present $^{77}$Se-NMR measurements on single-crystalline FeSe under\npressures up to 2 GPa. Based on the observation of the splitting and broadening\nof the NMR spectrum due to structural twin domains, we discovered that static,\nlocal nematic ordering exists well above the bulk nematic ordering temperature,\n$T_{\\rm s}$. The static, local nematic order and the low-energy stripe-type\nantiferromagnetic spin fluctuations, as revealed by NMR spin-lattice relaxation\nrate measurements, are both insensitive to pressure application. These NMR\nresults provide clear evidence for the microscopic cooperation between\nmagnetism and local nematicity in FeSe.\n",
"title": "NMR evidence for static local nematicity and its cooperative interplay with low-energy magnetic fluctuations in FeSe under pressure"
}
| null | null | null | null | true | null |
11395
| null |
Default
| null | null |
null |
{
"abstract": " The present study proposes LitStoryTeller, an interactive system for visually\nexploring the semantic structure of a scientific article. We demonstrate how\nLitStoryTeller could be used to answer some of the most fundamental research\nquestions, such as how a new method was built on top of existing methods, based\non what theoretical proof and experimental evidences. More importantly,\nLitStoryTeller can assist users to understand the full and interesting story a\nscientific paper, with a concise outline and important details. The proposed\nsystem borrows a metaphor from screen play, and visualizes the storyline of a\nscientific paper by arranging its characters (scientific concepts or\nterminologies) and scenes (paragraphs/sentences) into a progressive and\ninteractive storyline. Such storylines help to preserve the semantic structure\nand logical thinking process of a scientific paper. Semantic structures, such\nas scientific concepts and comparative sentences, are extracted using existing\nnamed entity recognition APIs and supervised classifiers, from a scientific\npaper automatically. Two supplementary views, ranked entity frequency view and\nentity co-occurrence network view, are provided to help users identify the\n\"main plot\" of such scientific storylines. When collective documents are ready,\nLitStoryTeller also provides a temporal entity evolution view and entity\ncommunity view for collection digestion.\n",
"title": "LitStoryTeller: An Interactive System for Visual Exploration of Scientific Papers Leveraging Named entities and Comparative Sentences"
}
| null | null |
[
"Computer Science"
] | null | true | null |
11396
| null |
Validated
| null | null |
null |
{
"abstract": " Conventional sound shielding structures typically prevent fluid transport\nbetween the exterior and interior. A design of a two-dimensional acoustic\nmetacage with subwavelength thickness which can shield acoustic waves from all\ndirections while allowing steady fluid flow is presented in this paper. The\nstructure is designed based on acoustic gradient-index metasurfaces composed of\nopen channels and shunted Helmholtz resonators. The strong parallel momentum on\nthe metacage surface rejects in-plane sound at an arbitrary angle of incidence\nwhich leads to low sound transmission through the metacage. The performance of\nthe proposed metacage is verified by numerical simulations and measurements on\na three-dimensional printed prototype. The acoustic metacage has potential\napplications in sound insulation where steady fluid flow is necessary or\nadvantageous.\n",
"title": "Acoustic Metacages for Omnidirectional Sound Shielding"
}
| null | null | null | null | true | null |
11397
| null |
Default
| null | null |
null |
{
"abstract": " Traditional dictionary learning methods are based on quadratic convex loss\nfunction and thus are sensitive to outliers. In this paper, we propose a\ngeneric framework for robust dictionary learning based on concave losses. We\nprovide results on composition of concave functions, notably regarding\nsuper-gradient computations, that are key for developing generic dictionary\nlearning algorithms applicable to smooth and non-smooth losses. In order to\nimprove identification of outliers, we introduce an initialization heuristic\nbased on undercomplete dictionary learning. Experimental results using\nsynthetic and real data demonstrate that our method is able to better detect\noutliers, is capable of generating better dictionaries, outperforming\nstate-of-the-art methods such as K-SVD and LC-KSVD.\n",
"title": "Concave losses for robust dictionary learning"
}
| null | null | null | null | true | null |
11398
| null |
Default
| null | null |
null |
{
"abstract": " We introduce a stop-code tolerant (SCT) approach to training recurrent\nconvolutional neural networks for lossy image compression. Our methods\nintroduce a multi-pass training method to combine the training goals of\nhigh-quality reconstructions in areas around stop-code masking as well as in\nhighly-detailed areas. These methods lead to lower true bitrates for a given\nrecursion count, both pre- and post-entropy coding, even using unstructured\nLZ77 code compression. The pre-LZ77 gains are achieved by trimming stop codes.\nThe post-LZ77 gains are due to the highly unequal distributions of 0/1 codes\nfrom the SCT architectures. With these code compressions, the SCT architecture\nmaintains or exceeds the image quality at all compression rates compared to\nJPEG and to RNN auto-encoders across the Kodak dataset. In addition, the SCT\ncoding results in lower variance in image quality across the extent of the\nimage, a characteristic that has been shown to be important in human ratings of\nimage quality\n",
"title": "Target-Quality Image Compression with Recurrent, Convolutional Neural Networks"
}
| null | null | null | null | true | null |
11399
| null |
Default
| null | null |
null |
{
"abstract": " We introduce dual matroids of 2-dimensional simplicial complexes. Under\ncertain necessary conditions, duals matroids are used to characterise\nembeddability in 3-space in a way analogous to Whitney's planarity criterion.\nWe further use dual matroids to extend a 3-dimensional analogue of\nKuratowski's theorem to the class of 2-dimensional simplicial complexes\nobtained from simply connected ones by identifying vertices or edges.\n",
"title": "Embedding simply connected 2-complexes in 3-space -- IV. Dual matroids"
}
| null | null | null | null | true | null |
11400
| null |
Default
| null | null |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.