arXiv subjects

Yue Zhang

Publications and source records attributed to Yue Zhang.

At least 811 records · Page 45Linked to original sources

arXiv2017

Two-dimensional boron on Pb (110) surface

We simulate boron on Pb(110) surface by using ab initio evolutionary methodology. Interestingly, the two-dimensional (2D) Dirac Pmmn boron can be formed because of good lattice matching. Unexpectedly, by increasing the thickness of 2D boron, a three-bonded graphene-like structure (P2_1/c boron) was revealed to possess double anisotropic Dirac cones. It is 20 meV/atom lower in energy than the Pmmn structure, indicating the most stable 2D boron with particular Dirac cones. The puckered structure of P2_1/c boron results in the peculiar Dirac cones, as well as substantial mechanical anisotropy. The calculated Young's modulus is 320 GPa.nm along zigzag direction, which is comparable with graphene.

cond-mat.mtrl-sci↗

arXiv2017

AMR-to-text Generation with Synchronous Node Replacement Grammar

This paper addresses the task of AMR-to-text generation by leveraging synchronous node replacement grammar. During training, graph-to-string rules are learned using a heuristic extraction algorithm. At test time, a graph transducer is applied to collapse input AMRs and generate output sentences. Evaluated on SemEval-2016 Task 8, our method gives a BLEU score of 25.62, which is the best reported so far.

cs.CL↗

arXiv2017

Neural Word Segmentation with Rich Pretraining

Neural word segmentation research has benefited from large-scale raw texts by leveraging them for pretraining character and word embeddings. On the other hand, statistical segmentation research has exploited richer sources of external information, such as punctuation, automatic segmentation and POS. We investigate the effectiveness of a range of external training sources for neural word segmentation by building a modular segmentation model, pretraining the most important submodule using rich external sources. Results show that such pretraining significantly improves the model, leading to accuracies competitive to the best methods on six benchmarks.

cs.CL↗

arXiv2017

A Reappraisal on Dark Matter Co-annihilating with a Top/Bottom Partner

We revisit the calculation of relic density of dark matter particles co-annihilating with a top or bottom partner, by properly including the QCD bound-states (onia) effects of the colored partners, as well as the relevant electroweak processes which become important in the low mass region. We carefully set up the complete framework that incorporates the relevant contributions and investigate their effects on the cosmologically preferred mass spectrum, which turn out to be comparable in size to those coming from the Sommerfeld enhancement. We apply the calculation to three scenarios: bino-stop and bino-sbottom co-annihilations in supersymmetry, and a vector dark matter co-annihilating with a fermionic top partner. In addition, we confront our analysis of the relic abundance with recent direct detection experiments and collider searches at the LHC, which have important implications in the bino-stop and bino-sbottom scenarios. In particular, in the bino-stop case recent LHC limits have excluded regions of parameter with a direct detection rate that is above the neutrino floor.

hep-ph↗

arXiv2017

Optimal Control of Connected Automated Vehicles at Urban Traffic Intersections: A Feasibility Enforcement Analysis

Earlier work has established a decentralized optimal control framework for coordinating online a continuous flow of connected automated vehicles (CAVs) entering a control zone and crossing two adjacent intersections in an urban area. A solution, when it exists, allows the vehicles to cross the intersections without the use of traffic lights, without creating congestion on the connecting road, and under the hard safety constraint of collision avoidance. We establish the conditions under which such solutions exist and show that they can be enforced through an appropriately designed feasibility enforcement zone that precedes the control zone. The proposed solution and overall control architecture are illustrated through simulation.

math.OC↗

arXiv2017

On the Unambiguous Distance of Multi-Carrier Phase Ranging with Random Hopped Frequencie

In a multi-carrier phase ranging (MPR) system, the distance that radio signal travels is estimated through phase shift of multiple carrier frequencies. Due to phase ambiguity, a unique estimation can only be obtained within the unambiguous distance (UD), which depends on the carrier frequencies used for ranging. Without external information, the maximum measurable distance of an MPR system is defined by its UD. The MPR system employing random frequency hopping (FH) waveform has a strong anti-jamming capability and sees promising potentials in many fields. However, it is challenging to depict its measurable distance. Different from current MPR system employing deterministic frequencies, the carrier frequency under the FH waveform hops randomly within the occupied bandwidth. Consequently, the UD is a random variable. In this paper, we try to find a deterministic value to depict the UD of an MPR system under the random spaced frequencies (RSF) configuration, serving as a metric of its measurable distance in engineering applications. It is safe to adopt the lower bound of the random UD as the metric, but the measurable distance may be drastically underestimated. Alternatively, we propose to adopt the upper bound of the random UD as the metric, because when the RSF set contains more than a dozen of carriers, i) we prove the probability that the random UD obtains its upper bound is very close to 1 if phase noise is not introduced; ii) simulations show that the upper bound can also be obtained reliably in the presence of phase noise. As a comparison, the practical measurable range of the MPR system under the traditional linearly spaced frequencies (LSF) configuration is only a fraction of its theoretical UD because the UD is more sensitive to phase error.

cs.NI↗

arXiv2017

Opinion Recommendation using Neural Memory Model

We present opinion recommendation, a novel task of jointly predicting a custom review with a rating score that a certain user would give to a certain product or service, given existing reviews and rating scores to the product or service by other users, and the reviews that the user has given to other products and services. A characteristic of opinion recommendation is the reliance of multiple data sources for multi-task joint learning, which is the strength of neural models. We use a single neural network to model users and products, capturing their correlation and generating customised product representations using a deep memory network, from which customised ratings and reviews are constructed jointly. Results show that our opinion recommendation system gives ratings that are closer to real user ratings on Yelp.com data compared with Yelp's own ratings, and our methods give better results compared to several pipelines baselines using state-of-the-art sentiment rating and summarization systems.

cs.CL↗

arXiv2017

Self-interacting Dark Matter Without Direct Detection Constraints

We explore the self-interacting dark matter scenario in a simple dark sector model where the dark matter interacts through a dark photon. Splitting a Dirac fermion dark matter into two levels using a small Majorana mass can evade strong direct detection constraints on the kinetic mixing between the dark and normal photons, thus allowing the dark sector to be more visible at high intensity and/or high energy experiments. It is pointed out that such a mass splitting has a strong impact on the dark matter self-interaction strength. We derive the new parameter space of a pseudo-Dirac self-interacting dark matter. Interestingly, with increasing mass splitting, a weak scale dark matter mass window survives that could be probed by the LHC and future colliders.

hep-ph↗

arXiv2016

Shift-Reduce Constituent Parsing with Neural Lookahead Features

Transition-based models can be fast and accurate for constituent parsing. Compared with chart-based models, they leverage richer features by extracting history information from a parser stack, which spans over non-local constituents. On the other hand, during incremental parsing, constituent information on the right hand side of the current word is not utilized, which is a relative weakness of shift-reduce parsing. To address this limitation, we leverage a fast neural model to extract lookahead features. In particular, we build a bidirectional LSTM model, which leverages the full sentence information to predict the hierarchy of constituents that each word starts and ends. The results are then passed to a strong transition-based constituent parser as lookahead features. The resulting parser gives 1.3% absolute improvement in WSJ and 2.3% in CTB compared to the baseline, given the highest reported accuracies for fully-supervised parsing.

cs.CL↗

arXiv2016

Bidirectional Tree-Structured LSTM with Head Lexicalization

Sequential LSTM has been extended to model tree structures, giving competitive results for a number of tasks. Existing methods model constituent trees by bottom-up combinations of constituent nodes, making direct use of input word information only for leaf nodes. This is different from sequential LSTMs, which contain reference to input words for each node. In this paper, we propose a method for automatic head-lexicalization for tree-structure LSTMs, propagating head words from leaf nodes to every constituent node. In addition, enabled by head lexicalization, we build a tree LSTM in the top-down direction, which corresponds to bidirectional sequential LSTM structurally. Experiments show that both extensions give better representations of tree structures. Our final model gives the best results on the Standford Sentiment Treebank and highly competitive results on the TREC question type classification task.

cs.CL↗

arXiv2016

Distance Metric Learning for Aspect Phrase Grouping

Aspect phrase grouping is an important task in aspect-level sentiment analysis. It is a challenging problem due to polysemy and context dependency. We propose an Attention-based Deep Distance Metric Learning (ADDML) method, by considering aspect phrase representation as well as context representation. First, leveraging the characteristics of the review text, we automatically generate aspect phrase sample pairs for distant supervision. Second, we feed word embeddings of aspect phrases and their contexts into an attention-based neural network to learn feature representation of contexts. Both aspect phrase embedding and context embedding are used to learn a deep feature subspace for measure the distances between aspect phrases for K-means clustering. Experiments on four review datasets show that the proposed method outperforms state-of-the-art strong baseline methods.

cs.CL↗

arXiv2016

A Corrected Parsimony Criterion for Reconstructing Phylogenies

In phylogenetic analysis, for non-molecular data, particularly morphology, parsimony optimization is the most commonly employed approach. In the past and present application of the parsimony principle, extra step numbers have been added across different characters without proper justification. This practice, however, has caused the impacts of characters to be inflated or deflated without a valid reason. To resolve this methodological deficiency, here I present a corrected parsimony criterion for reconstructing phylogenies. In essence, character rather than step is the most fundamental unit. Accordingly, the most parsimonious tree should maximize the sum or average of the phylogenetic signals, quantified by retention index, contributed by each character. Assigning proper weights to characters is a separate task that requires information other than the intra-character step number changing range.

q-bio.PE↗

arXiv2016

Training Dependency Parsers with Partial Annotation

Recently, these has been a surge on studying how to obtain partially annotated data for model supervision. However, there still lacks a systematic study on how to train statistical models with partial annotation (PA). Taking dependency parsing as our case study, this paper describes and compares two straightforward approaches for three mainstream dependency parsers. The first approach is previously proposed to directly train a log-linear graph-based parser (LLGPar) with PA based on a forest-based objective. This work for the first time proposes the second approach to directly training a linear graph-based parse (LGPar) and a linear transition-based parser (LTPar) with PA based on the idea of constrained decoding. We conduct extensive experiments on Penn Treebank under three different settings for simulating PA, i.e., random dependencies, most uncertain dependencies, and dependencies with divergent outputs from the three parsers. The results show that LLGPar is most effective in learning from PA and LTPar lags behind the graph-based counterparts by large margin. Moreover, LGPar and LTPar can achieve best performance by using LLGPar to complete PA into full annotation (FA).

cs.CL↗

arXiv2016

AMR-to-text generation as a Traveling Salesman Problem

The task of AMR-to-text generation is to generate grammatical text that sustains the semantic meaning for a given AMR graph. We at- tack the task by first partitioning the AMR graph into smaller fragments, and then generating the translation for each fragment, before finally deciding the order by solving an asymmetric generalized traveling salesman problem (AGTSP). A Maximum Entropy classifier is trained to estimate the traveling costs, and a TSP solver is used to find the optimized solution. The final model reports a BLEU score of 22.44 on the SemEval-2016 Task8 dataset.

cs.CL↗

arXiv2016

A gate-variable spin current demultiplexer based on graphene

Spintronics, which utilizes spin as information carrier, is a promising solution for nonvolatile memory and low-power computing in the post-Moore era. An important challenge is to realize long distance spin transport, together with efficient manipulation of spin current for novel logic-processing applications. Here, we describe a gate-variable spin current demultiplexer (GSDM) based on graphene, serving as a fundamental building block of reconfigurable spin current logic circuits. The concept relies on electrical gating of carrier density dependent conductivity and spin diffusion length in graphene. As a demo, GSDM is realized for both single-layer and bilayer graphene. The distribution and propagation of spin current in the two branches of GSDM depend on spin relaxation characteristics of graphene. Compared with Elliot-Yafet spin relaxation mechanism, D'yakonov-Perel mechanism results in more appreciable gate-tuning performance. These unique features of GSDM would give rise to abundant spin logic applications, such as on-chip spin current modulators and reconfigurable spin logic circuits.

cond-mat.mes-hall↗

arXiv2016

Stable Bound States of Asymmetric Dark Matter

The simplest renormalizable effective field theories with asymmetric dark matter bound states contain two additional gauge singlet fields one being the dark matter and the other a mediator particle that the dark matter annihilates into. We examine the physics of one such model with a Dirac fermion as the dark matter and a real scalar mediator. For a range of parameters the Yukawa coupling of the dark matter to the mediator gives rise to stable asymmetric dark matter bound states. We derive properties of the bound states including nuggets formed from $N\gg1$ dark matter particles. We also consider the formation of bound states in the early universe and direct detection of dark matter bound states. Many of our results also hold for symmetric dark matter.

hep-ph↗

arXiv2016

Strong CMB Constraint On P-Wave Annihilating Dark Matter

We consider a dark sector consisting of dark matter that is a Dirac fermion and a scalar mediator. This model has been extensively studied in the past. If the scalar couples to the dark matter in a parity conserving manner then dark matter annihilation to two mediators is dominated by the P-wave channel and hence is suppressed at very low momentum. The indirect detection constraint from the anisotropy of the Cosmic Microwave Background is usually thought to be absent in the model because of this suppression. In this letter we show that dark matter annihilation to bound states occurs through the S-wave and hence there is a constraint on the parameter space of the model from the Cosmic Microwave Background.

hep-ph↗

arXiv2016

Global analysis of isospin dependent microscopic nucleon-nucleus optical potential in Dirac Bruckner Hartree-Fock approach

The Microscopic Otical Model Potential is evaluated within a relativistic scheme which provides a natural and consistent relation between the spin-orbit part and the central part of the potential. The Dirac-Brueckner-Hartree-Fock (DBHF) approach provides such a microscopic relativistic scheme, which is based on a realistic nucleon-nucleon interaction and reproduce the saturation properties of symmetric nuclear matter without any adjustable parameter. Its solution using the projection technique within the subtracted T-matrix (STM) representation provides a reliable extension to asymmetric nuclear matter, which is important to describe the features of the isospin asymmetric nuclei. Therefore, the present work aims to perform a global analysis of the isospin-dependent nucleon-nucleus MOP based on the DBHF calculation in symmetric and asymmetric nuclear matter. The DBHF is used to evaluate the relativistic structure of the nucleon self-energies in nuclear matter at various densities and asymmetries. The Schrödinger equivalent potentials of finite nuclei are derived from these Dirac components by a local density approximation (LDA). The nucleon-nucleus scattering calculations are carried out for a broad spectrum $n$ and $p$ scattering experiments below 200 MeV with targets ranging from $^{12}$C to $^{208}$Pb.

nucl-th↗