arXiv subjects

Xiang Li

Publications and source records attributed to Xiang Li.

At least 1,405 records · Page 78Linked to original sources

arXiv2016

Two are Better than One: An Ensemble of Retrieval- and Generation-Based Dialog Systems

Open-domain human-computer conversation has attracted much attention in the field of NLP. Contrary to rule- or template-based domain-specific dialog systems, open-domain conversation usually requires data-driven approaches, which can be roughly divided into two categories: retrieval-based and generation-based systems. Retrieval systems search a user-issued utterance (called a query) in a large database, and return a reply that best matches the query. Generative approaches, typically based on recurrent neural networks (RNNs), can synthesize new replies, but they suffer from the problem of generating short, meaningless utterances. In this paper, we propose a novel ensemble of retrieval-based and generation-based dialog systems in the open domain. In our approach, the retrieved candidate, in addition to the original query, is fed to an RNN-based reply generator, so that the neural model is aware of more information. The generated reply is then fed back as a new candidate for post-reranking. Experimental results show that such ensemble outperforms each single part of it by a large margin.

cs.CL↗

arXiv2016

Statistical Properties of the Single Linkage Hierarchical Clustering Estimator

Distance-based hierarchical clustering (HC) methods are widely used in unsupervised data analysis but few authors take account of uncertainty in the distance data. We incorporate a statistical model of the uncertainty through corruption or noise in the pairwise distances and investigate the problem of estimating the HC as unknown parameters from measurements. Specifically, we focus on single linkage hierarchical clustering (SLHC) and study its geometry. We prove that under fairly reasonable conditions on the probability distribution governing measurements, SLHC is equivalent to maximum partial profile likelihood estimation (MPPLE) with some of the information contained in the data ignored. At the same time, we show that direct evaluation of SLHC on maximum likelihood estimation (MLE) of pairwise distances yields a consistent estimator. Consequently, a full MLE is expected to perform better than SLHC in getting the correct HC results for the ground truth metric.

stat.ML↗

arXiv2016

Search for gamma-ray line features from Milky Way satellites with Fermi-LAT Pass 8 data

With 91 months of the publicly available Fermi-LAT Pass 8 data, we analyze the gamma-ray emission from the Milky Way satellites to search for potential line signals due to the annihilation of dark matter particles into double photons. The searched targets include a sample of dwarf spheroidal galaxies, the Large Magellanic Cloud (LMC) and Small Magellanic Cloud (SMC). No significant line emission has been found neither in the stacked dwarf galaxy sample nor in the direction of LMC/SMC. The corresponding upper limits on the cross section of DM annihilation into two photons are derived. Compared with results of previous gamma-ray line searches with the Pass 7 data, the current constraints on the line emission from dwarf spheroidal galaxies has been significantly improved in a wide energy range. With the rapid increase of the sample of dwarf spheroidal galaxies (candidates), we expect that the sensitivity of gamma ray line searches will be significantly improved in the near future.

astro-ph.HE↗

arXiv2016

The Macronova in GRB 050709 and the GRB/macronova connection

GRB 050709 was the first short Gamma-ray Burst (sGRB) with an identified optical afterglow. In this work we report a re-analysis of the publicly available data of this event and the discovery of a Li-Paczynski macronova/kilonova that dominates the optical/IR signal at t $>$ 2.5 days. Such a signal would arise from ~0.05 M$_\odot$ r-process material launched by a compact binary merger. The implied mass ejection supports the suggestion that compact binary mergers are significant and possibly main sites of heavy r-process nucleosynthesis. We have re-analyzed all afterglow data from nearby short and hybrid GRBs. A statistical study of sGRB/macronova connection reveals that macronova may have taken place in all these GRBs though the fraction as low as 0.18 cannot be ruled out. The identification of two of the three macronova candidates in the I-band implies a more promising detection prospect for ground-based surveys.

astro-ph.HE↗

arXiv2016

GRB/GW association: Long-short GRB candidates, time-lag, measuring gravitational wave velocity and testing Einstein's equivalence principle

Short-duration gamma-ray bursts (SGRBs) are widely believed to be powered by the mergers of compact binaries, such as binary neutron stars or possibly neutron star-black hole binaries. Though the prospect of detecting SGRBs with gravitational wave (GW) signals by the advanced Laser Interferometer Gravitational-Wave Observatory (LIGO)/VIRGO network is promising, no known SGRB has been found within the expected advanced LIGO/VIRGO sensitivity range for binary neutron star systems. We find, however, that the two long-short GRBs (GRB 060505 and GRB 060614) may be within the horizon of advanced GW detectors. In the upcoming era of GW astronomy, the merger origin of some long-short GRBs, as favored by the macronova signature displayed in GRB 060614, can be unambiguously tested. The model-dependent time lags between the merger and the onset of the prompt emission of the GRB are estimated. The comparison of such time lags between model predictions and the real data expected in the era of the GW astronomy would be helpful in revealing the physical processes taking place at the central engine (including the launch of the relativistic outflow, the emergence of the outflow from the dense material ejected during the merger, and the radiation of gamma rays). We also show that the speed of GWs, with or without a simultaneous test of Einstein's equivalence principle, can be directly measured to an accuracy of $\sim 3\times 10^{-8}~{\rm cm~s^{-1}}$ or even better in the advanced LIGO/VIRGO era. The Astrophysical Journal, Volume

astro-ph.HE↗

arXiv2016

Implications of the Tentative Association between GW150914 and a {\it Fermi}-GBM Transient

The merger-driven Gamma-ray Bursts (GRBs) and their associated gravitational wave (GW) radiation, if both successfully detected, have some far-reaching implications, including for instance: (i) The statistical comparison of the physical properties of the short/long-short GRBs with and without GW detection can test the general origin model; (ii) Revealing the physical processes taking place at the central engine; (iii) Measuring the velocity of the Gravitational wave directly/accurately. In this work we discuss these implications in the case of possible association of GW150914/ GBM transient 150914. We compared GBM transient 150914 with other SGRBs and found that such an event {may be} a distinct outlier in some statistical diagrams, possibly due to its specific binary-black-hole merger origin. However, the presence of a "new" group of SGRBs with "unusual" physical parameters is also possible. If the outflow of GBM transient 150914 was launched by the accretion onto the nascent black hole, the magnetic activity rather than the neutrino process is likely responsible for the energy extraction and the accretion disk mass is estimated to be $\sim 10^{-5}~M_\odot$. The GW150914/GBM transient 150914 association, {if confirmed, would} provide the first opportunity to directly measure the GW velocity and its departure from the speed of the light {should be within} a factor of $\sim 10^{-17}$.

astro-ph.HE↗

arXiv2016

Magnons and Phonons Optically Driven Out of Local Equilibrium in a Magnetic Insulator

Magnons are the energy quanta of fundamental spin excitations, namely spin waves, and they can make a considerable contribution to energy transport in some magnetic materials in a similar manner as lattice vibration waves or phonons. The coupling and possible non-equilibrium between magnons and other energy carriers have been used to explain several recently discovered thermally driven spin transport and energy conversion phenomena. Here, we report experiments in which local non-equilibrium between magnons and phonons in a single crystalline bulk magnetic insulator, Y3Fe5O12 (yttrium iron garnet, or YIG), has been created optically within a focused laser spot and probed directly with the use of micro-Brillouin light scattering (BLS). By analyzing the experimental results with a thermally induced magnon diffusion model, we obtain the magnon diffusion length of thermal magnons. By explicitly establishing non-equilibrium between magnons and phonons, our studies represent an important step toward a quantitative understanding of various spin-heat coupling phenomena.

cond-mat.mtrl-sci↗

arXiv2016

Top-push Video-based Person Re-identification

Most existing person re-identification (re-id) models focus on matching still person images across disjoint camera views. Since only limited information can be exploited from still images, it is hard (if not impossible) to overcome the occlusion, pose and camera-view change, and lighting variation problems. In comparison, video-based re-id methods can utilize extra space-time information, which contains much more rich cues for matching to overcome the mentioned problems. However, we find that when using video-based representation, some inter-class difference can be much more obscure than the one when using still-image based representation, because different people could not only have similar appearance but also have similar motions and actions which are hard to align. To solve this problem, we propose a top-push distance learning model (TDL), in which we integrate a top-push constrain for matching video features of persons. The top-push constraint enforces the optimization on top-rank matching in re-id, so as to make the matching model more effective towards selecting more discriminative features to distinguish different persons. Our experiments show that the proposed video-based re-id framework outperforms the state-of-the-art video-based re-id methods.

cs.CV↗

arXiv2016

Search for a gamma-ray line feature from a group of nearby Galaxy clusters with Fermi LAT Pass 8 data

Galaxy clusters are the largest gravitationally bound objects in the universe and may be suitable targets for indirect dark matter searches. With 85 months of Fermi-LAT Pass 8 publicly available data, we analyze the gamma-ray emission in the directions of 16 nearby Galaxy Clusters with an unbinned likelihood analysis. No globally statistically-significant $γ-$ray line feature is identified and a tentative line signal may be present at $\sim 43$ GeV. The 95\% confidence level upper limits on the velocity-averaged cross section of dark matter particles annihilating into double $γ-$rays (i.e., $\langle σv \rangle_{χχ\rightarrow γγ}$) are derived. Unless very optimistic boost factors of dark matter annihilation in these Galaxy Clusters have been assumed, such constraints are much weaker than the bounds set by the Galactic $γ-$ray data.

astro-ph.HE↗

arXiv2016

An Enhanced Deep Feature Representation for Person Re-identification

Feature representation and metric learning are two critical components in person re-identification models. In this paper, we focus on the feature representation and claim that hand-crafted histogram features can be complementary to Convolutional Neural Network (CNN) features. We propose a novel feature extraction model called Feature Fusion Net (FFN) for pedestrian image representation. In FFN, back propagation makes CNN features constrained by the handcrafted features. Utilizing color histogram features (RGB, HSV, YCbCr, Lab and YIQ) and texture features (multi-scale and multi-orientation Gabor features), we get a new deep feature representation that is more discriminative and compact. Experiments on three challenging datasets (VIPeR, CUHK01, PRID450s) validates the effectiveness of our proposal.

cs.CV↗

arXiv2016

StalemateBreaker: A Proactive Content-Introducing Approach to Automatic Human-Computer Conversation

Existing open-domain human-computer conversation systems are typically passive: they either synthesize or retrieve a reply provided a human-issued utterance. It is generally presumed that humans should take the role to lead the conversation and introduce new content when a stalemate occurs, and that the computer only needs to "respond." In this paper, we propose StalemateBreaker, a conversation system that can proactively introduce new content when appropriate. We design a pipeline to determine when, what, and how to introduce new content during human-computer conversation. We further propose a novel reranking algorithm Bi-PageRank-HITS to enable rich interaction between conversation context and candidate replies. Experiments show that both the content-introducing approach and the reranking algorithm are effective. Our full StalemateBreaker model outperforms a state-of-the-practice conversation system by +14.4% p@1 when a stalemate occurs.

cs.CL↗

arXiv2016

Explore Spatiotemporal and Demographic Characteristics of Human Mobility via Twitter: A Case Study of Chicago

Characterizing human mobility patterns is essential for understanding human behaviors and the interactions with socioeconomic and natural environment. With the continuing advancement of location and Web 2.0 technologies, location-based social media (LBSM) have been gaining widespread popularity in the past few years. With an access to locations of users, profiles and the contents of the social media posts, the LBSM data provided a novel modality of data source for human mobility study. By exploiting the explicit location footprints and mining the latent demographic information implied in the LBSM data, the purpose of this paper is to investigate the spatiotemporal characteristics of human mobility with a particular focus on the impact of demography. We first collect geo-tagged Twitter feeds posted in the conterminous United States area, and organize the collection of feeds using the concept of space-time trajectory corresponding to each Twitter user. Commonly human mobility measures, including detected home and activity centers, are derived for each user trajectory. We then select a subset of Twitter users that have detected home locations in the city of Chicago as a case study, and apply name analysis to the names provided in user profiles to learn the implicit demographic information of Twitter users, including race/ethnicity, gender and age. Finally we explore the spatiotemporal distribution and mobility characteristics of Chicago Twitter users, and investigate the demographic impact by comparing the differences across three demographic dimensions (race/ethnicity, gender and age). We found that, although the human mobility measures of different demographic groups generally follow the generic laws (e.g., power law distribution), the demographic information, particular the race/ethnicity group, significantly affects the urban human mobility patterns.

cs.SI↗

arXiv2016

Network Clustering via Maximizing Modularity: Approximation Algorithms and Theoretical Limits

Many social networks and complex systems are found to be naturally divided into clusters of densely connected nodes, known as community structure (CS). Finding CS is one of fundamental yet challenging topics in network science. One of the most popular classes of methods for this problem is to maximize Newman's modularity. However, there is a little understood on how well we can approximate the maximum modularity as well as the implications of finding community structure with provable guarantees. In this paper, we settle definitely the approximability of modularity clustering, proving that approximating the problem within any (multiplicative) positive factor is intractable, unless P = NP. Yet we propose the first additive approximation algorithm for modularity clustering with a constant factor. Moreover, we provide a rigorous proof that a CS with modularity arbitrary close to maximum modularity QOPT might bear no similarity to the optimal CS of maximum modularity. Thus even when CS with near-optimal modularity are found, other verification methods are needed to confirm the significance of the structure.

cs.SI↗

arXiv2016

Search for gamma-ray emission from eight dwarf spheroidal galaxy candidates discovered in Year Two of Dark Energy Survey with Fermi-LAT data

Very recently the Dark Energy Survey (DES) Collaboration has released their second group of Dwarf spheroidal (dSph) galaxy candidates. With the publicly-available Pass 8 data of Fermi-LAT we search for $γ-$ray emissions from the directions of these eight newly discovered dSph galaxy candidates. No statistically significant $γ-$ray signal has been found in the combined analysis of these sources. With the empirically estimated J-factors of these sources, the constraint on the annihilation channel of $χχ\rightarrow τ^{+}τ^{-}$ is comparable to the bound set by the joint analysis of fifteen previously known dSphs with kinematically constrained J-factors for the dark matter mass $m_χ>250$ GeV. In the direction of Tucana III (DES J2356-5935), one of the nearest dSph galaxy candidates that is $\sim 25$ kpc away, there is a weak $γ-$ray signal and its peak test statistic (TS) value for the dark matter annihilation channel $χχ\rightarrow τ^{+}τ^{-1}$ is $\approx 6.7$ at $m_χ\sim 15$ GeV. The significance of the possible signal likely increases with time. More data is highly needed to pin down the physical origin of such a GeV excess.

astro-ph.HE↗

arXiv2016

A GeV source in the direction of Supernova Remnant CTB 37B

Supernova remnants (SNRs) are the most attractive candidates for the acceleration sites of Galactic cosmic rays. We report the detection of GeV $γ$-ray emission with the Pass 8 events recorded by Fermi Large Area Telescope (Fermi-LAT) in the vicinity of the shell type SNR CTB 37B that is likely associated with the TeV $γ-$ray source HESS J1713-381. The photon spectrum of CTB 37B is consistent with a power-law with an index of $1.89\pm0.08$ in the energy range of $0.5-500$ GeV, and the measured flux connects smoothly with that of HESS J1713-381 at a few hundred GeV. No significant spatial extension and time variation are detected. The multi-wavelength data can be well fitted with either a leptonic model or a hadronic one. However, parameters of both models suggest more efficient particle acceleration than typical SNRs. Meanwhile, the X-ray and $γ$-ray spectral properties of CTB 37B show that it is an interesting source bridging young SNRs dominated by non-thermal emission and old SNRs interacting with molecular clouds.

astro-ph.HE↗

arXiv2015

Maximum Likelihood Estimation for Single Linkage Hierarchical Clustering

We derive a statistical model for estimation of a dendrogram from single linkage hierarchical clustering (SLHC) that takes account of uncertainty through noise or corruption in the measurements of separation of data. Our focus is on just the estimation of the hierarchy of partitions afforded by the dendrogram, rather than the heights in the latter. The concept of estimating this "dendrogram structure'' is introduced, and an approximate maximum likelihood estimator (MLE) for the dendrogram structure is described. These ideas are illustrated by a simple Monte Carlo simulation that, at least for small data sets, suggests the method outperforms SLHC in the presence of noise.

stat.ML↗

arXiv2015

Identifying spatial invasion of pandemics on metapopulation networks via anatomizing arrival history

Spatial spread of infectious diseases among populations via the mobility of humans is highly stochastic and heterogeneous. Accurate forecast/mining of the spread process is often hard to be achieved by using statistical or mechanical models. Here we propose a new reverse problem, which aims to identify the stochastically spatial spread process itself from observable information regarding the arrival history of infectious cases in each subpopulation. We solved the problem by developing an efficient optimization algorithm based on dynamical programming, which comprises three procedures: i, anatomizing the whole spread process among all subpopulations into disjoint componential patches; ii, inferring the most probable invasion pathways underlying each patch via maximum likelihood estimation; iii, recovering the whole process by assembling the invasion pathways in each patch iteratively, without burdens in parameter calibrations and computer simulations. Based on the entropy theory, we introduced an identifiability measure to assess the difficulty level that an invasion pathway can be identified. Results on both artificial and empirical metapopulation networks show the robust performance in identifying actual invasion pathways driving pandemic spread.

cs.SI↗

arXiv2015

A possible Macronova in the late afterglow of the `long-short' burst GRB 060614

Long-duration ($>2$ s) $γ$-ray bursts that are believed to originate from the death of massive stars are expected to be accompanied by supernovae. GRB 060614, that lasted 102 s, lacks a supernova-like emission down to very stringent limits and its physical origin is still debated. Here we report the discovery of near-infrared bump that is significantly above the regular decaying afterglow. This red bump is inconsistent with even the weakest known supernova. However, it can arise from a Li-Paczyński macronova $-$ the radioactive decay of debris following a compact binary merger. If this interpretation is correct GRB 060614 arose from a compact binary merger rather than from the death of a massive star and it was a site of a significant production of heavy r-process elements. The significant ejected mass favors a black hole-neutron star merger but a double neutron star merger cannot be ruled out.

astro-ph.HE↗