Open Contents
연도
 2022173,597
 2021181,587
 2020178,291
 2019155,916
 2018140,392
 2017123,781
 2016113,436
 2015105,128
 201497,598
 201392,866
 201284,389
 201176,605
 201070,286
 200964,077
 200858,796
 200755,768
 200650,304
 200546,867
 200443,721
 200339,389
 200236,103
 200133,128
 200030,672
 199927,694
 199824,174
 199719,621
 199615,875
 199512,994
 199410,085
 19936,728
 19923,181
 1991370
 199025
 19898
 19881
 19861

arXivDocumentLevel Abstractive Summarization Gonçalo Raposo, Afonso Raposo, Ana Sofia Carmo... more(3) 20221206 The task of automatic text summarization produces a concise and fluent text summary while preserving key information and overall meaning. Recent approaches to documentlevel summarization have seen significant improvements in recent years by using models based on the Transformer architecture. However, the quadratic memory and time complexities with respect to the sequence length make them very expensive to use, especially with long sequences, as required by documentlevel summarization. Our work addresses the problem of documentlevel summarization by studying how efficient Transformer techniques can be used to improve the automatic summarization of very long texts. In particular, we will use the arXiv dataset, consisting of several scientific papers and the corresponding abstracts, as baselines for this work. Then, we propose a novel retrievalenhanced approach based on the architecture which reduces the cost of generating a summary of the entire document by processing smaller chunks. The results were below the baselines but suggest a more efficient memory a consumption and truthfulness.
Subject Source arXiv URL https://arxiv.org/abs/2212.03013view Article Title DocumentLevel Abstractive SummarizationAuthors Gonçalo Raposo; Afonso Raposo; Ana Sofia CarmoAbstract The task of automatic text summarization produces a concise and fluent text summary while preserving key information and overall meaning. Recent approaches to documentlevel summarization have seen significant improvements in recent years by using models based on the Transformer architecture. However, the quadratic memory and time complexities with respect to the sequence length make them very expensive to use, especially with long sequences, as required by documentlevel summarization. Our work addresses the problem of documentlevel summarization by studying how efficient Transformer techniques can be used to improve the automatic summarization of very long texts. In particular, we will use the arXiv dataset, consisting of several scientific papers and the corresponding abstracts, as baselines for this work. Then, we propose a novel retrievalenhanced approach based on the architecture which reduces the cost of generating a summary of the entire document by processing smaller chunks. The results were below the baselines but suggest a more efficient memory a consumption and truthfulness.Is Part Of 20221206 Identifier ISSN: Category cs.CL cs.IRLicense 
arXivHeat transfer in the extreme nearfield regime between metallic bodies Mauricio Gómez Viloria, Yangyu Guo, Samy Merabia, ... more(5) 20221206 We analyze the heat transfer between two metals separated by a vacuum gap in the extreme nearfield regime. In this crossover regime between conduction and radiation heat exchanges are mediated by photon, phonon and electron tunneling. We quantify the relative contribution of these carriers with respect to both the separation distance between the two bodies and the applied bias voltage. In presence of a weak bias ($V_{\rm b}<100$~mV), electrons and phonons can contribute equally to the heat transfer near contact, while the contribution of photons becomes negligible. On the other hand, for larger bias voltages, electrons play a dominant role. Moreover, we demonstrate that depending on the sign of this bias, electrons can either heat up the cold body or pump heat from it by Nottingham effect. Our results emphasize some inconsistencies in recent experimental results about heat exchanges in extreme nearfield regime and it sets a roadmap for future experiments.
Subject Source arXiv URL https://arxiv.org/abs/2212.03073view Article Title Heat transfer in the extreme nearfield regime between metallic bodiesAuthors Mauricio Gómez Viloria; Yangyu Guo; Samy Merabia; Philippe BenAbdallah; Riccardo MessinaAbstract We analyze the heat transfer between two metals separated by a vacuum gap in the extreme nearfield regime. In this crossover regime between conduction and radiation heat exchanges are mediated by photon, phonon and electron tunneling. We quantify the relative contribution of these carriers with respect to both the separation distance between the two bodies and the applied bias voltage. In presence of a weak bias ($V_{\rm b}<100$~mV), electrons and phonons can contribute equally to the heat transfer near contact, while the contribution of photons becomes negligible. On the other hand, for larger bias voltages, electrons play a dominant role. Moreover, we demonstrate that depending on the sign of this bias, electrons can either heat up the cold body or pump heat from it by Nottingham effect. Our results emphasize some inconsistencies in recent experimental results about heat exchanges in extreme nearfield regime and it sets a roadmap for future experiments.Is Part Of 20221206 Identifier ISSN: Category condmat.meshallLicense 
arXivState Space Closure: Revisiting Endless Online Level Generation via Reinforcement Learning Ziqi Wang, Tianye Shu, Jialin Liu... more(3) 20221206 In this paper we revisit endless online level generation with the recently proposed experiencedriven procedural content generation via reinforcement learning (EDRL) framework, from an observation that EDRL tends to generate recurrent patterns. Inspired by this phenomenon, we formulate a notion of state space closure, which means that any state that may appear in an infinitehorizon online generation process can be found in a finite horizon. Through theoretical analysis we find that though state space closure arises a concern about diversity, it makes the EDRL trained on a finitehorizon generalised to the infinitehorizon scenario without deterioration of content quality. Moreover, we verify the quality and diversity of contents generated by EDRL via empirical studies on the widely used Super Mario Bros. benchmark. Experimental results reveal that the current EDRL approach's ability of generating diverse game levels is limited due to the state space closure, whereas it does not suffer from reward deterioration given a horizon longer than the one of training. Concluding our findings and analysis, we argue that future works in generating online diverse and highquality contents via EDRL should address the issue of diversity on the premise of state space closure which ensures the quality.
Subject Source arXiv URL https://arxiv.org/abs/2212.02951view Article Title State Space Closure: Revisiting Endless Online Level Generation via Reinforcement LearningAuthors Ziqi Wang; Tianye Shu; Jialin LiuAbstract In this paper we revisit endless online level generation with the recently proposed experiencedriven procedural content generation via reinforcement learning (EDRL) framework, from an observation that EDRL tends to generate recurrent patterns. Inspired by this phenomenon, we formulate a notion of state space closure, which means that any state that may appear in an infinitehorizon online generation process can be found in a finite horizon. Through theoretical analysis we find that though state space closure arises a concern about diversity, it makes the EDRL trained on a finitehorizon generalised to the infinitehorizon scenario without deterioration of content quality. Moreover, we verify the quality and diversity of contents generated by EDRL via empirical studies on the widely used Super Mario Bros. benchmark. Experimental results reveal that the current EDRL approach's ability of generating diverse game levels is limited due to the state space closure, whereas it does not suffer from reward deterioration given a horizon longer than the one of training. Concluding our findings and analysis, we argue that future works in generating online diverse and highquality contents via EDRL should address the issue of diversity on the premise of state space closure which ensures the quality.Is Part Of 20221206 Identifier ISSN: Category cs.AILicense 
arXivUltrafast synthesis of SiC nanowire webs by floating catalysts rationalised through insitu measurements and thermodynamic calculations Isabel GómezPalos, Miguel VazquezPufleau, Jorge Valilla, ... more(6) 20221206 This work presents the synthesis of SiC nanowires floating in a gas stream through the vapourliquidsolid (VLS) mechanism using an aerosol of catalyst nanoparticles. These conditions lead to ultrafast growth at 8.5 {\mu}m/s (maximum of 50 {\mu}m/s), which is up to 3 orders of magnitude above conventional substratebased chemical vapour deposition. The high aspect ratio of the nanowires (up to 2200) favours their entanglement and the formation of freestanding network materials consisting entirely of SiCNWs. The floating catalyst chemical vapour deposition growth process is rationalised through insitu sampling of reaction products and catalyst aerosol from the gas phase, and thermodynamic calculations of the bulk ternary SiCFe phase diagram. The phase diagram suggests a description of the mechanistic path for the selective growth of SiCNWs, consistent with the observation that no other types of nanowires (Si or C) are grown by the catalyst. SiCNW growth occurs at 1130 {\deg}C, close to the calculated eutectic. According to the calculated phase diagram, upon addition of Si and C, the Ferich liquid segregates a carbon shell, and later enrichment of the liquid in Si leads to the formation of SiC. The exceptionally fast growth rate relative to substratebased processes is attributed to the increased availability of precursors for incorporation into the catalyst due to the high collision rate inherent to this new synthesis mode.
Subject Source arXiv URL https://arxiv.org/abs/2212.02901view Article Title Ultrafast synthesis of SiC nanowire webs by floating catalysts rationalised through insitu measurements and thermodynamic calculationsAuthors Isabel GómezPalos; Miguel VazquezPufleau; Jorge Valilla; Álvaro Ridruejo; Damien Tourret; Juan J. VilatelaAbstract This work presents the synthesis of SiC nanowires floating in a gas stream through the vapourliquidsolid (VLS) mechanism using an aerosol of catalyst nanoparticles. These conditions lead to ultrafast growth at 8.5 {\mu}m/s (maximum of 50 {\mu}m/s), which is up to 3 orders of magnitude above conventional substratebased chemical vapour deposition. The high aspect ratio of the nanowires (up to 2200) favours their entanglement and the formation of freestanding network materials consisting entirely of SiCNWs. The floating catalyst chemical vapour deposition growth process is rationalised through insitu sampling of reaction products and catalyst aerosol from the gas phase, and thermodynamic calculations of the bulk ternary SiCFe phase diagram. The phase diagram suggests a description of the mechanistic path for the selective growth of SiCNWs, consistent with the observation that no other types of nanowires (Si or C) are grown by the catalyst. SiCNW growth occurs at 1130 {\deg}C, close to the calculated eutectic. According to the calculated phase diagram, upon addition of Si and C, the Ferich liquid segregates a carbon shell, and later enrichment of the liquid in Si leads to the formation of SiC. The exceptionally fast growth rate relative to substratebased processes is attributed to the increased availability of precursors for incorporation into the catalyst due to the high collision rate inherent to this new synthesis mode.Is Part Of 20221206 Identifier ISSN: DOI 10.1039/D2NR06016GCategory condmat.mtrlsci physics.appphLicense 
arXivLessons from the Mathematics of TwoDimensional Euclidean Quantum Gravity Timothy Budd 20221206 The search for a mathematical foundation for the path integral of Euclidean quantum gravity calls for the construction of random geometry on the spacetime manifold. Following developments in physics on the twodimensional theory, random geometry on the 2sphere has in recent years received much attention in the mathematical literature, which has led to a fully rigorous implementation of the path integral formulation of twodimensional Euclidean quantum gravity. In this chapter we review several important mathematical developments that may serve as guiding principles for approaching Euclidean quantum gravity in dimensions higher than two. Our starting point is the discrete geometry encoded by random planar maps, which realizes a lattice discretization of the path integral. We recap the enumeration of planar maps via their generating functions and show how bijections with trees explain the surprising simplicity of some of these. Then we explain how to handle infinite planar maps and to analyze their exploration via the peeling process. The aforementioned trees provide the basis for the construction of the universal continuum limit of the random discrete geometries, known as the Brownian sphere, which represents the random geometry underlying twodimensional Euclidean quantum gravity in the absence of matter.
Subject Source arXiv URL https://arxiv.org/abs/2212.03031view Article Title Lessons from the Mathematics of TwoDimensional Euclidean Quantum GravityAuthors Timothy BuddAbstract The search for a mathematical foundation for the path integral of Euclidean quantum gravity calls for the construction of random geometry on the spacetime manifold. Following developments in physics on the twodimensional theory, random geometry on the 2sphere has in recent years received much attention in the mathematical literature, which has led to a fully rigorous implementation of the path integral formulation of twodimensional Euclidean quantum gravity. In this chapter we review several important mathematical developments that may serve as guiding principles for approaching Euclidean quantum gravity in dimensions higher than two. Our starting point is the discrete geometry encoded by random planar maps, which realizes a lattice discretization of the path integral. We recap the enumeration of planar maps via their generating functions and show how bijections with trees explain the surprising simplicity of some of these. Then we explain how to handle infinite planar maps and to analyze their exploration via the peeling process. The aforementioned trees provide the basis for the construction of the universal continuum limit of the random discrete geometries, known as the Brownian sphere, which represents the random geometry underlying twodimensional Euclidean quantum gravity in the absence of matter.Is Part Of 20221206 Identifier ISSN: Category grqcLicense 
arXivSearches for MassAsymmetric Compact Binary Coalescence Events using Neural Networks in the LIGO/Virgo Third Observation Period M. AndresCarcasona, A. MenendezVazquez, M. Martinez, ... more(4) 20221206 We present the results on the search for the coalescence of compact binary mergers with very asymmetric mass configurations using convolutional neural networks and the LIGO/Virgo data for the O3 observation period. Twodimensional images in time and frequency are used as input. Masses in the range between 0.01 Msun and 20 Msun are considered. We explore neural networks trained with input information from a single interferometer, pairs of interferometers, or all three interferometers together, indicating that the use of the maximum information available leads to an improved performance. A scan over the O3 data set using the convolutional neural networks for detection results into no significant excess from an onlynoise hypothesis. The results are translated into 90% confidence level upper limits on the merger rate as a function of the mass parameters of the binary system.
Subject Source arXiv URL https://arxiv.org/abs/2212.02829view Article Title Searches for MassAsymmetric Compact Binary Coalescence Events using Neural Networks in the LIGO/Virgo Third Observation PeriodAuthors M. AndresCarcasona; A. MenendezVazquez; M. Martinez; Ll. M. MirAbstract We present the results on the search for the coalescence of compact binary mergers with very asymmetric mass configurations using convolutional neural networks and the LIGO/Virgo data for the O3 observation period. Twodimensional images in time and frequency are used as input. Masses in the range between 0.01 Msun and 20 Msun are considered. We explore neural networks trained with input information from a single interferometer, pairs of interferometers, or all three interferometers together, indicating that the use of the maximum information available leads to an improved performance. A scan over the O3 data set using the convolutional neural networks for detection results into no significant excess from an onlynoise hypothesis. The results are translated into 90% confidence level upper limits on the merger rate as a function of the mass parameters of the binary system.Is Part Of 20221206 Identifier ISSN: Category grqcLicense 
arXivHow to Compare Fuzzers Philipp Görz, Björn Mathis, Keno Hassler, ... more(7) 20221206 Fuzzing is a key method to discover vulnerabilities in programs. Despite considerable progress in this area in the past years, measuring and comparing the effectiveness of fuzzers is still an open research question. In software testing, the gold standard for evaluating test quality is mutation analysis, assessing the ability of a test to detect synthetic bugs; if a set of tests fails to detect such mutations, it will also fail to detect real bugs. Mutation analysis subsumes various coverage measures and provides a large and diverse set of faults that can be arbitrarily hard to trigger and detect, thus preventing the problems of saturation and overfitting. Unfortunately, the cost of traditional mutation analysis is exorbitant for fuzzing, as mutations need independent evaluation. In this paper, we apply modern mutation analysis techniques that pool multiple mutations; allowing us, for the first time, to evaluate and compare fuzzers with mutation analysis. We introduce an evaluation bench for fuzzers and apply it to a number of popular fuzzers and subjects. In a comprehensive evaluation, we show how it allows us to assess fuzzer performance and measure the impact of improved techniques. While we find that today's fuzzers can detect only a small percentage of mutations, this should be seen as a challenge for future research  notably in improving (1) detecting failures beyond generic crashes (2) triggering mutations (and thus faults).
Subject Source arXiv URL https://arxiv.org/abs/2212.03075view Article Title How to Compare FuzzersAuthors Philipp Görz; Björn Mathis; Keno Hassler; Emre Güler; Thorsten Holz; Andreas Zeller; Rahul GopinathAbstract Fuzzing is a key method to discover vulnerabilities in programs. Despite considerable progress in this area in the past years, measuring and comparing the effectiveness of fuzzers is still an open research question. In software testing, the gold standard for evaluating test quality is mutation analysis, assessing the ability of a test to detect synthetic bugs; if a set of tests fails to detect such mutations, it will also fail to detect real bugs. Mutation analysis subsumes various coverage measures and provides a large and diverse set of faults that can be arbitrarily hard to trigger and detect, thus preventing the problems of saturation and overfitting. Unfortunately, the cost of traditional mutation analysis is exorbitant for fuzzing, as mutations need independent evaluation. In this paper, we apply modern mutation analysis techniques that pool multiple mutations; allowing us, for the first time, to evaluate and compare fuzzers with mutation analysis. We introduce an evaluation bench for fuzzers and apply it to a number of popular fuzzers and subjects. In a comprehensive evaluation, we show how it allows us to assess fuzzer performance and measure the impact of improved techniques. While we find that today's fuzzers can detect only a small percentage of mutations, this should be seen as a challenge for future research  notably in improving (1) detecting failures beyond generic crashes (2) triggering mutations (and thus faults).Is Part Of 20221206 Identifier ISSN: Category cs.SE cs.CRLicense 
arXivOn the statistical relation between the halo mass function and the internal structure of dark matter haloes T. R. G. Richardson, P. S. Corasaniti 20221206 Context. Large and complete galaxy cluster samples can be used to infer cosmological parameter constraints from number count measurements. Key to their interpretation is the availability of accurately calibrated estimates of the halo mass function from cosmological simulations. Galaxy cluster masses are usually defined as the mass within a spherical region enclosing a given matter overdensity (in units of the critical density). However, this may differ from the mass definition of the numerical halo catalogues that are used to calibrate the halo mass function. Aims. In this article, we present a generic non parametric formalism that allows one to accurately map the halo mass function between different mass overdensity definitions using the distribution of halo sparsities defined as the ratios of both masses. We show that changing mass definitions reduces to modelling the distribution of halo sparsities. Methods. Using standard transformation rules of random variates, we derive relations between the halo mass function at different overdensities and the distribution of halo sparsities. Results. We show that these relations reproduce the Nbody halo mass functions from the Uchuu simulation within the statistical errors at a few percent level. Furthermore, these relations allow one to relate the halo mass functions at different overdensities to parametric descriptions of the halo density profile. In particular, we will discuss the case of the concentrationmass relation of the NavarroFrenkWhite profile. Finally, we will show that the use of such relations allows to predict the distribution of sparsities of a sample of haloes of a given mass, thus opening the way to inferring cosmological constraints from individual galaxy cluster sparsity measurements.
Subject Source arXiv URL https://arxiv.org/abs/2212.03233view Article Title On the statistical relation between the halo mass function and the internal structure of dark matter haloesAuthors T. R. G. Richardson; P. S. CorasanitiAbstract Context. Large and complete galaxy cluster samples can be used to infer cosmological parameter constraints from number count measurements. Key to their interpretation is the availability of accurately calibrated estimates of the halo mass function from cosmological simulations. Galaxy cluster masses are usually defined as the mass within a spherical region enclosing a given matter overdensity (in units of the critical density). However, this may differ from the mass definition of the numerical halo catalogues that are used to calibrate the halo mass function. Aims. In this article, we present a generic non parametric formalism that allows one to accurately map the halo mass function between different mass overdensity definitions using the distribution of halo sparsities defined as the ratios of both masses. We show that changing mass definitions reduces to modelling the distribution of halo sparsities. Methods. Using standard transformation rules of random variates, we derive relations between the halo mass function at different overdensities and the distribution of halo sparsities. Results. We show that these relations reproduce the Nbody halo mass functions from the Uchuu simulation within the statistical errors at a few percent level. Furthermore, these relations allow one to relate the halo mass functions at different overdensities to parametric descriptions of the halo density profile. In particular, we will discuss the case of the concentrationmass relation of the NavarroFrenkWhite profile. Finally, we will show that the use of such relations allows to predict the distribution of sparsities of a sample of haloes of a given mass, thus opening the way to inferring cosmological constraints from individual galaxy cluster sparsity measurements.Is Part Of 20221206 Identifier ISSN: Category astroph.COLicense 
arXivAn explicit formula for the BenjaminOno equation Patrick Gerard 20221206 We establish an explicit formula for the general solution of the BenjaminOno equation on the torus and on the line. Contents 1. Introduction 1 1.1. The BenjaminOno equation 1 1.2. The Lax pair 2 1.3. The explicit formula on the torus 3 1.4. The explicit formula on the line 3 1.5. Organization of the paper 4 2. Proof of the explicit formula on the torus 4 3. Proof of the explicit formula on the line 6 4. Final remarks 8 References 9
Subject 수학 Source arXiv URL https://arxiv.org/abs/2212.03139view Article Title An explicit formula for the BenjaminOno equationAuthors Patrick GerardAbstract We establish an explicit formula for the general solution of the BenjaminOno equation on the torus and on the line. Contents 1. Introduction 1 1.1. The BenjaminOno equation 1 1.2. The Lax pair 2 1.3. The explicit formula on the torus 3 1.4. The explicit formula on the line 3 1.5. Organization of the paper 4 2. Proof of the explicit formula on the torus 4 3. Proof of the explicit formula on the line 6 4. Final remarks 8 References 9Is Part Of 20221206 Identifier ISSN: Category math.APLicense 
arXivGeneration of perfectly entangled two and three qubits states by classical random interaction Javed Akram 20221206 This study examines the possibility of finding perfect entanglers for a Hamiltonian which corresponds to several quantum information platforms of interest at the present time. However, in this study, we use a superconducting circuit that stands out from other quantumcomputing devices, especially because Transmon qubits can be coupled via capacitors or microwave cavities, which enable us to combine high coherence, fast gates, and high flexibility in its design parameters. There are currently two factors limiting the performance of superconducting processors: timing mismatch and the limitation of entangling gates to two qubits. In this work, we present a twoqubit SWAP and a threequbit Fredkin gate, additionally, we also demonstrate a perfect adiabatic entanglement generation between two and three programmable superconducting qubits. Furthermore, in this study, we also demonstrate the impact of random dephasing, emission, and absorption noises on the quantum gates and entanglement. It is demonstrated by numerical simulation that the CSWAP gate and $W$state generation can be achieved perfectly in one step with high reliability under weak coupling conditions. Hence, our scheme could contribute to quantum teleportation, quantum communication, and some other areas of quantum information processing.
Subject Source arXiv URL https://arxiv.org/abs/2212.03115view Article Title Generation of perfectly entangled two and three qubits states by classical random interactionAuthors Javed AkramAbstract This study examines the possibility of finding perfect entanglers for a Hamiltonian which corresponds to several quantum information platforms of interest at the present time. However, in this study, we use a superconducting circuit that stands out from other quantumcomputing devices, especially because Transmon qubits can be coupled via capacitors or microwave cavities, which enable us to combine high coherence, fast gates, and high flexibility in its design parameters. There are currently two factors limiting the performance of superconducting processors: timing mismatch and the limitation of entangling gates to two qubits. In this work, we present a twoqubit SWAP and a threequbit Fredkin gate, additionally, we also demonstrate a perfect adiabatic entanglement generation between two and three programmable superconducting qubits. Furthermore, in this study, we also demonstrate the impact of random dephasing, emission, and absorption noises on the quantum gates and entanglement. It is demonstrated by numerical simulation that the CSWAP gate and $W$state generation can be achieved perfectly in one step with high reliability under weak coupling conditions. Hence, our scheme could contribute to quantum teleportation, quantum communication, and some other areas of quantum information processing.Is Part Of 20221206 Identifier ISSN: Category quantph condmat.suprconLicense