arXiv subjects

Xing Wang

Publications and source records attributed to Xing Wang.

At least 217 records · Page 12Linked to original sources

arXiv2019

Self-Attention with Structural Position Representations

Although self-attention networks (SANs) have advanced the state-of-the-art on various NLP tasks, one criticism of SANs is their ability of encoding positions of input words (Shaw et al., 2018). In this work, we propose to augment SANs with structural position representations to model the latent structure of the input sentence, which is complementary to the standard sequential positional representations. Specifically, we use dependency tree to represent the grammatical structure of a sentence, and propose two strategies to encode the positional relationships among words in the dependency tree. Experimental results on NIST Chinese-to-English and WMT14 English-to-German translation tasks show that the proposed approach consistently boosts performance over both the absolute and relative sequential position representations.

cs.CL↗

arXiv2019

Transverse Parton Distribution and Fragmentation Functions at NNLO: the Quark Case

We revisit the calculation of perturbative quark transverse momentum dependent parton distribution functions and fragmentation functions using the exponential regulator for rapidity divergences. We show that the exponential regulator provides a consistent framework for the calculation of various ingredients in transverse momentum dependent factorization. Compared to existing regulators in the literature, the exponential regulator has a couple of advantages which we explain in detail. As a result, the calculation is greatly simplified and we are able to obtain the next-to-next-to-leading order results up to $\mathcal{O}(ε^2)$ in dimensional regularization. These terms are necessary for a higher order calculation which is made possible with the simplification brought by the new regulator. As a by-product, we have obtained the two-loop quark jet function for the Energy-Energy Correlator in the back-to-back limit, which is the last missing ingredient for its N$^3$LL resummation.

hep-ph↗

arXiv2019

Exploiting Sentential Context for Neural Machine Translation

In this work, we present novel approaches to exploit sentential context for neural machine translation (NMT). Specifically, we first show that a shallow sentential context extracted from the top encoder layer only, can improve translation performance via contextualizing the encoding representations of individual words. Next, we introduce a deep sentential context, which aggregates the sentential context representations from all the internal layers of the encoder to form a more comprehensive context representation. Experimental results on the WMT14 English-to-German and English-to-French benchmarks show that our model consistently improves performance over the strong TRANSFORMER model (Vaswani et al., 2017), demonstrating the necessity and effectiveness of exploiting sentential context for NMT.

cs.CL↗

arXiv2019

Multiple Independent Subspace Clusterings

Multiple clustering aims at discovering diverse ways of organizing data into clusters. Despite the progress made, it's still a challenge for users to analyze and understand the distinctive structure of each output clustering. To ease this process, we consider diverse clusterings embedded in different subspaces, and analyze the embedding subspaces to shed light into the structure of each clustering. To this end, we provide a two-stage approach called MISC (Multiple Independent Subspace Clusterings). In the first stage, MISC uses independent subspace analysis to seek multiple and statistical independent (i.e. non-redundant) subspaces, and determines the number of subspaces via the minimum description length principle. In the second stage, to account for the intrinsic geometric structure of samples embedded in each subspace, MISC performs graph regularized semi-nonnegative matrix factorization to explore clusters. It additionally integrates the kernel trick into matrix factorization to handle non-linearly separable clusters. Experimental results on synthetic datasets show that MISC can find different interesting clusterings from the sought independent subspaces, and it also outperforms other related and competitive approaches on real-world datasets.

cs.LG↗

arXiv2019

Dark Matter Blind Spots at One-Loop

We evaluate the impact of one-loop electroweak corrections to the spin-independent dark matter (DM) scattering cross-section with nucleons ($σ_{\rm SI}$), in models with a so-called blind spot for direct detection, where the leading-order prediction for the relevant DM coupling to the Higgs boson, and therefore $σ_{\rm SI}$, are vanishingly small. Adopting a simple illustrative scenario in which the DM state results from the mixing of electroweak singlet and doublet fermions, we compute the relevant higher order corrections to the scalar effective operator contributions to $σ_{\rm SI}$, stemming from both triangle and box diagrams involving the SM and dark sector fields. It is observed that in a significant region of the singlet-doublet model-space, the one-loop corrections ``unblind'' the tree-level blind spots and lead to detectable SI scattering rates at future multi-ton scale liquid Xenon experiments, with $σ_{\rm SI}$ reaching values up to a few times $10^{-47} {~\rm cm}^2$, for a weak scale DM with $\mathcal{O}(1)$ Yukawa couplings. Furthermore, we find that there always exists a new SI blind spot at the next-to-leading order, which is perturbatively shifted from the leading order one in the singlet-doublet mass parameters. For comparison, we also present the tree-level spin-dependent scattering cross-sections near the SI blind-spot region, that could lead to a larger signal. Our results can be mapped to the blind-spot scenario for bino-Higgsino DM in the MSSM, with other sfermions, the heavier Higgs boson, and the wino decoupled.

hep-ph↗

arXiv2019

Electroweak Dark Matter at Future Hadron Colliders

In a large class of scenarios, dark matter (DM) particles that belong to a multiplet of the standard model (SM) weak interactions are challenging to probe in direct detection experiments due to loop-suppressed cross-sections. Direct production at colliders is thus crucial to look for such DM candidates, and under current estimates, future runs of the 14-TeV LHC are projected to probe masses of around 300 GeV for DM belonging to an SU(2) doublet (Higgsino-like), and 900 GeV for SU(2) triplet (wino-like). We examine how far this mass reach can be extended at the proposed 27-TeV high-energy upgrade of the LHC (HE-LHC), and compare the results to the case for a 100-TeV hadron collider. Following a detector setup similar to that of the ATLAS tracking system for the Run-2 LHC upgrade, with a new Insertable B-Layer (IBL), a disappearing charged track analysis at the HE-LHC can probe Higgsino-like (wino-like) DM mass of up to 600 GeV (2.1 TeV) at the 95% C.L. The monojet and missing transverse momentum search, on the otherhand, has a weaker reach of 490 GeV (700 GeV) at 95% C.L. for the Higgsino-like (wino-like) states. The mass range accessible in the collider searches can be complementary to the indirect detection probes using gamma rays from dwarf-spheroidal galaxies.

hep-ph↗

arXiv2019

Modeling Recurrence for Transformer

Recently, the Transformer model that is based solely on attention mechanisms, has advanced the state-of-the-art on various machine translation tasks. However, recent studies reveal that the lack of recurrence hinders its further improvement of translation capacity. In response to this problem, we propose to directly model recurrence for Transformer with an additional recurrence encoder. In addition to the standard recurrent neural network, we introduce a novel attentive recurrent network to leverage the strengths of both attention and recurrent networks. Experimental results on the widely-used WMT14 English-German and WMT17 Chinese-English translation tasks demonstrate the effectiveness of the proposed approach. Our studies also reveal that the proposed model benefits from a short-cut that bridges the source and target sequences with a single recurrent layer, which outperforms its deep counterpart.

cs.CL↗

arXiv2019

Information Aggregation for Multi-Head Attention with Routing-by-Agreement

Multi-head attention is appealing for its ability to jointly extract different types of information from multiple representation subspaces. Concerning the information aggregation, a common practice is to use a concatenation followed by a linear transformation, which may not fully exploit the expressiveness of multi-head attention. In this work, we propose to improve the information aggregation for multi-head attention with a more powerful routing-by-agreement algorithm. Specifically, the routing algorithm iteratively updates the proportion of how much a part (i.e. the distinct information learned from a specific subspace) should be assigned to a whole (i.e. the final output representation), based on the agreement between parts and wholes. Experimental results on linguistic probing tasks and machine translation tasks prove the superiority of the advanced information aggregation over the standard linear transformation.

cs.CL↗

arXiv2019

On unconventional fall of rainbow spring

In this study we experimentally show that a stretched rainbow spring under gravity or extra weight may exhibit unconventional fall motion. Specially, when the rainbow spring is released from a high place, its lower end remains stationary until the spring wires stack together and then all the parts of the rainbow spring falls down together. We utilize a high-speed camera to record the fall process of one plastic and one metal rainbow spring under different loading conditions to systematically investigate this unconventional physical phenomenon. We use the time of the elastic wave propagating the length of the spring to predict the duration of the lower end of the rainbow spring remaining still. The findings from this study elucidate this physical phenomenon which has potential for the areas requiring temporary absolute space positioning.

physics.class-ph↗

arXiv2019

Context-Aware Self-Attention Networks

Self-attention model have shown its flexibility in parallel computation and the effectiveness on modeling both long- and short-term dependencies. However, it calculates the dependencies between representations without considering the contextual information, which have proven useful for modeling dependencies among neural representations in various natural language tasks. In this work, we focus on improving self-attention networks through capturing the richness of context. To maintain the simplicity and flexibility of the self-attention networks, we propose to contextualize the transformations of the query and key layers, which are used to calculates the relevance between elements. Specifically, we leverage the internal representations that embed both global and deep contexts, thus avoid relying on external resources. Experimental results on WMT14 English-German and WMT17 Chinese-English translation tasks demonstrate the effectiveness and universality of the proposed methods. Furthermore, we conducted extensive analyses to quantity how the context vectors participate in the self-attention model.

cs.CL↗

arXiv2019

Dynamic Layer Aggregation for Neural Machine Translation with Routing-by-Agreement

With the promising progress of deep neural networks, layer aggregation has been used to fuse information across layers in various fields, such as computer vision and machine translation. However, most of the previous methods combine layers in a static fashion in that their aggregation strategy is independent of specific hidden states. Inspired by recent progress on capsule networks, in this paper we propose to use routing-by-agreement strategies to aggregate layers dynamically. Specifically, the algorithm learns the probability of a part (individual layer representations) assigned to a whole (aggregated representations) in an iterative way and combines parts accordingly. We implement our algorithm on top of the state-of-the-art neural machine translation model TRANSFORMER and conduct experiments on the widely-used WMT14 English-German and WMT17 Chinese-English translation datasets. Experimental results across language pairs show that the proposed approach consistently outperforms the strong baseline model and a representative static aggregation model.

cs.CL↗

arXiv2018

Learning to Refine Source Representations for Neural Machine Translation

Neural machine translation (NMT) models generally adopt an encoder-decoder architecture for modeling the entire translation process. The encoder summarizes the representation of input sentence from scratch, which is potentially a problem if the sentence is ambiguous. When translating a text, humans often create an initial understanding of the source sentence and then incrementally refine it along the translation on the target side. Starting from this intuition, we propose a novel encoder-refiner-decoder framework, which dynamically refines the source representations based on the generated target-side information at each decoding step. Since the refining operations are time-consuming, we propose a strategy, leveraging the power of reinforcement learning models, to decide when to refine at specific decoding steps. Experimental results on both Chinese-English and English-German translation tasks show that the proposed approach significantly and consistently improves translation performance over the standard encoder-decoder framework. Furthermore, when refining strategy is applied, results still show reasonable improvement over the baseline without much decrease in decoding speed.

cs.CL↗

arXiv2018

Charm-quark Yukawa Coupling in $h\rightarrow c\bar{c}γ$ at LHC

It is extremely challenging to probe the charm-quark Yukawa coupling at hadron colliders primarily due to the large Standard Model (SM) background (including $h\to b\bar b$) and the lack of an effective trigger for the signal $h\to c\bar c$. We examine the feasibility of probing this coupling at the LHC via a Higgs radiative decay $h\rightarrow c\bar{c}γ$. The existence of an additional photon in the final state may help for the signal identification and background suppression. Adopting a refined triggering strategy and utilizing basic machine learning, we find that a coupling limit of about 8 times the SM value may be reached with $2σ$ sensitivity after the High Luminosity LHC (HL-LHC). Our result is comparable and complementary to other projections for direct and indirect probes of $h\to c\bar c$ at the HL-LHC. Without a significant change in detector capabilities, there would be no significant improvement for this search from higher energy hadron colliders.

hep-ph↗

arXiv2018

Resummation for rapidity distributions in top-quark pair production

We extend our framework for the simultaneous resummation of soft and small-mass logarithms to rapidity distributions in top quark pair production. We give numerical results for the rapidity distribution of the top quark or the anti-top quark, as well as the rapidity distribution of the $t\bar{t}$ pair, finding that resummation effects stabilize the dependence of the differential cross sections on the choice of factorization scale. We compare our results with recent measurements at the Large Hadron Collider and find good agreement. Our results may be useful in the extraction of the gluon parton distribution function from $t\bar{t}$ production.

hep-ph↗

arXiv2018

A Deep Dive into Blockchain Selfish Mining

This paper studies a fundamental problem regarding the security of blockchain on how the existence of multiple misbehaving pools influences the profitability of selfish mining. Each selfish miner maintains a private chain and makes it public opportunistically for the purpose of acquiring more rewards incommensurate to his Hashrate. We establish a novel Markov chain model to characterize all the state transitions of public and private chains. The minimum requirement of Hashrate together with the minimum delay of being profitable is derived in close-form. The former reduces to 21.48% with the symmetric selfish miners, while their competition with asymmetric Hashrates puts forward a higher requirement of the profitable threshold. The profitable delay increases with the decrease of the Hashrate of selfish miners, making the mining pools more cautious on performing selfish mining.

cs.CR↗

arXiv2018

Exploiting Deep Representations for Neural Machine Translation

Advanced neural machine translation (NMT) models generally implement encoder and decoder as multiple layers, which allows systems to model complex functions and capture complicated linguistic structures. However, only the top layers of encoder and decoder are leveraged in the subsequent process, which misses the opportunity to exploit the useful information embedded in other layers. In this work, we propose to simultaneously expose all of these signals with layer aggregation and multi-layer attention mechanisms. In addition, we introduce an auxiliary regularization term to encourage different layers to capture diverse information. Experimental results on widely-used WMT14 English-German and WMT17 Chinese-English translation data demonstrate the effectiveness and universality of the proposed approach.

cs.CL↗

arXiv2018

Network Modeling and Pathway Inference from Incomplete Data ("PathInf")

In this work, we developed a network inference method from incomplete data ("PathInf") , as massive and non-uniformly distributed missing values is a common challenge in practical problems. PathInf is a two-stages inference model. In the first stage, it applies a data summarization model based on maximum likelihood to deal with the massive distributed missing values by transforming the observation-wise items in the data into state matrix. In the second stage, transition pattern (i.e. pathway) among variables is inferred as a graph inference problem solved by greedy algorithm with constraints. The proposed method was validated and compared with the state-of-art Bayesian network method on the simulation data, and shown consistently superior performance. By applying the PathInf on the lymph vascular metastasis data, we obtained the holistic pathways of the lymph node metastasis with novel discoveries on the jumping metastasis among nodes that are physically apart. The discovery indicates the possible presence of sentinel node groups in the lung lymph nodes which have been previously speculated yet never found. The pathway map can also improve the current dissection examination protocol for better individualized treatment planning, for higher diagnostic accuracy and reducing the patients trauma.

stat.ML↗

arXiv2018

Resummation for (boosted) top-quark pair production at NNLO+NNLL' in QCD

We construct predictions for top quark pair differential distributions at hadron colliders that combine state-of-the-art NNLO QCD calculations with double resummation at NNLL' accuracy of threshold logarithms arising from soft gluon emissions and of small mass logarithms. This is the first time a resummed calculation at full NNLO+NNLL' accuracy in QCD for a process with non-trivial color structure has been completed at the differential level. Of main interest to us is the stability of the M_{tt} and top-quark p_T distributions in the boosted regime where fixed order calculations may become strongly dependent on the choice of dynamic scales. With the help of numeric and analytic arguments we confirm that the choice for the factorization and renormalization scales advocated recently by some of the authors is indeed optimal. We further derive a set of optimized kinematics-dependent scales for the matching functions which appear in the resummed calculations. Our NNLO+NNLL' prediction for the top-pair invariant mass is significantly less sensitive to the choice of factorization scale than the fixed order prediction, even at NNLO. Notably, the resummed and fixed order calculations are in nearly perfect agreement with each other in the full M_{tt} range when the optimal dynamic scale is used. For the top-quark p_T distribution the resummation performed here has less of an impact and instead we find that upgrading the matching with fixed-order from NLO+NNLL' to NNLO+NNLL' to be an important effect, a point to be kept in mind when using NLO-based Monte Carlo event generators to calculate this distribution.

hep-ph↗