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": " Physical-layer group secret-key (GSK) generation is an effective way of\ngenerating secret keys in wireless networks, wherein the nodes exploit inherent\nrandomness in the wireless channels to generate group keys, which are\nsubsequently applied to secure messages while broadcasting, relaying, and other\nnetwork-level communications. While existing GSK protocols focus on securing\nthe common source of randomness from external eavesdroppers, they assume that\nthe legitimate nodes of the group are trusted. In this paper, we address\ninsider attacks from the legitimate participants of the wireless network during\nthe key generation process. Instead of addressing conspicuous attacks such as\nswitching-off communication, injecting noise, or denying consensus on group\nkeys, we introduce stealth attacks that can go undetected against\nstate-of-the-art GSK schemes. We propose two forms of attacks, namely: (i)\ndifferent-key attacks, wherein an insider attempts to generate different keys\nat different nodes, especially across nodes that are out of range so that they\nfail to recover group messages despite possessing the group key, and (ii)\nlow-rate key attacks, wherein an insider alters the common source of randomness\nso as to reduce the key-rate. We also discuss various detection techniques,\nwhich are based on detecting anomalies and inconsistencies on the channel\nmeasurements at the legitimate nodes. Through simulations we show that GSK\ngeneration schemes are vulnerable to insider-threats, especially on topologies\nthat cannot support additional secure links between neighbouring nodes to\nverify the attacks.\n",
"title": "Insider-Attacks on Physical-Layer Group Secret-Key Generation in Wireless Networks"
} | null | null | null | null | true | null | 2801 | null | Default | null | null |
null | {
"abstract": " Interactive reinforcement learning (IRL) extends traditional reinforcement\nlearning (RL) by allowing an agent to interact with parent-like trainers during\na task. In this paper, we present an IRL approach using dynamic audio-visual\ninput in terms of vocal commands and hand gestures as feedback. Our\narchitecture integrates multi-modal information to provide robust commands from\nmultiple sensory cues along with a confidence value indicating the\ntrustworthiness of the feedback. The integration process also considers the\ncase in which the two modalities convey incongruent information. Additionally,\nwe modulate the influence of sensory-driven feedback in the IRL task using\ngoal-oriented knowledge in terms of contextual affordances. We implement a\nneural network architecture to predict the effect of performed actions with\ndifferent objects to avoid failed-states, i.e., states from which it is not\npossible to accomplish the task. In our experimental setup, we explore the\ninterplay of multimodal feedback and task-specific affordances in a robot\ncleaning scenario. We compare the learning performance of the agent under four\ndifferent conditions: traditional RL, multi-modal IRL, and each of these two\nsetups with the use of contextual affordances. Our experiments show that the\nbest performance is obtained by using audio-visual feedback with\naffordancemodulated IRL. The obtained results demonstrate the importance of\nmulti-modal sensory processing integrated with goal-oriented knowledge in IRL\ntasks.\n",
"title": "Multi-modal Feedback for Affordance-driven Interactive Reinforcement Learning"
} | null | null | null | null | true | null | 2802 | null | Default | null | null |
null | {
"abstract": " As South and Central American countries prepare for increased birth defects\nfrom Zika virus outbreaks and plan for mitigation strategies to minimize\nongoing and future outbreaks, understanding important characteristics of Zika\noutbreaks and how they vary across regions is a challenging and important\nproblem. We developed a mathematical model for the 2015 Zika virus outbreak\ndynamics in Colombia, El Salvador, and Suriname. We fit the model to publicly\navailable data provided by the Pan American Health Organization, using\nApproximate Bayesian Computation to estimate parameter distributions and\nprovide uncertainty quantification. An important model input is the at-risk\nsusceptible population, which can vary with a number of factors including\nclimate, elevation, population density, and socio-economic status. We informed\nthis initial condition using the highest historically reported dengue incidence\nmodified by the probable dengue reporting rates in the chosen countries. The\nmodel indicated that a country-level analysis was not appropriate for Colombia.\nWe then estimated the basic reproduction number, or the expected number of new\nhuman infections arising from a single infected human, to range between 4 and 6\nfor El Salvador and Suriname with a median of 4.3 and 5.3, respectively. We\nestimated the reporting rate to be around 16% in El Salvador and 18% in\nSuriname with estimated total outbreak sizes of 73,395 and 21,647 people,\nrespectively. The uncertainty in parameter estimates highlights a need for\nresearch and data collection that will better constrain parameter ranges.\n",
"title": "Estimating the reproductive number, total outbreak size, and reporting rates for Zika epidemics in South and Central America"
} | null | null | null | null | true | null | 2803 | null | Default | null | null |
null | {
"abstract": " In this study, we systematically investigate the impact of class imbalance on\nclassification performance of convolutional neural networks (CNNs) and compare\nfrequently used methods to address the issue. Class imbalance is a common\nproblem that has been comprehensively studied in classical machine learning,\nyet very limited systematic research is available in the context of deep\nlearning. In our study, we use three benchmark datasets of increasing\ncomplexity, MNIST, CIFAR-10 and ImageNet, to investigate the effects of\nimbalance on classification and perform an extensive comparison of several\nmethods to address the issue: oversampling, undersampling, two-phase training,\nand thresholding that compensates for prior class probabilities. Our main\nevaluation metric is area under the receiver operating characteristic curve\n(ROC AUC) adjusted to multi-class tasks since overall accuracy metric is\nassociated with notable difficulties in the context of imbalanced data. Based\non results from our experiments we conclude that (i) the effect of class\nimbalance on classification performance is detrimental; (ii) the method of\naddressing class imbalance that emerged as dominant in almost all analyzed\nscenarios was oversampling; (iii) oversampling should be applied to the level\nthat completely eliminates the imbalance, whereas the optimal undersampling\nratio depends on the extent of imbalance; (iv) as opposed to some classical\nmachine learning models, oversampling does not cause overfitting of CNNs; (v)\nthresholding should be applied to compensate for prior class probabilities when\noverall number of properly classified cases is of interest.\n",
"title": "A systematic study of the class imbalance problem in convolutional neural networks"
} | null | null | null | null | true | null | 2804 | null | Default | null | null |
null | {
"abstract": " Let $\\Gamma \\leq \\mathrm{Aut}(T_{d_1}) \\times \\mathrm{Aut}(T_{d_2})$ be a\ngroup acting freely and transitively on the product of two regular trees of\ndegree $d_1$ and $d_2$. We develop an algorithm which computes the closure of\nthe projection of $\\Gamma$ on $\\mathrm{Aut}(T_{d_t})$ under the hypothesis that\n$d_t \\geq 6$ is even and that the local action of $\\Gamma$ on $T_{d_t}$\ncontains $\\mathrm{Alt}(d_t)$. We show that if $\\Gamma$ is torsion-free and $d_1\n= d_2 = 6$, exactly seven closed subgroups of $\\mathrm{Aut}(T_6)$ arise in this\nway. We also construct two new infinite families of virtually simple lattices\nin $\\mathrm{Aut}(T_{6}) \\times \\mathrm{Aut}(T_{4n})$ and in\n$\\mathrm{Aut}(T_{2n}) \\times \\mathrm{Aut}(T_{2n+1})$ respectively, for all $n\n\\geq 2$. In particular we provide an explicit presentation of a torsion-free\ninfinite simple group on $5$ generators and $10$ relations, that splits as an\namalgamated free product of two copies of $F_3$ over $F_{11}$. We include\ninformation arising from computer-assisted exhaustive searches of lattices in\nproducts of trees of small degrees. In an appendix by Pierre-Emmanuel Caprace,\nsome of our results are used to show that abstract and relative commensurator\ngroups of free groups are almost simple, providing partial answers to questions\nof Lubotzky and Lubotzky-Mozes-Zimmer.\n",
"title": "New simple lattices in products of trees and their projections"
} | null | null | [
"Mathematics"
]
| null | true | null | 2805 | null | Validated | null | null |
null | {
"abstract": " We present in this paper algorithms for solving stiff PDEs on the unit sphere\nwith spectral accuracy in space and fourth-order accuracy in time. These are\nbased on a variant of the double Fourier sphere method in coefficient space\nwith multiplication matrices that differ from the usual ones, and\nimplicit-explicit time-stepping schemes. Operating in coefficient space with\nthese new matrices allows one to use a sparse direct solver, avoids the\ncoordinate singularity and maintains smoothness at the poles, while\nimplicit-explicit schemes circumvent severe restrictions on the time-steps due\nto stiffness. A comparison is made against exponential integrators and it is\nfound that implicit-explicit schemes perform best. Implementations in MATLAB\nand Chebfun make it possible to compute the solution of many PDEs to high\naccuracy in a very convenient fashion.\n",
"title": "Fourth-order time-stepping for stiff PDEs on the sphere"
} | null | null | null | null | true | null | 2806 | null | Default | null | null |
null | {
"abstract": " Machine learning algorithms for prediction are increasingly being used in\ncritical decisions affecting human lives. Various fairness formalizations, with\nno firm consensus yet, are employed to prevent such algorithms from\nsystematically discriminating against people based on certain attributes\nprotected by law. The aim of this article is to survey how fairness is\nformalized in the machine learning literature for the task of prediction and\npresent these formalizations with their corresponding notions of distributive\njustice from the social sciences literature. We provide theoretical as well as\nempirical critiques of these notions from the social sciences literature and\nexplain how these critiques limit the suitability of the corresponding fairness\nformalizations to certain domains. We also suggest two notions of distributive\njustice which address some of these critiques and discuss avenues for\nprospective fairness formalizations.\n",
"title": "On Formalizing Fairness in Prediction with Machine Learning"
} | null | null | null | null | true | null | 2807 | null | Default | null | null |
null | {
"abstract": " An introduction to the Zwanzig-Mori-Götze-Wölfle memory function\nformalism (or generalized Drude formalism) is presented. This formalism is used\nextensively in analyzing the experimentally obtained optical conductivity of\nstrongly correlated systems like cuprates and Iron based superconductors etc.\nFor a broader perspective both the generalised Langevin equation approach and\nthe projection operator approach for the memory function formalism are given.\nThe Götze-Wölfle perturbative expansion of memory function is presented\nand its application to the computation of the dynamical conductivity of metals\nis also reviewd. This review of the formalism contains all the mathematical\ndetails for pedagogical purposes.\n",
"title": "The Memory Function Formalism: A Review"
} | null | null | null | null | true | null | 2808 | null | Default | null | null |
null | {
"abstract": " The multilabel learning problem with large number of labels, features, and\ndata-points has generated a tremendous interest recently. A recurring theme of\nthese problems is that only a few labels are active in any given datapoint as\ncompared to the total number of labels. However, only a small number of\nexisting work take direct advantage of this inherent extreme sparsity in the\nlabel space. By the virtue of Restricted Isometry Property (RIP), satisfied by\nmany random ensembles, we propose a novel procedure for multilabel learning\nknown as RIPML. During the training phase, in RIPML, labels are projected onto\na random low-dimensional subspace followed by solving a least-square problem in\nthis subspace. Inference is done by a k-nearest neighbor (kNN) based approach.\nWe demonstrate the effectiveness of RIPML by conducting extensive simulations\nand comparing results with the state-of-the-art linear dimensionality reduction\nbased approaches.\n",
"title": "RIPML: A Restricted Isometry Property based Approach to Multilabel Learning"
} | null | null | null | null | true | null | 2809 | null | Default | null | null |
null | {
"abstract": " We correct one erroneous statement made in our recent paper \"Medial axis and\nsingularities\".\n",
"title": "Erratum to: Medial axis and singularities"
} | null | null | null | null | true | null | 2810 | null | Default | null | null |
null | {
"abstract": " Next-generation 802.11ax WLANs will make extensive use of multi-user\ncommunications in both downlink (DL) and uplink (UL) directions to achieve high\nand efficient spectrum utilization in scenarios with many user stations per\naccess point. It will become possible with the support of multi-user (MU)\nmultiple input, multiple output (MIMO) and orthogonal frequency division\nmultiple access (OFDMA) transmissions. In this paper, we first overview the\nnovel characteristics introduced by IEEE 802.11ax to implement AP-initiated\nOFDMA and MU-MIMO transmissions in both downlink and uplink directions. Namely,\nwe describe the changes made at the physical layer and at the medium access\ncontrol layer to support OFDMA, the use of \\emph{trigger frames} to schedule\nuplink multi-user transmissions, and the new \\emph{multi-user RTS/CTS\nmechanism} to protect large multi-user transmissions from collisions. Then, in\norder to study the achievable throughput of an 802.11ax network, we use both\nmathematical analysis and simulations to numerically quantify the benefits of\nMU transmissions and the impact of 802.11ax overheads on the WLAN saturation\nthroughput. Results show the advantages of MU transmissions in scenarios with\nmany user stations, also providing some novel insights on the conditions in\nwhich 802.11ax WLANs are able to maximize their performance, such as the\nexistence of an optimal number of active user stations in terms of throughput,\nor the need to provide strict prioritization to AP-initiated MU transmissions\nto avoid collisions with user stations.\n",
"title": "AP-initiated Multi-User Transmissions in IEEE 802.11ax WLANs"
} | null | null | null | null | true | null | 2811 | null | Default | null | null |
null | {
"abstract": " The belief that three dimensional space is infinite and flat in the absence\nof matter is a canon of physics that has been in place since the time of\nNewton. The assumption that space is flat at infinity has guided several modern\nphysical theories. But what do we actually know to support this belief? A\nsimple argument, called the \"Telescope Principle\", asserts that all that we can\nknow about space is bounded by observations. Physical theories are best when\nthey can be verified by observations, and that should also apply to the\ngeometry of space. The Telescope Principle is simple to state, but it leads to\nvery interesting insights into relativity and Yang-Mills theory via projective\nequivalences of their respective spaces.\n",
"title": "What do we know about the geometry of space?"
} | null | null | [
"Physics"
]
| null | true | null | 2812 | null | Validated | null | null |
null | {
"abstract": " Twinning is an important deformation mode of hexagonal close-packed metals.\nThe crystallographic theory is based on the 150-years old concept of simple\nshear. The habit plane of the twin is the shear plane, it is invariant. Here we\npresent Electron BackScatter Diffraction observations and crystallographic\nanalysis of a millimeter size twin in a magnesium single crystal whose straight\nhabit plane, unambiguously determined both the parent crystal and in its twin,\nis not an invariant plane. This experimental evidence demonstrates that\nmacroscopic deformation twinning can be obtained by a mechanism that is not a\nsimple shear. Beside, this unconventional twin is often co-formed with a new\nconventional twin that exhibits the lowest shear magnitude ever reported in\nmetals. The existence of unconventional twinning introduces a shift of paradigm\nand calls for the development of a new theory for the displacive\ntransformations\n",
"title": "Evidence of new twinning modes in magnesium questioning the shear paradigm"
} | null | null | null | null | true | null | 2813 | null | Default | null | null |
null | {
"abstract": " Femtosecond optical pulses at mid-infrared frequencies have opened up the\nnonlinear control of lattice vibrations in solids. So far, all applications\nhave relied on second order phonon nonlinearities, which are dominant at field\nstrengths near 1 MVcm-1. In this regime, nonlinear phononics can transiently\nchange the average lattice structure, and with it the functionality of a\nmaterial. Here, we achieve an order-of-magnitude increase in field strength,\nand explore higher-order lattice nonlinearities. We drive up to five phonon\nharmonics of the A1 mode in LiNbO3. Phase-sensitive measurements of atomic\ntrajectories in this regime are used to experimentally reconstruct the\ninteratomic potential and to benchmark ab-initio calculations for this\nmaterial. Tomography of the Free Energy surface by high-order nonlinear\nphononics will impact many aspects of materials research, including the study\nof classical and quantum phase transitions.\n",
"title": "Probing the Interatomic Potential of Solids by Strong-Field Nonlinear Phononics"
} | null | null | [
"Physics"
]
| null | true | null | 2814 | null | Validated | null | null |
null | {
"abstract": " In Optical diffraction tomography, the multiply scattered field is a\nnonlinear function of the refractive index of the object. The Rytov method is a\nlinear approximation of the forward model, and is commonly used to reconstruct\nimages. Recently, we introduced a reconstruction method based on the Beam\nPropagation Method (BPM) that takes the nonlinearity into account. We refer to\nthis method as Learning Tomography (LT). In this paper, we carry out\nsimulations in order to assess the performance of LT over the linear iterative\nmethod. Each algorithm has been rigorously assessed for spherical objects, with\nsynthetic data generated using the Mie theory. By varying the RI contrast and\nthe size of the objects, we show that the LT reconstruction is more accurate\nand robust than the reconstruction based on the linear model. In addition, we\nshow that LT is able to correct distortion that is evident in Rytov\napproximation due to limitations in phase unwrapping. More importantly, the\ncapacity of LT in handling multiple scattering problem are demonstrated by\nsimulations of multiple cylinders using the Mie theory and confirmed by\nexperimental results of two spheres.\n",
"title": "Assessment of learning tomography using Mie theory"
} | null | null | null | null | true | null | 2815 | null | Default | null | null |
null | {
"abstract": " Optical tweezers have enabled important insights into intracellular transport\nthrough the investigation of motor proteins, with their ability to manipulate\nparticles at the microscale, affording femto Newton force resolution. Its use\nto realize a constant force clamp has enabled vital insights into the behavior\nof motor proteins under different load conditions. However, the varying nature\nof disturbances and the effect of thermal noise pose key challenges to force\nregulation. Furthermore, often the main aim of many studies is to determine the\nmotion of the motor and the statistics related to the motion, which can be at\nodds with the force regulation objective. In this article, we propose a mixed\nobjective H2-Hinfinity optimization framework using a model-based design, that\nachieves the dual goals of force regulation and real time motion estimation\nwith quantifiable guarantees. Here, we minimize the Hinfinity norm for the\nforce regulation and error in step estimation while maintaining the H2 norm of\nthe noise on step estimate within user specified bounds. We demonstrate the\nefficacy of the framework through extensive simulations and an experimental\nimplementation using an optical tweezer setup with live samples of the motor\nprotein kinesin; where regulation of forces below 1 pico Newton with errors\nbelow 10 percent is obtained while simultaneously providing real time estimates\nof motor motion.\n",
"title": "Single Molecule Studies Under Constant Force Using Model Based Robust Control Design"
} | null | null | null | null | true | null | 2816 | null | Default | null | null |
null | {
"abstract": " In this Essay we investigate the observational signatures of Loop Quantum\nCosmology (LQC) in the CMB data. First, we concentrate on the dynamics of LQC\nand we provide the basic cosmological functions. We then obtain the power\nspectrum of scalar and tensor perturbations in order to study the performance\nof LQC against the latest CMB data. We find that LQC provides a robust\nprediction for the main slow-roll parameters, like the scalar spectral index\nand the tensor-to-scalar fluctuation ratio, which are in excellent agreement\nwithin $1\\sigma$ with the values recently measured by the Planck collaboration.\nThis result indicates that LQC can be seen as an alternative scenario with\nrespect to that of standard inflation.\n",
"title": "Measuring the effects of Loop Quantum Cosmology in the CMB data"
} | null | null | null | null | true | null | 2817 | null | Default | null | null |
null | {
"abstract": " In many modern machine learning applications, structures of underlying\nmathematical models often yield nonconvex optimization problems. Due to the\nintractability of nonconvexity, there is a rising need to develop efficient\nmethods for solving general nonconvex problems with certain performance\nguarantee. In this work, we investigate the accelerated proximal gradient\nmethod for nonconvex programming (APGnc). The method compares between a usual\nproximal gradient step and a linear extrapolation step, and accepts the one\nthat has a lower function value to achieve a monotonic decrease. In specific,\nunder a general nonsmooth and nonconvex setting, we provide a rigorous argument\nto show that the limit points of the sequence generated by APGnc are critical\npoints of the objective function. Then, by exploiting the\nKurdyka-{\\L}ojasiewicz (\\KL) property for a broad class of functions, we\nestablish the linear and sub-linear convergence rates of the function value\nsequence generated by APGnc. We further propose a stochastic variance reduced\nAPGnc (SVRG-APGnc), and establish its linear convergence under a special case\nof the \\KL property. We also extend the analysis to the inexact version of\nthese methods and develop an adaptive momentum strategy that improves the\nnumerical performance.\n",
"title": "Convergence Analysis of Proximal Gradient with Momentum for Nonconvex Optimization"
} | null | null | null | null | true | null | 2818 | null | Default | null | null |
null | {
"abstract": " Correlation networks were used to detect characteristics which, although\nfixed over time, have an important influence on the evolution of prices over\ntime. Potentially important features were identified using the websites and\nwhitepapers of cryptocurrencies with the largest userbases. These were assessed\nusing two datasets to enhance robustness: one with fourteen cryptocurrencies\nbeginning from 9 November 2017, and a subset with nine cryptocurrencies\nstarting 9 September 2016, both ending 6 March 2018. Separately analysing the\nsubset of cryptocurrencies raised the number of data points from 115 to 537,\nand improved robustness to changes in relationships over time. Excluding USD\nTether, the results showed a positive association between different\ncryptocurrencies that was statistically significant. Robust, strong positive\nassociations were observed for six cryptocurrencies where one was a fork of the\nother; Bitcoin / Bitcoin Cash was an exception. There was evidence for the\nexistence of a group of cryptocurrencies particularly associated with Cardano,\nand a separate group correlated with Ethereum. The data was not consistent with\na token's functionality or creation mechanism being the dominant determinants\nof the evolution of prices over time but did suggest that factors other than\nspeculation contributed to the price.\n",
"title": "Exploring the Interconnectedness of Cryptocurrencies using Correlation Networks"
} | null | null | null | null | true | null | 2819 | null | Default | null | null |
null | {
"abstract": " Existing neural conversational models process natural language primarily on a\nlexico-syntactic level, thereby ignoring one of the most crucial components of\nhuman-to-human dialogue: its affective content. We take a step in this\ndirection by proposing three novel ways to incorporate affective/emotional\naspects into long short term memory (LSTM) encoder-decoder neural conversation\nmodels: (1) affective word embeddings, which are cognitively engineered, (2)\naffect-based objective functions that augment the standard cross-entropy loss,\nand (3) affectively diverse beam search for decoding. Experiments show that\nthese techniques improve the open-domain conversational prowess of\nencoder-decoder networks by enabling them to produce emotionally rich responses\nthat are more interesting and natural.\n",
"title": "Affective Neural Response Generation"
} | null | null | null | null | true | null | 2820 | null | Default | null | null |
null | {
"abstract": " In this paper, energy efficient power allocation for downlink massive MIMO\nsystems is investigated. A constrained non-convex optimization problem is\nformulated to maximize the energy efficiency (EE), which takes into account the\nquality of service (QoS) requirements. By exploiting the properties of\nfractional programming and the lower bound of the user data rate, the\nnon-convex optimization problem is transformed into a convex optimization\nproblem. The Lagrangian dual function method is utilized to convert the\nconstrained convex problem into an unconstrained convex one. Due to the\nmulti-variable coupling problem caused by the intra-user interference, it is\nintractable to derive an explicit solution to the above optimization problem.\nExploiting the standard interference function, we propose an implicit iterative\nalgorithm to solve the unconstrained convex optimization problem and obtain the\noptimal power allocation scheme. Simulation results show that the proposed\niterative algorithm converges in just a few iterations, and demonstrate the\nimpact of the number of users and the number of antennas on the EE.\n",
"title": "Energy Efficient Power Allocation in Massive MIMO Systems based on Standard Interference Function"
} | null | null | null | null | true | null | 2821 | null | Default | null | null |
null | {
"abstract": " The composite fermion (CF) formalism produces wave functions that are not\nalways linearly independent. This is especially so in the low angular momentum\nregime in the lowest Landau level, where a subclass of CF states, known as\nsimple states, gives a good description of the low energy spectrum. For the\ntwo-component Bose gas, explicit bases avoiding the large number of redundant\nstates have been found. We generalize one of these bases to the $M$-component\nBose gas and prove its validity. We also show that the numbers of linearly\nindependent simple states for different values of angular momentum are given by\ncoefficients of $q$-multinomials.\n",
"title": "Composite fermion basis for M-component Bose gases"
} | null | null | null | null | true | null | 2822 | null | Default | null | null |
null | {
"abstract": " In the $k$-Cut problem, we are given an edge-weighted graph $G$ and an\ninteger $k$, and have to remove a set of edges with minimum total weight so\nthat $G$ has at least $k$ connected components. Prior work on this problem\ngives, for all $h \\in [2,k]$, a $(2-h/k)$-approximation algorithm for $k$-cut\nthat runs in time $n^{O(h)}$. Hence to get a $(2 - \\varepsilon)$-approximation\nalgorithm for some absolute constant $\\varepsilon$, the best runtime using\nprior techniques is $n^{O(k\\varepsilon)}$. Moreover, it was recently shown that\ngetting a $(2 - \\varepsilon)$-approximation for general $k$ is NP-hard,\nassuming the Small Set Expansion Hypothesis.\nIf we use the size of the cut as the parameter, an FPT algorithm to find the\nexact $k$-Cut is known, but solving the $k$-Cut problem exactly is $W[1]$-hard\nif we parameterize only by the natural parameter of $k$. An immediate question\nis: \\emph{can we approximate $k$-Cut better in FPT-time, using $k$ as the\nparameter?}\nWe answer this question positively. We show that for some absolute constant\n$\\varepsilon > 0$, there exists a $(2 - \\varepsilon)$-approximation algorithm\nthat runs in time $2^{O(k^6)} \\cdot \\widetilde{O} (n^4)$. This is the first FPT\nalgorithm that is parameterized only by $k$ and strictly improves the\n$2$-approximation.\n",
"title": "An FPT Algorithm Beating 2-Approximation for $k$-Cut"
} | null | null | null | null | true | null | 2823 | null | Default | null | null |
null | {
"abstract": " This paper analyses in detail the dynamics in a neighbourhood of a\nGénot-Brogliato point, colloquially termed the G-spot, which physically\nrepresents so-called dynamic jam in rigid body mechanics with unilateral\ncontact and Coulomb friction. Such singular points arise in planar rigid body\nproblems with slipping point contacts at the intersection between the\nconditions for onset of lift-off and for the Painlevé paradox. The G-spot can\nbe approached in finite time by an open set of initial conditions in a general\nclass of problems. The key question addressed is what happens next. In\nprinciple trajectories could, at least instantaneously, lift off, continue in\nslip, or undergo a so-called impact without collision. Such impacts are\nnon-local in momentum space and depend on properties evaluated away from the\nG-spot. The results are illustrated on a particular physical example, namely\nthe a frictional impact oscillator first studied by Leine et al.\nThe answer is obtained via an analysis that involves a consistent contact\nregularisation with a stiffness proportional to $1/\\varepsilon^2$. Taking a\nsingular limit as $\\varepsilon \\to 0$, one finds an inner and an outer\nasymptotic zone in the neighbourhood of the G-spot. Two distinct cases are\nfound according to whether the contact force becomes infinite or remains finite\nas the G-spot is approached. In the former case it is argued that there can be\nno such canards and so an impact without collision must occur. In the latter\ncase, the canard trajectory acts as a dividing surface between trajectories\nthat momentarily lift off and those that do not before taking the impact. The\norientation of the initial condition set leading to each eventuality is shown\nto change each time a certain positive parameter $\\beta$ passes through an\ninteger.\n",
"title": "Dynamics beyond dynamic jam; unfolding the Painlevé paradox singularity"
} | null | null | null | null | true | null | 2824 | null | Default | null | null |
null | {
"abstract": " This paper deals with existence and regularity of positive solutions of\nsingular elliptic problems on a smooth bounded domain with Dirichlet boundary\nconditions involving the $\\Phi$-Laplacian operator. The proof of existence is\nbased on a variant of the generalized Galerkin method that we developed\ninspired on ideas by Browder and a comparison principle. By using a kind of\nMoser iteration scheme we show $L^{\\infty}(\\Omega)$-regularity for positive\nsolutions\n",
"title": "Existence and regularity of positive solutions of quasilinear elliptic problems with singular semilinear term"
} | null | null | null | null | true | null | 2825 | null | Default | null | null |
null | {
"abstract": " Among the manifold takes on world literature, it is our goal to contribute to\nthe discussion from a digital point of view by analyzing the representation of\nworld literature in Wikipedia with its millions of articles in hundreds of\nlanguages. As a preliminary, we introduce and compare three different\napproaches to identify writers on Wikipedia using data from DBpedia, a\ncommunity project with the goal of extracting and providing structured\ninformation from Wikipedia. Equipped with our basic set of writers, we analyze\nhow they are represented throughout the 15 biggest Wikipedia language versions.\nWe combine intrinsic measures (mostly examining the connectedness of articles)\nwith extrinsic ones (analyzing how often articles are frequented by readers)\nand develop methods to evaluate our results. The better part of our findings\nseems to convey a rather conservative, old-fashioned version of world\nliterature, but a version derived from reproducible facts revealing an implicit\nliterary canon based on the editing and reading behavior of millions of people.\nWhile still having to solve some known issues, the introduced methods will help\nus build an observatory of world literature to further investigate its\nrepresentativeness and biases.\n",
"title": "World Literature According to Wikipedia: Introduction to a DBpedia-Based Framework"
} | null | null | [
"Computer Science"
]
| null | true | null | 2826 | null | Validated | null | null |
null | {
"abstract": " We investigate task clustering for deep-learning based multi-task and\nfew-shot learning in a many-task setting. We propose a new method to measure\ntask similarities with cross-task transfer performance matrix for the deep\nlearning scenario. Although this matrix provides us critical information\nregarding similarity between tasks, its asymmetric property and unreliable\nperformance scores can affect conventional clustering methods adversely.\nAdditionally, the uncertain task-pairs, i.e., the ones with extremely\nasymmetric transfer scores, may collectively mislead clustering algorithms to\noutput an inaccurate task-partition. To overcome these limitations, we propose\na novel task-clustering algorithm by using the matrix completion technique. The\nproposed algorithm constructs a partially-observed similarity matrix based on\nthe certainty of cluster membership of the task-pairs. We then use a matrix\ncompletion algorithm to complete the similarity matrix. Our theoretical\nanalysis shows that under mild constraints, the proposed algorithm will\nperfectly recover the underlying \"true\" similarity matrix with a high\nprobability. Our results show that the new task clustering method can discover\ntask clusters for training flexible and superior neural network models in a\nmulti-task learning setup for sentiment classification and dialog intent\nclassification tasks. Our task clustering approach also extends metric-based\nfew-shot learning methods to adapt multiple metrics, which demonstrates\nempirical advantages when the tasks are diverse.\n",
"title": "Robust Task Clustering for Deep Many-Task Learning"
} | null | null | null | null | true | null | 2827 | null | Default | null | null |
null | {
"abstract": " The roles played by learning and memorization represent an important topic in\ndeep learning research. Recent work on this subject has shown that the\noptimization behavior of DNNs trained on shuffled labels is qualitatively\ndifferent from DNNs trained with real labels. Here, we propose a novel\npermutation approach that can differentiate memorization from learning in deep\nneural networks (DNNs) trained as usual (i.e., using the real labels to guide\nthe learning, rather than shuffled labels). The evaluation of weather the DNN\nhas learned and/or memorized, happens in a separate step where we compare the\npredictive performance of a shallow classifier trained with the features\nlearned by the DNN, against multiple instances of the same classifier, trained\non the same input, but using shuffled labels as outputs. By evaluating these\nshallow classifiers in validation sets that share structure with the training\nset, we are able to tell apart learning from memorization. Application of our\npermutation approach to multi-layer perceptrons and convolutional neural\nnetworks trained on image data corroborated many findings from other groups.\nMost importantly, our illustrations also uncovered interesting dynamic patterns\nabout how DNNs memorize over increasing numbers of training epochs, and support\nthe surprising result that DNNs are still able to learn, rather than only\nmemorize, when trained with pure Gaussian noise as input.\n",
"title": "Detecting Learning vs Memorization in Deep Neural Networks using Shared Structure Validation Sets"
} | null | null | null | null | true | null | 2828 | null | Default | null | null |
null | {
"abstract": " Evaluating the return on ad spend (ROAS), the causal effect of advertising on\nsales, is critical to advertisers for understanding the performance of their\nexisting marketing strategy as well as how to improve and optimize it. Media\nMix Modeling (MMM) has been used as a convenient analytical tool to address the\nproblem using observational data. However it is well recognized that MMM\nsuffers from various fundamental challenges: data collection, model\nspecification and selection bias due to ad targeting, among others\n\\citep{chan2017,wolfe2016}.\nIn this paper, we study the challenge associated with measuring the impact of\nsearch ads in MMM, namely the selection bias due to ad targeting. Using causal\ndiagrams of the search ad environment, we derive a statistically principled\nmethod for bias correction based on the \\textit{back-door} criterion\n\\citep{pearl2013causality}. We use case studies to show that the method\nprovides promising results by comparison with results from randomized\nexperiments. We also report a more complex case study where the advertiser had\nspent on more than a dozen media channels but results from a randomized\nexperiment are not available. Both our theory and empirical studies suggest\nthat in some common, practical scenarios, one may be able to obtain an\napproximately unbiased estimate of search ad ROAS.\n",
"title": "Bias Correction For Paid Search In Media Mix Modeling"
} | null | null | null | null | true | null | 2829 | null | Default | null | null |
null | {
"abstract": " Neville's algorithm is known to provide an efficient and numerically stable\nsolution for polynomial interpolations. In this paper, an extension of this\nalgorithm is presented which includes the derivatives of the interpolating\npolynomial.\n",
"title": "Neville's algorithm revisited"
} | null | null | [
"Computer Science"
]
| null | true | null | 2830 | null | Validated | null | null |
null | {
"abstract": " Traditional linear methods for forecasting multivariate time series are not\nable to satisfactorily model the non-linear dependencies that may exist in\nnon-Gaussian series. We build on the theory of learning vector-valued functions\nin the reproducing kernel Hilbert space and develop a method for learning\nprediction functions that accommodate such non-linearities. The method not only\nlearns the predictive function but also the matrix-valued kernel underlying the\nfunction search space directly from the data. Our approach is based on learning\nmultiple matrix-valued kernels, each of those composed of a set of input\nkernels and a set of output kernels learned in the cone of positive\nsemi-definite matrices. In addition to superior predictive performance in the\npresence of strong non-linearities, our method also recovers the hidden dynamic\nrelationships between the series and thus is a new alternative to existing\ngraphical Granger techniques.\n",
"title": "Forecasting and Granger Modelling with Non-linear Dynamical Dependencies"
} | null | null | null | null | true | null | 2831 | null | Default | null | null |
null | {
"abstract": " The continually increasing number of documents produced each year\nnecessitates ever improving information processing methods for searching,\nretrieving, and organizing text. Central to these information processing\nmethods is document classification, which has become an important application\nfor supervised learning. Recently the performance of these traditional\nclassifiers has degraded as the number of documents has increased. This is\nbecause along with this growth in the number of documents has come an increase\nin the number of categories. This paper approaches this problem differently\nfrom current document classification methods that view the problem as\nmulti-class classification. Instead we perform hierarchical classification\nusing an approach we call Hierarchical Deep Learning for Text classification\n(HDLTex). HDLTex employs stacks of deep learning architectures to provide\nspecialized understanding at each level of the document hierarchy.\n",
"title": "HDLTex: Hierarchical Deep Learning for Text Classification"
} | null | null | null | null | true | null | 2832 | null | Default | null | null |
null | {
"abstract": " We present a method for efficient learning of control policies for multiple\nrelated robotic motor skills. Our approach consists of two stages, joint\ntraining and specialization training. During the joint training stage, a neural\nnetwork policy is trained with minimal information to disambiguate the motor\nskills. This forces the policy to learn a common representation of the\ndifferent tasks. Then, during the specialization training stage we selectively\nsplit the weights of the policy based on a per-weight metric that measures the\ndisagreement among the multiple tasks. By splitting part of the control policy,\nit can be further trained to specialize to each task. To update the control\npolicy during learning, we use Trust Region Policy Optimization with\nGeneralized Advantage Function (TRPOGAE). We propose a modification to the\ngradient update stage of TRPO to better accommodate multi-task learning\nscenarios. We evaluate our approach on three continuous motor skill learning\nproblems in simulation: 1) a locomotion task where three single legged robots\nwith considerable difference in shape and size are trained to hop forward, 2) a\nmanipulation task where three robot manipulators with different sizes and joint\ntypes are trained to reach different locations in 3D space, and 3) locomotion\nof a two-legged robot, whose range of motion of one leg is constrained in\ndifferent ways. We compare our training method to three baselines. The first\nbaseline uses only joint training for the policy, the second trains independent\npolicies for each task, and the last randomly selects weights to split. We show\nthat our approach learns more efficiently than each of the baseline methods.\n",
"title": "Multi-task Learning with Gradient Guided Policy Specialization"
} | null | null | null | null | true | null | 2833 | null | Default | null | null |
null | {
"abstract": " We prove upper bounds for the mean square of the remainder in the prime\ngeodesic theorem, for every cofinite Fuchsian group, which improve on average\non the best known pointwise bounds. The proof relies on the Selberg trace\nformula. For the modular group we prove a refined upper bound by using the\nKuznetsov trace formula.\n",
"title": "Mean square in the prime geodesic theorem"
} | null | null | null | null | true | null | 2834 | null | Default | null | null |
null | {
"abstract": " We define a novel, extensional, three-valued semantics for higher-order logic\nprograms with negation. The new semantics is based on interpreting the types of\nthe source language as three-valued Fitting-monotonic functions at all levels\nof the type hierarchy. We prove that there exists a bijection between such\nFitting-monotonic functions and pairs of two-valued-result functions where the\nfirst member of the pair is monotone-antimonotone and the second member is\nantimonotone-monotone. By deriving an extension of consistent approximation\nfixpoint theory (Denecker et al. 2004) and utilizing the above bijection, we\ndefine an iterative procedure that produces for any given higher-order logic\nprogram a distinguished extensional model. We demonstrate that this model is\nactually a minimal one. Moreover, we prove that our construction generalizes\nthe familiar well-founded semantics for classical logic programs, making in\nthis way our proposal an appealing formulation for capturing the well-founded\nsemantics for higher-order logic programs. This paper is under consideration\nfor acceptance in TPLP.\n",
"title": "Approximation Fixpoint Theory and the Well-Founded Semantics of Higher-Order Logic Programs"
} | null | null | null | null | true | null | 2835 | null | Default | null | null |
null | {
"abstract": " Spectroscopic surveys require fast and efficient analysis methods to maximize\ntheir scientific impact. Here we apply a deep neural network architecture to\nanalyze both SDSS-III APOGEE DR13 and synthetic stellar spectra. When our\nconvolutional neural network model (StarNet) is trained on APOGEE spectra, we\nshow that the stellar parameters (temperature, gravity, and metallicity) are\ndetermined with similar precision and accuracy as the APOGEE pipeline. StarNet\ncan also predict stellar parameters when trained on synthetic data, with\nexcellent precision and accuracy for both APOGEE data and synthetic data, over\na wide range of signal-to-noise ratios. In addition, the statistical\nuncertainties in the stellar parameter determinations are comparable to the\ndifferences between the APOGEE pipeline results and those determined\nindependently from optical spectra. We compare StarNet to other data-driven\nmethods; for example, StarNet and the Cannon 2 show similar behaviour when\ntrained with the same datasets, however StarNet performs poorly on small\ntraining sets like those used by the original Cannon. The influence of the\nspectral features on the stellar parameters is examined via partial derivatives\nof the StarNet model results with respect to the input spectra. While StarNet\nwas developed using the APOGEE observed spectra and corresponding ASSET\nsynthetic data, we suggest that this technique is applicable to other\nwavelength ranges and other spectral surveys.\n",
"title": "An Application of Deep Neural Networks in the Analysis of Stellar Spectra"
} | null | null | null | null | true | null | 2836 | null | Default | null | null |
null | {
"abstract": " Microservice Architecture (MSA) is a novel service-based architectural style\nfor distributed software systems. Compared to Service-oriented Architecture\n(SOA), MSA puts a stronger focus on self-containment of services. Each\nmicroservice is responsible for realizing exactly one business or technological\ncapability that is distinct from other services' capabilities. Additionally, on\nthe implementation and operation level, microservices are self-contained in\nthat they are developed, tested, deployed and operated independently from each\nother. Next to these characteristics that distinguish MSA from SOA, both\narchitectural styles rely on services as building blocks of distributed\nsoftware architecture and hence face similar challenges regarding, e.g.,\nservice identification, composition and provisioning. However, in contrast to\nMSA, SOA may rely on an extensive body of knowledge to tackle these challenges.\nThus, due to both architectural styles being service-based, the question arises\nto what degree MSA might draw on existing findings of SOA research and\npractice. In this paper we address this question in the field of Model-driven\nDevelopment (MDD) for design and operation of service-based architectures.\nTherefore, we present an analysis of existing MDD approaches to SOA, which\ncomprises the identification and semantic clustering of modeling concepts for\nSOA design and operation. For each concept cluster, the analysis assesses its\napplicability to MDD of MSA (MSA-MDD) and assigns it to a specific modeling\nviewpoint. The goal of the presented analysis is to provide a conceptual\nfoundation for an MSA-MDD metamodel.\n",
"title": "Analysis of Service-oriented Modeling Approaches for Viewpoint-specific Model-driven Development of Microservice Architecture"
} | null | null | null | null | true | null | 2837 | null | Default | null | null |
null | {
"abstract": " RoboJam is a machine-learning system for generating music that assists users\nof a touchscreen music app by performing responses to their short\nimprovisations. This system uses a recurrent artificial neural network to\ngenerate sequences of touchscreen interactions and absolute timings, rather\nthan high-level musical notes. To accomplish this, RoboJam's network uses a\nmixture density layer to predict appropriate touch interaction locations in\nspace and time. In this paper, we describe the design and implementation of\nRoboJam's network and how it has been integrated into a touchscreen music app.\nA preliminary evaluation analyses the system in terms of training, musical\ngeneration and user interaction.\n",
"title": "RoboJam: A Musical Mixture Density Network for Collaborative Touchscreen Interaction"
} | null | null | null | null | true | null | 2838 | null | Default | null | null |
null | {
"abstract": " We report measurements of the de Haas-van Alphen effect in the layered\nheavy-fermion compound CePt$_2$In$_7$ in high magnetic fields up to 35 T. Above\nan angle-dependent threshold field, we observed several de Haas-van Alphen\nfrequencies originating from almost ideally two-dimensional Fermi surfaces. The\nfrequencies are similar to those previously observed to develop only above a\nmuch higher field of 45 T, where a clear anomaly was detected and proposed to\noriginate from a change in the electronic structure [M. M. Altarawneh et al.,\nPhys. Rev. B 83, 081103 (2011)]. Our experimental results are compared with\nband structure calculations performed for both CePt$_2$In$_7$ and\nLaPt$_2$In$_7$, and the comparison suggests localized $f$ electrons in\nCePt$_2$In$_7$. This conclusion is further supported by comparing\nexperimentally observed Fermi surfaces in CePt$_2$In$_7$ and PrPt$_2$In$_7$,\nwhich are found to be almost identical. The measured effective masses in\nCePt$_2$In$_7$ are only moderately enhanced above the bare electron mass $m_0$,\nfrom 2$m_0$ to 6$m_0$.\n",
"title": "Quasi-two-dimensional Fermi surfaces with localized $f$ electrons in the layered heavy-fermion compound CePt$_2$In$_7$"
} | null | null | null | null | true | null | 2839 | null | Default | null | null |
null | {
"abstract": " We associate an Albert form to any pair of cyclic algebras of prime degree\n$p$ over a field $F$ with $\\operatorname{char}(F)=p$ which coincides with the\nclassical Albert form when $p=2$. We prove that if every Albert form is\nisotropic then $H^4(F)=0$. As a result, we obtain that if $F$ is a linked field\nwith $\\operatorname{char}(F)=2$ then its $u$-invariant is either $0,2,4$ or\n$8$.\n",
"title": "Differential Forms, Linked Fields and the $u$-Invariant"
} | null | null | null | null | true | null | 2840 | null | Default | null | null |
null | {
"abstract": " A nonlinear cyclic system with delay and the overall negative feedback is\nconsidered. The characteristic equation of the linearized system is studied in\ndetail. Sufficient conditions for the oscillation of all solutions and for the\nexistence of monotone solutions are derived in terms of roots of the\ncharacteristic equation.\n",
"title": "A cyclic system with delay and its characteristic equation"
} | null | null | null | null | true | null | 2841 | null | Default | null | null |
null | {
"abstract": " Automatic welding of tubular TKY joints is an important and challenging task\nfor the marine and offshore industry. In this paper, a framework for tubular\njoint detection and motion planning is proposed. The pose of the real tubular\njoint is detected using RGB-D sensors, which is used to obtain a\nreal-to-virtual mapping for positioning the workpiece in a virtual environment.\nFor motion planning, a Bi-directional Transition based Rapidly exploring Random\nTree (BiTRRT) algorithm is used to generate trajectories for reaching the\ndesired goals. The complete framework is verified with experiments, and the\nresults show that the robot welding torch is able to transit without collision\nto desired goals which are close to the tubular joint.\n",
"title": "Object Detection and Motion Planning for Automated Welding of Tubular Joints"
} | null | null | [
"Computer Science"
]
| null | true | null | 2842 | null | Validated | null | null |
null | {
"abstract": " We prove that the killing rate of certain degree-lowering \"recursion\noperators\" on a polynomial algebra over a finite field grows slower than\nlinearly in the degree of the polynomial attacked. We also explain the\nmotivating application: obtaining a lower bound for the Krull dimension of a\nlocal component of a big mod-p Hecke algebra in the genus-zero case. We sketch\nthe application for p=2 and p=3 in level one. The case p=2 was first\nestablished in by Nicolas and Serre in 2012 using different methods.\n",
"title": "Nilpotence order growth of recursion operators in characteristic p"
} | null | null | null | null | true | null | 2843 | null | Default | null | null |
null | {
"abstract": " Neural models have become ubiquitous in automatic speech recognition systems.\nWhile neural networks are typically used as acoustic models in more complex\nsystems, recent studies have explored end-to-end speech recognition systems\nbased on neural networks, which can be trained to directly predict text from\ninput acoustic features. Although such systems are conceptually elegant and\nsimpler than traditional systems, it is less obvious how to interpret the\ntrained models. In this work, we analyze the speech representations learned by\na deep end-to-end model that is based on convolutional and recurrent layers,\nand trained with a connectionist temporal classification (CTC) loss. We use a\npre-trained model to generate frame-level features which are given to a\nclassifier that is trained on frame classification into phones. We evaluate\nrepresentations from different layers of the deep model and compare their\nquality for predicting phone labels. Our experiments shed light on important\naspects of the end-to-end model such as layer depth, model complexity, and\nother design choices.\n",
"title": "Analyzing Hidden Representations in End-to-End Automatic Speech Recognition Systems"
} | null | null | null | null | true | null | 2844 | null | Default | null | null |
null | {
"abstract": " Diffusion MRI (dMRI) is a valuable tool in the assessment of tissue\nmicrostructure. By fitting a model to the dMRI signal it is possible to derive\nvarious quantitative features. Several of the most popular dMRI signal models\nare expansions in an appropriately chosen basis, where the coefficients are\ndetermined using some variation of least-squares. However, such approaches lack\nany notion of uncertainty, which could be valuable in e.g. group analyses. In\nthis work, we use a probabilistic interpretation of linear least-squares\nmethods to recast popular dMRI models as Bayesian ones. This makes it possible\nto quantify the uncertainty of any derived quantity. In particular, for\nquantities that are affine functions of the coefficients, the posterior\ndistribution can be expressed in closed-form. We simulated measurements from\nsingle- and double-tensor models where the correct values of several quantities\nare known, to validate that the theoretically derived quantiles agree with\nthose observed empirically. We included results from residual bootstrap for\ncomparison and found good agreement. The validation employed several different\nmodels: Diffusion Tensor Imaging (DTI), Mean Apparent Propagator MRI (MAP-MRI)\nand Constrained Spherical Deconvolution (CSD). We also used in vivo data to\nvisualize maps of quantitative features and corresponding uncertainties, and to\nshow how our approach can be used in a group analysis to downweight subjects\nwith high uncertainty. In summary, we convert successful linear models for dMRI\nsignal estimation to probabilistic models, capable of accurate uncertainty\nquantification.\n",
"title": "Bayesian uncertainty quantification in linear models for diffusion MRI"
} | null | null | null | null | true | null | 2845 | null | Default | null | null |
null | {
"abstract": " Empirical risk minimization (ERM) is ubiquitous in machine learning and\nunderlies most supervised learning methods. While there has been a large body\nof work on algorithms for various ERM problems, the exact computational\ncomplexity of ERM is still not understood. We address this issue for multiple\npopular ERM problems including kernel SVMs, kernel ridge regression, and\ntraining the final layer of a neural network. In particular, we give\nconditional hardness results for these problems based on complexity-theoretic\nassumptions such as the Strong Exponential Time Hypothesis. Under these\nassumptions, we show that there are no algorithms that solve the aforementioned\nERM problems to high accuracy in sub-quadratic time. We also give similar\nhardness results for computing the gradient of the empirical loss, which is the\nmain computational burden in many non-convex learning tasks.\n",
"title": "On the Fine-Grained Complexity of Empirical Risk Minimization: Kernel Methods and Neural Networks"
} | null | null | null | null | true | null | 2846 | null | Default | null | null |
null | {
"abstract": " Deep learning searches for nonlinear factors for predicting asset returns.\nPredictability is achieved via multiple layers of composite factors as opposed\nto additive ones. Viewed in this way, asset pricing studies can be revisited\nusing multi-layer deep learners, such as rectified linear units (ReLU) or\nlong-short-term-memory (LSTM) for time-series effects. State-of-the-art\nalgorithms including stochastic gradient descent (SGD), TensorFlow and dropout\ndesign provide imple- mentation and efficient factor exploration. To illustrate\nour methodology, we revisit the equity market risk premium dataset of Welch and\nGoyal (2008). We find the existence of nonlinear factors which explain\npredictability of returns, in particular at the extremes of the characteristic\nspace. Finally, we conclude with directions for future research.\n",
"title": "Deep Learning for Predicting Asset Returns"
} | null | null | null | null | true | null | 2847 | null | Default | null | null |
null | {
"abstract": " ADMM is a popular algorithm for solving convex optimization problems.\nApplying this algorithm to distributed consensus optimization problem results\nin a fully distributed iterative solution which relies on processing at the\nnodes and communication between neighbors. Local computations usually suffer\nfrom different types of errors, due to e.g., observation or quantization noise,\nwhich can degrade the performance of the algorithm. In this work, we focus on\nanalyzing the convergence behavior of distributed ADMM for consensus\noptimization in presence of additive node error. We specifically show that (a\nnoisy) ADMM converges linearly under certain conditions and also examine the\nassociated convergence point. Numerical results are provided which demonstrate\nthe effectiveness of the presented analysis.\n",
"title": "Analysis of Distributed ADMM Algorithm for Consensus Optimization in Presence of Error"
} | null | null | [
"Computer Science",
"Mathematics"
]
| null | true | null | 2848 | null | Validated | null | null |
null | {
"abstract": " We prove finite jet determination for (finitely) smooth CR diffeomorphisms of\n(finitely) smooth Levi degenerate hypersurfaces in $\\mathbb{C}^{n+1}$ by\nconstructing generalized stationary discs glued to such hypersurfaces.\n",
"title": "Jet determination of smooth CR automorphisms and generalized stationary discs"
} | null | null | null | null | true | null | 2849 | null | Default | null | null |
null | {
"abstract": " We present a mathematical analysis of a non-convex energy landscape for\nrobust subspace recovery. We prove that an underlying subspace is the only\nstationary point and local minimizer in a specified neighborhood under\ndeterministic conditions on a dataset. If the deterministic condition is\nsatisfied, we further show that a geodesic gradient descent method over the\nGrassmannian manifold can exactly recover the underlying subspace when the\nmethod is properly initialized. Proper initialization by principal component\nanalysis is guaranteed with a similar deterministic condition. Under slightly\nstronger assumptions, the gradient descent method with a special shrinking step\nsize scheme achieves linear convergence. The practicality of the deterministic\ncondition is demonstrated on some statistical models of data, and the method\nachieves almost state-of-the-art recovery guarantees on the Haystack Model for\ndifferent regimes of sample size and ambient dimension. In particular, when the\nambient dimension is fixed and the sample size is large enough, we show that\nour gradient method can exactly recover the underlying subspace for any fixed\nfraction of outliers (less than 1).\n",
"title": "A Well-Tempered Landscape for Non-convex Robust Subspace Recovery"
} | null | null | null | null | true | null | 2850 | null | Default | null | null |
null | {
"abstract": " Inspired by biophysical principles underlying nonlinear dendritic computation\nin neural circuits, we develop a scheme to train deep neural networks to make\nthem robust to adversarial attacks. Our scheme generates highly nonlinear,\nsaturated neural networks that achieve state of the art performance on gradient\nbased adversarial examples on MNIST, despite never being exposed to\nadversarially chosen examples during training. Moreover, these networks exhibit\nunprecedented robustness to targeted, iterative schemes for generating\nadversarial examples, including second-order methods. We further identify\nprinciples governing how these networks achieve their robustness, drawing on\nmethods from information geometry. We find these networks progressively create\nhighly flat and compressed internal representations that are sensitive to very\nfew input dimensions, while still solving the task. Moreover, they employ\nhighly kurtotic weight distributions, also found in the brain, and we\ndemonstrate how such kurtosis can protect even linear classifiers from\nadversarial attack.\n",
"title": "Biologically inspired protection of deep networks from adversarial attacks"
} | null | null | [
"Computer Science",
"Statistics"
]
| null | true | null | 2851 | null | Validated | null | null |
null | {
"abstract": " The entropy of a random variable is well-known to equal the exponential\ngrowth rate of the volumes of its typical sets. In this paper, we show that for\nany log-concave random variable $X$, the sequence of the $\\lfloor n\\theta\n\\rfloor^{\\text{th}}$ intrinsic volumes of the typical sets of $X$ in dimensions\n$n \\geq 1$ grows exponentially with a well-defined rate. We denote this rate by\n$h_X(\\theta)$, and call it the $\\theta^{\\text{th}}$ intrinsic entropy of $X$.\nWe show that $h_X(\\theta)$ is a continuous function of $\\theta$ over the range\n$[0,1]$, thereby providing a smooth interpolation between the values 0 and\n$h(X)$ at the endpoints 0 and 1, respectively.\n",
"title": "Intrinsic entropies of log-concave distributions"
} | null | null | null | null | true | null | 2852 | null | Default | null | null |
null | {
"abstract": " In this paper, we consider solving a class of nonconvex and nonsmooth\nproblems frequently appearing in signal processing and machine learning\nresearch. The traditional alternating direction method of multipliers\nencounters troubles in both mathematics and computations in solving the\nnonconvex and nonsmooth subproblem. In view of this, we propose a reweighted\nalternating direction method of multipliers. In this algorithm, all subproblems\nare convex and easy to solve. We also provide several guarantees for the\nconvergence and prove that the algorithm globally converges to a critical point\nof an auxiliary function with the help of the Kurdyka-{\\L}ojasiewicz property.\nSeveral numerical results are presented to demonstrate the efficiency of the\nproposed algorithm.\n",
"title": "Iteratively Linearized Reweighted Alternating Direction Method of Multipliers for a Class of Nonconvex Problems"
} | null | null | null | null | true | null | 2853 | null | Default | null | null |
null | {
"abstract": " In this paper, we present a novel deep fusion architecture for audio\nclassification tasks. The multi-channel model presented is formed using deep\nconvolution layers where different acoustic features are passed through each\nchannel. To enable dissemination of information across the channels, we\nintroduce attention feature maps that aid in the alignment of frames. The\noutput of each channel is merged using interaction parameters that non-linearly\naggregate the representative features. Finally, we evaluate the performance of\nthe proposed architecture on three benchmark datasets:- DCASE-2016 and LITIS\nRouen (acoustic scene recognition), and CHiME-Home (tagging). Our experimental\nresults suggest that the architecture presented outperforms the standard\nbaselines and achieves outstanding performance on the task of acoustic scene\nrecognition and audio tagging.\n",
"title": "Acoustic Features Fusion using Attentive Multi-channel Deep Architecture"
} | null | null | null | null | true | null | 2854 | null | Default | null | null |
null | {
"abstract": " Recently, two-dimensional canonical correlation analysis (2DCCA) has been\nsuccessfully applied for image feature extraction. The method instead of\nconcatenating the columns of the images to the one-dimensional vectors,\ndirectly works with two-dimensional image matrices. Although 2DCCA works well\nin different recognition tasks, it lacks a probabilistic interpretation. In\nthis paper, we present a probabilistic framework for 2DCCA called probabilistic\n2DCCA (P2DCCA) and an iterative EM based algorithm for optimizing the\nparameters. Experimental results on synthetic and real data demonstrate\nsuperior performance in loading factor estimation for P2DCCA compared to 2DCCA.\nFor real data, three subsets of AR face database and also the UMIST face\ndatabase confirm the robustness of the proposed algorithm in face recognition\ntasks with different illumination conditions, facial expressions, poses and\nocclusions.\n",
"title": "An EM Based Probabilistic Two-Dimensional CCA with Application to Face Recognition"
} | null | null | null | null | true | null | 2855 | null | Default | null | null |
null | {
"abstract": " The wide adoption of smartphones and mobile applications has brought\nsignificant changes to not only how individuals behave in the real world, but\nalso how groups of users interact with each other when organizing group events.\nUnderstanding how users make event decisions as a group and identifying the\ncontributing factors can offer important insights for social group studies and\nmore effective system and application design for group event scheduling.\nIn this work, we have designed a new mobile application called\nOutWithFriendz, which enables users of our mobile app to organize group events,\ninvite friends, suggest and vote on event time and venue. We have deployed\nOutWithFriendz at both Apple App Store and Google Play, and conducted a\nlarge-scale user study spanning over 500 users and 300 group events. Our\nanalysis has revealed several important observations regarding group event\nplanning process including the importance of user mobility, individual\npreferences, host preferences, and group voting process.\n",
"title": "Understanding Group Event Scheduling via the OutWithFriendz Mobile Application"
} | null | null | [
"Computer Science"
]
| null | true | null | 2856 | null | Validated | null | null |
null | {
"abstract": " Hybridized molecule/metal interfaces are ubiquitous in molecular and organic\ndevices. The energy level alignment (ELA) of frontier molecular levels relative\nto the metal Fermi level (EF) is critical to the conductance and functionality\nof these devices. However, a clear understanding of the ELA that includes\nmany-electron self-energy effects is lacking. Here, we investigate the\nmany-electron effects on the ELA using state-of-the-art, benchmark GW\ncalculations on prototypical chemisorbed molecules on Au(111), in eleven\ndifferent geometries. The GW ELA is in good agreement with photoemission for\nmonolayers of benzene-diamine on Au(111). We find that in addition to static\nimage charge screening, the frontier levels in most of these geometries are\nrenormalized by additional screening from substrate-mediated intermolecular\nCoulomb interactions. For weakly chemisorbed systems, such as amines and\npyridines on Au, this additional level renormalization (~1.5 eV) comes solely\nfrom static screened exchange energy, allowing us to suggest computationally\nmore tractable schemes to predict the ELA at such interfaces. However, for more\nstrongly chemisorbed thiolate layers, dynamical effects are present. Our ab\ninitio results constitute an important step towards the understanding and\nmanipulation of functional molecular/organic systems for both fundamental\nstudies and applications.\n",
"title": "Energy Level Alignment at Hybridized Organic-metal Interfaces: the Role of Many-electron Effects"
} | null | null | null | null | true | null | 2857 | null | Default | null | null |
null | {
"abstract": " Detection of the mostly geomagnetically generated radio emission of\ncosmic-ray air showers provides an alternative to air-Cherenkov and\nair-fluorescence detection, since it is not limited to clear nights. Like these\nestablished methods, the radio signal is sensitive to the calorimetric energy\nand the position of the maximum of the electromagnetic shower component. This\nmakes antenna arrays an ideal extension for particle-detector arrays above a\nthreshold energy of about 100 PeV of the primary cosmic-ray particles. In the\nlast few years the digital radio technique for cosmic-ray air showers again\nmade significant progress, and there now is a consistent picture of the\nemission mechanisms confirmed by several measurements. Recent results by the\nantenna arrays AERA and Tunka-Rex confirm that the absolute accuracy for the\nshower energy is as good as the other detection techniques. Moreover, the\nsensitivity to the shower maximum of the radio signal has been confirmed in\ndirect comparison to air-Cherenkov measurements by Tunka-Rex. The dense antenna\narray LOFAR can already compete with the established techniques in accuracy for\ncosmic-ray mass-composition. In the future, a new generation of radio\nexperiments might drive the field: either by providing extremely large exposure\nfor inclined cosmic-ray or neutrino showers or, like the SKA core in Australia\nwith its several 10,000 antennas, by providing extremely detailed measurements.\n",
"title": "Radio detection of Extensive Air Showers (ECRS 2016)"
} | null | null | null | null | true | null | 2858 | null | Default | null | null |
null | {
"abstract": " We study how the regret guarantees of nonstochastic multi-armed bandits can\nbe improved, if the effective range of the losses in each round is small (e.g.\nthe maximal difference between two losses in a given round). Despite a recent\nimpossibility result, we show how this can be made possible under certain mild\nadditional assumptions, such as availability of rough estimates of the losses,\nor advance knowledge of the loss of a single, possibly unspecified arm. Along\nthe way, we develop a novel technique which might be of independent interest,\nto convert any multi-armed bandit algorithm with regret depending on the loss\nrange, to an algorithm with regret depending only on the effective range, while\navoiding predictably bad arms altogether.\n",
"title": "Bandit Regret Scaling with the Effective Loss Range"
} | null | null | [
"Computer Science",
"Statistics"
]
| null | true | null | 2859 | null | Validated | null | null |
null | {
"abstract": " The paper presents a novel concept that analyzes and visualizes worldwide\nfashion trends. Our goal is to reveal cutting-edge fashion trends without\ndisplaying an ordinary fashion style. To achieve the fashion-based analysis, we\ncreated a new fashion culture database (FCDB), which consists of 76 million\ngeo-tagged images in 16 cosmopolitan cities. By grasping a fashion trend of\nmixed fashion styles,the paper also proposes an unsupervised fashion trend\ndescriptor (FTD) using a fashion descriptor, a codeword vetor, and temporal\nanalysis. To unveil fashion trends in the FCDB, the temporal analysis in FTD\neffectively emphasizes consecutive features between two different times. In\nexperiments, we clearly show the analysis of fashion trends and fashion-based\ncity similarity. As the result of large-scale data collection and an\nunsupervised analyzer, the proposed approach achieves world-level fashion\nvisualization in a time series. The code, model, and FCDB will be publicly\navailable after the construction of the project page.\n",
"title": "Changing Fashion Cultures"
} | null | null | null | null | true | null | 2860 | null | Default | null | null |
null | {
"abstract": " We study classes of atomic models At_T of a countable, complete first-order\ntheory T . We prove that if At_T is not pcl-small, i.e., there is an atomic\nmodel N that realizes uncountably many types over pcl(a) for some finite tuple\na from N, then there are 2^aleph1 non-isomorphic atomic models of T, each of\nsize aleph1.\n",
"title": "A strong failure of aleph_0-stability for atomic classes"
} | null | null | null | null | true | null | 2861 | null | Default | null | null |
null | {
"abstract": " We study the problem of estimating the mean of a random vector $X$ given a\nsample of $N$ independent, identically distributed points. We introduce a new\nestimator that achieves a purely sub-Gaussian performance under the only\ncondition that the second moment of $X$ exists. The estimator is based on a\nnovel concept of a multivariate median.\n",
"title": "Sub-Gaussian estimators of the mean of a random vector"
} | null | null | [
"Mathematics",
"Statistics"
]
| null | true | null | 2862 | null | Validated | null | null |
null | {
"abstract": " We study the problem of containing epidemic spreading processes in temporal\nnetworks. We specifically focus on the problem of finding a resource allocation\nto suppress epidemic infection, provided that an empirical time-series data of\nconnectivities between nodes is available. Although this problem is of\npractical relevance, it has not been clear how an empirical time-series data\ncan inform our strategy of resource allocations, due to the computational\ncomplexity of the problem. In this direction, we present a computationally\nefficient framework for finding a resource allocation that satisfies a given\nbudget constraint and achieves a given control performance. The framework is\nbased on convex programming and, moreover, allows the performance measure to be\ndescribed by a wide class of functionals called posynomials with nonnegative\nexponents. We illustrate our theoretical results using a data of temporal\ninteraction networks within a primary school.\n",
"title": "Resource Allocation for Containing Epidemics from Temporal Network Data"
} | null | null | null | null | true | null | 2863 | null | Default | null | null |
null | {
"abstract": " In this paper, we investigate the possibility of applying plan\ntransformations to general manipulation plans in order to specialize them to\nthe specific situation at hand. We present a framework for optimizing execution\nand achieving higher performance by autonomously transforming robot's behavior\nat runtime. We show that plans employed by robotic agents in real-world\nenvironments can be transformed, despite their control structures being very\ncomplex due to the specifics of acting in the real world. The evaluation is\ncarried out on a plan of a PR2 robot performing pick and place tasks, to which\nwe apply three example transformations, as well as on a large amount of\nexperiments in a fast plan projection environment.\n",
"title": "Towards Plan Transformations for Real-World Pick and Place Tasks"
} | null | null | [
"Computer Science"
]
| null | true | null | 2864 | null | Validated | null | null |
null | {
"abstract": " In computer vision applications, such as domain adaptation (DA), few shot\nlearning (FSL) and zero-shot learning (ZSL), we encounter new objects and\nenvironments, for which insufficient examples exist to allow for training\n\"models from scratch,\" and methods that adapt existing models, trained on the\npresented training environment, to the new scenario are required. We propose a\nnovel visual attribute encoding method that encodes each image as a\nlow-dimensional probability vector composed of prototypical part-type\nprobabilities. The prototypes are learnt to be representative of all training\ndata. At test-time we utilize this encoding as an input to a classifier. At\ntest-time we freeze the encoder and only learn/adapt the classifier component\nto limited annotated labels in FSL; new semantic attributes in ZSL. We conduct\nextensive experiments on benchmark datasets. Our method outperforms\nstate-of-art methods trained for the specific contexts (ZSL, FSL, DA).\n",
"title": "Learning for New Visual Environments with Limited Labels"
} | null | null | null | null | true | null | 2865 | null | Default | null | null |
null | {
"abstract": " Among mobile cloud applications, mobile cloud gaming has gained a significant\npopularity in the recent years. In mobile cloud games, textures, game objects,\nand game events are typically streamed from a server to the mobile client.\nOne of the challenges in cloud mobile gaming is how to efficiently multicast\ngaming contents and updates in Massively Multi-player Online Games (MMOGs).\nThis report surveys the state of art techniques introduced for game\nsynchronization and multicasting mechanisms to decrease latency and bandwidth\nconsumption, and discuss several schemes that have been proposed in this area\nthat can be applied to any networked gaming context. From our point of view,\ngaming applications demand high interactivity. Therefore, concentrating on\ngaming applications will eventually cover a wide range of applications without\nviolating the limited scope of this survey.\n",
"title": "A Survey of Bandwidth and Latency Enhancement Approaches for Mobile Cloud Game Multicasting"
} | null | null | null | null | true | null | 2866 | null | Default | null | null |
null | {
"abstract": " Cosmic ray intensities (CRIs) recorded by sixteen neutron monitors have been\nused to study its dependence on the tilt angles (TA) of the heliospheric\ncurrent sheet (HCS) during period 1976-2014, which covers three solar activity\ncycles 21, 22 and 23. The median primary rigidity covers the range 16-33 GV.\nOur results have indicated that the CRIs are directly sensitive to, and\norganized by, the interplanetary magnetic field (IMF) and its neutral sheet\ninclinations. The observed differences in the sensitivity of cosmic ray\nintensity to changes in the neutral sheet tilt angles before and after the\nreversal of interplanetary magnetic field polarity have been studied. Much\nstronger intensity-tilt angle correlation was found when the solar magnetic\nfield in the North Polar Region was directed inward than it was outward. The\nrigidity dependence of sensitivities of cosmic rays differs according to the\nIMF polarity, for the periods 1981-1988 and 2001-2008 (qA < 0) it was R-1.00\nand R-1.48 respectively, while for the 1991-1998 epoch (qA > 0) it was R-1.35.\nHysteresis loops between TA and CRIs have been examined during three solar\nactivity cycles 21, 22 and 23. A consider differences in time lags during qA >\n0 and qA < 0 polarity states of the heliosphere have been observed. We also\nfound that the cosmic ray intensity decreases at much faster rate with increase\nof tilt angle during qA < 0 than qA > 0, indicating stronger response to the\ntilt angle changes during qA < 0. Our results are discussed in the light of 3D\nmodulation models including the gradient, curvature drifts and the tilt of the\nheliospheric current sheet.\n",
"title": "Modulation of High-Energy Particles and the Heliospheric Current Sheet Tilts throughout 1976-2014"
} | null | null | null | null | true | null | 2867 | null | Default | null | null |
null | {
"abstract": " In many developing countries, public transit plays an important role in daily\nlife. However, few existing methods have considered the influence of public\ntransit in their models. In this work, we present a dual-perspective view of\nthe epidemic spreading process of the individual that involves both\ncontamination in places (such as work places and homes) and public transit\n(such as buses and trains). In more detail, we consider a group of individuals\nwho travel to some places using public transit, and introduce public transit\ninto the epidemic spreading process. A novel modeling framework is proposed\nconsidering place-based infections and the public-transit-based infections. In\nthe urban scenario, we investigate the public transit trip contribution rate\n(PTTCR) in the epidemic spreading process of the individual, and assess the\nimpact of the public transit trip contribution rate by evaluating the volume of\ninfectious people. Scenarios for strategies such as public transit and school\nclosure were tested and analyzed. Our simulation results suggest that\nindividuals with a high public transit trip contribution rate will increase the\nvolume of infectious people when an infectious disease outbreak occurs by\naffecting the social network through the public transit trip contribution rate.\n",
"title": "Detecting the impact of public transit on the transmission of epidemics"
} | null | null | null | null | true | null | 2868 | null | Default | null | null |
null | {
"abstract": " We consider a class of magnetic fields defined over the interior of a\nmanifold $M$ which go to infinity at its boundary and whose direction near the\nboundary of $M$ is controlled by a closed 1-form $\\sigma_\\infty \\in\n\\Gamma(T^*\\partial M)$. We are able to show that charged particles in the\ninterior of $M$ under the influence of such fields can only escape the manifold\nthrough the zero locus of $\\sigma_\\infty$. In particular in the case where the\n1-form is nowhere vanishing we conclude that the particles become confined to\nits interior for all time.\n",
"title": "The Hamiltonian Dynamics of Magnetic Confinement in Toroidal Domains"
} | null | null | null | null | true | null | 2869 | null | Default | null | null |
null | {
"abstract": " Some lung diseases are related to bronchial airway structures and morphology.\nAlthough airway segmentation from chest CT volumes is an important task in the\ncomputer-aided diagnosis and surgery assistance systems for the chest, complete\n3-D airway structure segmentation is a quite challenging task due to its\ncomplex tree-like structure. In this paper, we propose a new airway\nsegmentation method from 3D chest CT volumes based on volume of interests (VOI)\nusing gradient vector flow (GVF). This method segments the bronchial regions by\napplying the cavity enhancement filter (CEF) to trace the bronchial tree\nstructure from the trachea. It uses the CEF in the VOI to segment each branch.\nAnd a tube-likeness function based on GVF and the GVF magnitude map in each VOI\nare utilized to assist predicting the positions and directions of child\nbranches. By calculating the tube-likeness function based on GVF and the GVF\nmagnitude map, the airway-like candidate structures are identified and their\ncentrelines are extracted. Based on the extracted centrelines, we can detect\nthe branch points of the bifurcations and directions of the airway branches in\nthe next level. At the same time, a leakage detection is performed to avoid the\nleakage by analysing the pixel information and the shape information of airway\ncandidate regions extracted in the VOI. Finally, we unify all of the extracted\nbronchial regions to form an integrated airway tree. Preliminary experiments\nusing four cases of chest CT volumes demonstrated that the proposed method can\nextract more bronchial branches in comparison with other methods.\n",
"title": "Airway segmentation from 3D chest CT volumes based on volume of interest using gradient vector flow"
} | null | null | null | null | true | null | 2870 | null | Default | null | null |
null | {
"abstract": " In this paper, we present the design and implementation of a robust motion\nformation distributed control algorithm for a team of mobile robots. The\nprimary task for the team is to form a geometric shape, which can be freely\ntranslated and rotated at the same time. This approach makes the robots to\nbehave as a cohesive whole, which can be useful in tasks such as collaborative\ntransportation. The robustness of the algorithm relies on the fact that each\nrobot employs only local measurements from a laser sensor which does not need\nto be off-line calibrated. Furthermore, robots do not need to exchange any\ninformation with each other. Being free of sensor calibration and not requiring\na communication channel helps the scaling of the overall system to a large\nnumber of robots. In addition, since the robots do not need any off-board\nlocalization system, but require only relative positions with respect to their\nneighbors, it can be aimed to have a full autonomous team that operates in\nenvironments where such localization systems are not available. The\ncomputational cost of the algorithm is inexpensive and the resources from a\nstandard microcontroller will suffice. This fact makes the usage of our\napproach appealing as a support for other more demanding algorithms, e.g.,\nprocessing images from onboard cameras. We validate the performance of the\nalgorithm with a team of four mobile robots equipped with low-cost commercially\navailable laser scanners.\n",
"title": "Multi-robot motion-formation distributed control with sensor self-calibration: experimental validation"
} | null | null | null | null | true | null | 2871 | null | Default | null | null |
null | {
"abstract": " In recent years, a number of prominent computer scientists, along with\nacademics in fields such as philosophy and physics, have lent credence to the\nnotion that machines may one day become as large as humans. Many have further\nargued that machines could even come to exceed human size by a significant\nmargin. However, there are at least seven distinct arguments that preclude this\noutcome. We show that it is not only implausible that machines will ever exceed\nhuman size, but in fact impossible.\n",
"title": "On the Impossibility of Supersized Machines"
} | null | null | null | null | true | null | 2872 | null | Default | null | null |
null | {
"abstract": " There are a number of examples of variations of Hodge structure of maximum\ndimension. However, to our knowledge, those that are global on the level of the\nperiod domain are totally geodesic subspaces that arise from an orbit of a\nsubgroup of the group of the period domain. That is, they are defined by Lie\ntheory rather than by algebraic geometry. In this note, we give an example of a\nvariation of maximum dimension which is nowhere tangent to a geodesic\nvariation. The period domain in question, which classifies weight two Hodge\nstructures with $h^{2,0} = 2$ and $h^{1,1} = 28$, is of dimension $57$. The\nhorizontal tangent bundle has codimension one, thus it is an example of a\nholomorphic contact structure, with local integral manifolds of dimension 28.\nThe group of the period domain is $SO(4,28)$, and one can produce global\nintegral manifolds as orbits of the action of subgroups isomorphic to\n$SU(2,14)$. Our example is given by the variation of Hodge structure on the\nsecond cohomology of weighted projective hypersurfaces of degree $10$ in a\nweighted projective three-space with weights $1, 1, 2, 5$\n",
"title": "Non-geodesic variations of Hodge structure of maximum dimension"
} | null | null | null | null | true | null | 2873 | null | Default | null | null |
null | {
"abstract": " Recent results in coupled or temporal graphical models offer schemes for\nestimating the relationship structure between features when the data come from\nrelated (but distinct) longitudinal sources. A novel application of these ideas\nis for analyzing group-level differences, i.e., in identifying if trends of\nestimated objects (e.g., covariance or precision matrices) are different across\ndisparate conditions (e.g., gender or disease). Often, poor effect sizes make\ndetecting the differential signal over the full set of features difficult: for\nexample, dependencies between only a subset of features may manifest\ndifferently across groups. In this work, we first give a parametric model for\nestimating trends in the space of SPD matrices as a function of one or more\ncovariates. We then generalize scan statistics to graph structures, to search\nover distinct subsets of features (graph partitions) whose temporal dependency\nstructure may show statistically significant group-wise differences. We\ntheoretically analyze the Family Wise Error Rate (FWER) and bounds on Type 1\nand Type 2 error. On a cohort of individuals with risk factors for Alzheimer's\ndisease (but otherwise cognitively healthy), we find scientifically interesting\ngroup differences where the default analysis, i.e., models estimated on the\nfull graph, do not survive reasonable significance thresholds.\n",
"title": "Finding Differentially Covarying Needles in a Temporally Evolving Haystack: A Scan Statistics Perspective"
} | null | null | null | null | true | null | 2874 | null | Default | null | null |
null | {
"abstract": " A distributed algorithm is described for finding a common fixed point of a\nfamily of m>1 nonlinear maps M_i : R^n -> R^n assuming that each map is a\nparacontraction and that at least one such common fixed point exists. The\ncommon fixed point is simultaneously computed by m agents assuming each agent i\nknows only M_i, the current estimates of the fixed point generated by its\nneighbors, and nothing more. Each agent recursively updates its estimate of a\nfixed point by utilizing the current estimates generated by each of its\nneighbors. Neighbor relations are characterized by a time-varying directed\ngraph N(t). It is shown under suitably general conditions on N(t), that the\nalgorithm causes all agents estimates to converge to the same common fixed\npoint of the m nonlinear maps.\n",
"title": "A Distributed Algorithm for Computing a Common Fixed Point of a Finite Family of Paracontractions"
} | null | null | null | null | true | null | 2875 | null | Default | null | null |
null | {
"abstract": " For a metric measure space, we treat the set of distributions of 1-Lipschitz\nfunctions, which is called the 1-measurement. On the 1-measurement, we have a\npartial order relation by the Lipschitz order introduced by Gromov. The aim of\nthis paper is to study the maximum and maximal elements of the 1-measurement\nwith respect to the Lipschitz order. We present a necessary condition of a\nmetric measure space for the existence of the maximum of the 1-measurement. We\nalso consider a metric measure space that has the maximum of its 1-measurement.\n",
"title": "The maximum of the 1-measurement of a metric measure space"
} | null | null | [
"Mathematics"
]
| null | true | null | 2876 | null | Validated | null | null |
null | {
"abstract": " Distributed ledger technologies rely on consensus protocols confronting\ntraders with random waiting times until the transfer of ownership is\naccomplished. This time-consuming settlement process exposes arbitrageurs to\nprice risk and imposes limits to arbitrage. We derive theoretical arbitrage\nboundaries under general assumptions and show that they increase with expected\nlatency, latency uncertainty, spot volatility, and risk aversion. Using\nhigh-frequency data from the Bitcoin network, we estimate arbitrage boundaries\ndue to settlement latency of on average 124 basis points, covering 88 percent\nof the observed cross-exchange price differences. Settlement through\ndecentralized systems thus induces non-trivial frictions affecting market\nefficiency and price formation.\n",
"title": "Limits to Arbitrage in Markets with Stochastic Settlement Latency"
} | null | null | null | null | true | null | 2877 | null | Default | null | null |
null | {
"abstract": " We study the complexity of approximations to the normalized information\ndistance. We introduce a hierarchy of computable approximations by considering\nthe number of oscillations. This is a function version of the difference\nhierarchy for sets. We show that the normalized information distance is not in\nany level of this hierarchy, strengthening previous nonapproximability results.\nAs an ingredient to the proof, we also prove a conditional undecidability\nresult about independence.\n",
"title": "Normalized Information Distance and the Oscillation Hierarchy"
} | null | null | [
"Computer Science",
"Mathematics"
]
| null | true | null | 2878 | null | Validated | null | null |
null | {
"abstract": " As training data rapid growth, large-scale parallel training with multi-GPUs\ncluster is widely applied in the neural network model learning currently.We\npresent a new approach that applies exponential moving average method in\nlarge-scale parallel training of neural network model. It is a non-interference\nstrategy that the exponential moving average model is not broadcasted to\ndistributed workers to update their local models after model synchronization in\nthe training process, and it is implemented as the final model of the training\nsystem. Fully-connected feed-forward neural networks (DNNs) and deep\nunidirectional Long short-term memory (LSTM) recurrent neural networks (RNNs)\nare successfully trained with proposed method for large vocabulary continuous\nspeech recognition on Shenma voice search data in Mandarin. The character error\nrate (CER) of Mandarin speech recognition further degrades than\nstate-of-the-art approaches of parallel training.\n",
"title": "Exponential Moving Average Model in Parallel Speech Recognition Training"
} | null | null | [
"Computer Science"
]
| null | true | null | 2879 | null | Validated | null | null |
null | {
"abstract": " Hyperparameter tuning is the black art of automatically finding a good\ncombination of control parameters for a data miner. While widely applied in\nempirical Software Engineering, there has not been much discussion on which\nhyperparameter tuner is best for software analytics. To address this gap in the\nliterature, this paper applied a range of hyperparameter optimizers (grid\nsearch, random search, differential evolution, and Bayesian optimization) to\ndefect prediction problem. Surprisingly, no hyperparameter optimizer was\nobserved to be `best' and, for one of the two evaluation measures studied here\n(F-measure), hyperparameter optimization, in 50\\% cases, was no better than\nusing default configurations.\nWe conclude that hyperparameter optimization is more nuanced than previously\nbelieved. While such optimization can certainly lead to large improvements in\nthe performance of classifiers used in software analytics, it remains to be\nseen which specific optimizers should be applied to a new dataset.\n",
"title": "Is One Hyperparameter Optimizer Enough?"
} | null | null | null | null | true | null | 2880 | null | Default | null | null |
null | {
"abstract": " We present Deep Generalized Canonical Correlation Analysis (DGCCA) -- a\nmethod for learning nonlinear transformations of arbitrarily many views of\ndata, such that the resulting transformations are maximally informative of each\nother. While methods for nonlinear two-view representation learning (Deep CCA,\n(Andrew et al., 2013)) and linear many-view representation learning\n(Generalized CCA (Horst, 1961)) exist, DGCCA is the first CCA-style multiview\nrepresentation learning technique that combines the flexibility of nonlinear\n(deep) representation learning with the statistical power of incorporating\ninformation from many independent sources, or views. We present the DGCCA\nformulation as well as an efficient stochastic optimization algorithm for\nsolving it. We learn DGCCA representations on two distinct datasets for three\ndownstream tasks: phonetic transcription from acoustic and articulatory\nmeasurements, and recommending hashtags and friends on a dataset of Twitter\nusers. We find that DGCCA representations soundly beat existing methods at\nphonetic transcription and hashtag recommendation, and in general perform no\nworse than standard linear many-view techniques.\n",
"title": "Deep Generalized Canonical Correlation Analysis"
} | null | null | null | null | true | null | 2881 | null | Default | null | null |
null | {
"abstract": " A main question in graphical models and causal inference is whether, given a\nprobability distribution $P$ (which is usually an underlying distribution of\ndata), there is a graph (or graphs) to which $P$ is faithful. The main goal of\nthis paper is to provide a theoretical answer to this problem. We work with\ngeneral independence models, which contain probabilistic independence models as\na special case. We exploit a generalization of ordering, called preordering, of\nthe nodes of (mixed) graphs. This allows us to provide sufficient conditions\nfor a given independence model to be Markov to a graph with the minimum\npossible number of edges, and more importantly, necessary and sufficient\nconditions for a given probability distribution to be faithful to a graph. We\npresent our results for the general case of mixed graphs, but specialize the\ndefinitions and results to the better-known subclasses of undirected\n(concentration) and bidirected (covariance) graphs as well as directed acyclic\ngraphs.\n",
"title": "Faithfulness of Probability Distributions and Graphs"
} | null | null | null | null | true | null | 2882 | null | Default | null | null |
null | {
"abstract": " We give a lower bound for the multipliers of repelling periodic points of\nentire functions. The bound is deduced from a bound for the multipliers of\nfixed points of composite entire functions.\n",
"title": "On the multipliers of repelling periodic points of entire functions"
} | null | null | null | null | true | null | 2883 | null | Default | null | null |
null | {
"abstract": " We show that, within a linear approximation of BCS theory, a weak homogeneous\nmagnetic field lowers the critical temperature by an explicit constant times\nthe field strength, up to higher order terms. This provides a rigorous\nderivation and generalization of results obtained in the physics literature\nfrom WHH theory of the upper critical magnetic field. A new ingredient in our\nproof is a rigorous phase approximation to control the effects of the magnetic\nfield.\n",
"title": "The BCS critical temperature in a weak homogeneous magnetic field"
} | null | null | null | null | true | null | 2884 | null | Default | null | null |
null | {
"abstract": " Predicting personality is essential for social applications supporting\nhuman-centered activities, yet prior modeling methods with users written text\nrequire too much input data to be realistically used in the context of social\nmedia. In this work, we aim to drastically reduce the data requirement for\npersonality modeling and develop a model that is applicable to most users on\nTwitter. Our model integrates Word Embedding features with Gaussian Processes\nregression. Based on the evaluation of over 1.3K users on Twitter, we find that\nour model achieves comparable or better accuracy than state of the art\ntechniques with 8 times fewer data.\n",
"title": "25 Tweets to Know You: A New Model to Predict Personality with Social Media"
} | null | null | [
"Computer Science"
]
| null | true | null | 2885 | null | Validated | null | null |
null | {
"abstract": " We propose a novel couple mappings method for low resolution face recognition\nusing deep convolutional neural networks (DCNNs). The proposed architecture\nconsists of two branches of DCNNs to map the high and low resolution face\nimages into a common space with nonlinear transformations. The branch\ncorresponding to transformation of high resolution images consists of 14 layers\nand the other branch which maps the low resolution face images to the common\nspace includes a 5-layer super-resolution network connected to a 14-layer\nnetwork. The distance between the features of corresponding high and low\nresolution images are backpropagated to train the networks. Our proposed method\nis evaluated on FERET data set and compared with state-of-the-art competing\nmethods. Our extensive experimental results show that the proposed method\nsignificantly improves the recognition performance especially for very low\nresolution probe face images (11.4% improvement in recognition accuracy).\nFurthermore, it can reconstruct a high resolution image from its corresponding\nlow resolution probe image which is comparable with state-of-the-art\nsuper-resolution methods in terms of visual quality.\n",
"title": "Low Resolution Face Recognition Using a Two-Branch Deep Convolutional Neural Network Architecture"
} | null | null | null | null | true | null | 2886 | null | Default | null | null |
null | {
"abstract": " Various sectors are likely to carry a set of emerging applications while\ntargeting a reliable communication with low latency transmission. To address\nthis issue, upon a spectrally-efficient transmission, this paper investigates\nthe performance of a one full-dulpex (FD) relay system, and considers for that\npurpose, two basic relaying schemes, namely the symbol-by-symbol transmission,\ni.e., amplify-and-forward (AF) and the block-by-block transmission, i.e.,\nselective decode-and-forward (SDF). The conducted analysis presents an\nexhaustive comparison, covering both schemes, over two different transmission\nmodes, i.e., the non combining mode where the best link, direct or relay link\nis decoded and the signals combining mode, where direct and relay links are\ncombined at the receiver side. While targeting latency purpose as a necessity,\nsimulations show a refined results of performed comparisons, and reveal that AF\nrelaying scheme is more adapted to combining mode, whereas the SDF relaying\nscheme is more suitable for non combining mode.\n",
"title": "A Comparative Study of Full-Duplex Relaying Schemes for Low Latency Applications"
} | null | null | null | null | true | null | 2887 | null | Default | null | null |
null | {
"abstract": " Let $G$ be the circulant graph $C_n(S)$ with $S\\subseteq\\{ 1,\\ldots,\\left\n\\lfloor\\frac{n}{2}\\right \\rfloor\\}$ and let $I(G)$ be its edge ideal in the\nring $K[x_0,\\ldots,x_{n-1}]$. Under the hypothesis that $n$ is prime we : 1)\ncompute the regularity index of $R/I(G)$; 2) compute the Castelnuovo-Mumford\nregularity when $R/I(G)$ is Cohen-Macaulay; 3) prove that the circulant graphs\nwith $S=\\{1,\\ldots,s\\}$ are sequentially $S_2$ . We end characterizing the\nCohen-Macaulay circulant graphs of Krull dimension $2$ and computing their\nCohen-Macaulay type and Castelnuovo-Mumford regularity.\n",
"title": "Some algebraic invariants of edge ideal of circulant graphs"
} | null | null | null | null | true | null | 2888 | null | Default | null | null |
null | {
"abstract": " Barrier options are one of the most widely traded exotic options on stock\nexchanges. In this paper, we develop a new stochastic simulation method for\npricing barrier options and estimating the corresponding execution\nprobabilities. We show that the proposed method always outperforms the standard\nMonte Carlo approach and becomes substantially more efficient when the\nunderlying asset has high volatility, while it performs better than multilevel\nMonte Carlo for special cases of barrier options and underlying assets. These\ntheoretical findings are confirmed by numerous simulation results.\n",
"title": "Efficient Pricing of Barrier Options on High Volatility Assets using Subset Simulation"
} | null | null | null | null | true | null | 2889 | null | Default | null | null |
null | {
"abstract": " Generalized-ensemble Monte Carlo simulations such as the multicanonical\nmethod and similar techniques are among the most efficient approaches for\nsimulations of systems undergoing discontinuous phase transitions or with\nrugged free- energy landscapes. As Markov chain methods, they are inherently\nserial computationally. It was demonstrated recently, however, that a\ncombination of independent simulations that communicate weight updates at\nvariable intervals allows for the efficient utilization of parallel\ncomputational resources for multicanonical simulations. Implementing this\napproach for the many-thread architecture provided by current generations of\ngraphics processing units (GPUs), we show how it can be efficiently employed\nwith of the order of $10^4$ parallel walkers and beyond, thus constituting a\nversatile tool for Monte Carlo simulations in the era of massively parallel\ncomputing. We provide the fully documented source code for the approach applied\nto the paradigmatic example of the two-dimensional Ising model as starting\npoint and reference for practitioners in the field.\n",
"title": "Massively parallel multicanonical simulations"
} | null | null | null | null | true | null | 2890 | null | Default | null | null |
null | {
"abstract": " We compared positions of the Gaia first data release (DR1) secondary data set\nat its faint limit with CCD positions of stars in 20 fields observed with the\nVLT/FORS2 camera. The FORS2 position uncertainties are smaller than one\nmilli-arcsecond (mas) and allowed us to perform an independent verification of\nthe DR1 astrometric precision. In the fields that we observed with FORS2, we\nprojected the Gaia DR1 positions into the CCD plane, performed a polynomial fit\nbetween the two sets of matching stars, and carried out statistical analyses of\nthe residuals in positions. The residual RMS roughly matches the expectations\ngiven by the Gaia DR1 uncertainties, where we identified three regimes in terms\nof Gaia DR1 precision: for G = 17-20 stars we found that the formal DR1\nposition uncertainties of stars with DR1 precisions in the range of 0.5-5 mas\nare underestimated by 63 +/- 5\\%, whereas the DR1 uncertainties of stars in the\nrange 7-10 mas are overestimated by a factor of two. For the best-measured and\ngenerally brighter G = 16-18 stars with DR1 positional uncertainties of <0.5\nmas, we detected 0.44 +/- 0.13 mas excess noise in the residual RMS, whose\norigin can be in both FORS2 and Gaia DR1. By adopting Gaia DR1 as the absolute\nreference frame we refined the pixel scale determination of FORS2, leading to\nminor updates to the parallaxes of 20 ultracool dwarfs that we published\npreviously. We also updated the FORS2 absolute parallax of the Luhman 16 binary\nbrown dwarf system to 501.42 +/- 0.11 mas\n",
"title": "Gaia and VLT astrometry of faint stars: Precision of Gaia DR1 positions and updated VLT parallaxes of ultracool dwarfs"
} | null | null | null | null | true | null | 2891 | null | Default | null | null |
null | {
"abstract": " The analysis of manifold-valued data requires efficient tools from Riemannian\ngeometry to cope with the computational complexity at stake. This complexity\narises from the always-increasing dimension of the data, and the absence of\nclosed-form expressions to basic operations such as the Riemannian logarithm.\nIn this paper, we adapt a generic numerical scheme recently introduced for\ncomputing parallel transport along geodesics in a Riemannian manifold to\nfinite-dimensional manifolds of diffeomorphisms. We provide a qualitative and\nquantitative analysis of its behavior on high-dimensional manifolds, and\ninvestigate an application with the prediction of brain structures progression.\n",
"title": "Parallel transport in shape analysis: a scalable numerical scheme"
} | null | null | null | null | true | null | 2892 | null | Default | null | null |
null | {
"abstract": " The paper presents the graph Fourier transform (GFT) of a signal in terms of\nits spectral decomposition over the Jordan subspaces of the graph adjacency\nmatrix $A$. This representation is unique and coordinate free, and it leads to\nunambiguous definition of the spectral components (\"harmonics\") of a graph\nsignal. This is particularly meaningful when $A$ has repeated eigenvalues, and\nit is very useful when $A$ is defective or not diagonalizable (as it may be the\ncase with directed graphs). Many real world large sparse graphs have defective\nadjacency matrices. We present properties of the GFT and show it to satisfy a\ngeneralized Parseval inequality and to admit a total variation ordering of the\nspectral components. We express the GFT in terms of spectral projectors and\npresent an illustrative example for a real world large urban traffic dataset.\n",
"title": "Spectral Projector-Based Graph Fourier Transforms"
} | null | null | [
"Computer Science"
]
| null | true | null | 2893 | null | Validated | null | null |
null | {
"abstract": " Dempster-Shafer evidence theory is wildly applied in multi-sensor data\nfusion. However, lots of uncertainty and interference exist in practical\nsituation, especially in the battle field. It is still an open issue to model\nthe reliability of sensor reports. Many methods are proposed based on the\nrelationship among collected data. In this letter, we proposed a quantum\nmechanical approach to evaluate the reliability of sensor reports, which is\nbased on the properties of a sensor itself. The proposed method is used to\nmodify the combining of evidences.\n",
"title": "Quantum Mechanical Approach to Modelling Reliability of Sensor Reports"
} | null | null | null | null | true | null | 2894 | null | Default | null | null |
null | {
"abstract": " Could we use Computer Vision in the Internet of Things for using pictures as\nsensors? This is the principal hypothesis that we want to resolve. Currently,\nin order to create safety areas, cities, or homes, people use IP cameras.\nNevertheless, this system needs people who watch the camera images, watch the\nrecording after something occurred, or watch when the camera notifies them of\nany movement. These are the disadvantages. Furthermore, there are many Smart\nCities and Smart Homes around the world. This is why we thought of using the\nidea of the Internet of Things to add a way of automating the use of IP\ncameras. In our case, we propose the analysis of pictures through Computer\nVision to detect people in the analysed pictures. With this analysis, we are\nable to obtain if these pictures contain people and handle the pictures as if\nthey were sensors with two possible states. Notwithstanding, Computer Vision is\na very complicated field. This is why we needed a second hypothesis: Could we\nwork with Computer Vision in the Internet of Things with a good accuracy to\nautomate or semi-automate this kind of events? The demonstration of these\nhypotheses required a testing over our Computer Vision module to check the\npossibilities that we have to use this module in a possible real environment\nwith a good accuracy. Our proposal, as a possible solution, is the analysis of\nentire sequence instead of isolated pictures for using pictures as sensors in\nthe Internet of Things.\n",
"title": "Midgar: Detection of people through computer vision in the Internet of Things scenarios to improve the security in Smart Cities, Smart Towns, and Smart Homes"
} | null | null | null | null | true | null | 2895 | null | Default | null | null |
null | {
"abstract": " Interpretability of deep neural networks is a recently emerging area of\nmachine learning research targeting a better understanding of how models\nperform feature selection and derive their classification decisions. In this\npaper, two neural network architectures are trained on spectrogram and raw\nwaveform data for audio classification tasks on a newly created audio dataset\nand layer-wise relevance propagation (LRP), a previously proposed\ninterpretability method, is applied to investigate the models' feature\nselection and decision making. It is demonstrated that the networks are highly\nreliant on feature marked as relevant by LRP through systematic manipulation of\nthe input data. Our results show that by making deep audio classifiers\ninterpretable, one can analyze and compare the properties and strategies of\ndifferent models beyond classification accuracy, which potentially opens up new\nways for model improvements.\n",
"title": "Interpreting and Explaining Deep Neural Networks for Classification of Audio Signals"
} | null | null | null | null | true | null | 2896 | null | Default | null | null |
null | {
"abstract": " Graph matching in two correlated random graphs refers to the task of\nidentifying the correspondence between vertex sets of the graphs. Recent\nresults have characterized the exact information-theoretic threshold for graph\nmatching in correlated Erdős-Rényi graphs. However, very little is known\nabout the existence of efficient algorithms to achieve graph matching without\nseeds. In this work we identify a region in which a straightforward $O(n^2\\log\nn)$-time canonical labeling algorithm, initially introduced in the context of\ngraph isomorphism, succeeds in matching correlated Erdős-Rényi graphs.\nThe algorithm has two steps. In the first step, all vertices are labeled by\ntheir degrees and a trivial minimum distance matching (i.e., simply sorting\nvertices according to their degrees) matches a fixed number of highest degree\nvertices in the two graphs. Having identified this subset of vertices, the\nremaining vertices are matched using a matching algorithm for bipartite graphs.\n",
"title": "On the Performance of a Canonical Labeling for Matching Correlated Erdős-Rényi Graphs"
} | null | null | null | null | true | null | 2897 | null | Default | null | null |
null | {
"abstract": " Semi-supervised learning deals with the problem of how, if possible, to take\nadvantage of a huge amount of unclassified data, to perform a classification in\nsituations when, typically, there is little labeled data. Even though this is\nnot always possible (it depends on how useful, for inferring the labels, it\nwould be to know the distribution of the unlabeled data), several algorithm\nhave been proposed recently. %but in general they are not proved to outperform\nA new algorithm is proposed, that under almost necessary conditions, %and it is\nproved that it attains asymptotically the performance of the best theoretical\nrule as the amount of unlabeled data tends to infinity. The set of necessary\nassumptions, although reasonable, show that semi-supervised classification only\nworks for very well conditioned problems. The focus is on understanding when\nand why semi-supervised learning works when the size of the initial training\nsample remains fixed and the asymptotic is on the size of the unlabeled data.\nThe performance of the algorithm is assessed in the well known \"Isolet\"\nreal-data of phonemes, where a strong dependence on the choice of the initial\ntraining sample is shown.\n",
"title": "On semi-supervised learning"
} | null | null | [
"Statistics"
]
| null | true | null | 2898 | null | Validated | null | null |
null | {
"abstract": " We obtain an essential spectral gap for a convex co-compact hyperbolic\nsurface $M=\\Gamma\\backslash\\mathbb H^2$ which depends only on the dimension\n$\\delta$ of the limit set. More precisely, we show that when $\\delta>0$ there\nexists $\\varepsilon_0=\\varepsilon_0(\\delta)>0$ such that the Selberg zeta\nfunction has only finitely many zeroes $s$ with $\\Re s>\\delta-\\varepsilon_0$.\nThe proof uses the fractal uncertainty principle approach developed by\nDyatlov-Zahl [arXiv:1504.06589]. The key new component is a Fourier decay bound\nfor the Patterson-Sullivan measure, which may be of independent interest. This\nbound uses the fact that transformations in the group $\\Gamma$ are nonlinear,\ntogether with estimates on exponential sums due to Bourgain which follow from\nthe discretized sum-product theorem in $\\mathbb R$.\n",
"title": "Fourier dimension and spectral gaps for hyperbolic surfaces"
} | null | null | null | null | true | null | 2899 | null | Default | null | null |
null | {
"abstract": " In this work several semantic approaches to concept-based query expansion and\nreranking schemes are studied and compared with different ontology-based\nexpansion methods in web document search and retrieval. In particular, we focus\non concept-based query expansion schemes, where, in order to effectively\nincrease the precision of web document retrieval and to decrease the users\nbrowsing time, the main goal is to quickly provide users with the most suitable\nquery expansion. Two key tasks for query expansion in web document retrieval\nare to find the expansion candidates, as the closest concepts in web document\ndomain, and to rank the expanded queries properly. The approach we propose aims\nat improving the expansion phase for better web document retrieval and\nprecision. The basic idea is to measure the distance between candidate concepts\nusing the PMING distance, a collaborative semantic proximity measure, i.e. a\nmeasure which can be computed by using statistical results from web search\nengine. Experiments show that the proposed technique can provide users with\nmore satisfying expansion results and improve the quality of web document\nretrieval.\n",
"title": "Semantic Evolutionary Concept Distances for Effective Information Retrieval in Query Expansion"
} | null | null | null | null | true | null | 2900 | null | Default | null | null |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.