arXiv subjects
Publications and source records attributed to Xi Zhang.
In this paper, we analyze the asymptotic behaviour of the Hermitian-Yang-Mills flow over a compact non-Kähler manifold $(X,g)$ with the Hermitian metric $g$ satisfying the Gauduchon and Astheno-Kähler condition.
In networks, multiple contagions, such as information and purchasing behaviors, may interact with each other as they spread simultaneously. However, most of the existing information diffusion models are built on the assumption that each individual contagion spreads independently, regardless of their interactions. Gaining insights into such interaction is crucial to understand the contagion adoption behaviors, and thus can make better predictions. In this paper, we study the contagion adoption behavior under a set of interactions, specifically, the interactions among users, contagions' contents and sentiments, which are learned from social network structures and texts. We then develop an effective and efficient interaction-aware diffusion (IAD) framework, incorporating these interactions into a unified model. We also present a generative process to distinguish user roles, a co-training method to determine contagions' categories and a new topic model to obtain topic-specific sentiments. Evaluation on large-scale Weibo dataset demonstrates that our proposal can learn how different users, contagion categories and sentiments interact with each other efficiently. With these interactions, we can make a more accurate prediction than the state-of-art baselines. Moreover, we can better understand how the interactions influence the propagation process and thus can suggest useful directions for information promotion or suppression in viral marketing.
Generating images from word descriptions is a challenging task. Generative adversarial networks(GANs) are shown to be able to generate realistic images of real-life objects. In this paper, we propose a new neural network architecture of LSTM Conditional Generative Adversarial Networks to generate images of real-life objects. Our proposed model is trained on the Oxford-102 Flowers and Caltech-UCSD Birds-200-2011 datasets. We demonstrate that our proposed model produces the better results surpassing other state-of-art approaches.
A High-performance Atmospheric Radiation Package (HARP) is developed for studying multiple-scattering planetary atmospheres. HARP is an open-source program written in C++ that utilizes high-level data structure and parallel-computing algorithms. It is generic in three aspects. First, the construction of the model atmospheric profile is generic. The program can either take in an atmospheric profile or construct an adiabatic thermal and compositional profile, taking into account the clouds and latent heat release due to condensation. Second, the calculation of opacity is generic, based on line-by-line molecular transitions and tabulated continuum data, along with a table of correlated-k opacity provided as an option to speed up the calculation of energy fluxes. Third, the selection of the solver for the radiative transfer equation is generic. The solver is not hardwired in the program. Instead, based on the purpose, a variety of radiative transfer solvers can be chosen to couple with the atmosphere model and the opacity model. We use the program to investigate the radiative heating and cooling rates of all four giant planets in the Solar System. Our Jupiter's result is consistent with previous publications. Saturn has a nearly perfect balance between the heating rate and cooling rate. Uranus has the least radiative fluxes because of the lack of CH4 and its photochemical products. Both Uranus and Neptune suffer from a severe energy deficit in their stratospheres. Possible ways to resolve this issue are discussed. Finally, we recalculate the radiative time constants of all four giant planet atmospheres and find that the traditional values from (Conrath BJ, Gierasch PJ, Leroy SS. Temperature and Circulation in the Stratosphere of the Outer Planets. Icar. 1990;83:255-81) are significantly overestimated.
With the success of the Cyber Grand Challenge (CGC) sponsored by DARPA, the topic of Autonomous Cyber Reasoning System (CRS) has recently attracted extensive attention from both industry and academia. Utilizing automated system to detect, exploit and patch software vulnerabilities seems so attractive because of its scalability and cost-efficiency compared with the human expert based solution. In this paper, we give an extensive survey of former representative works related to the underlying technologies of a CRS, including vulnerability detection, exploitation and patching. As an important supplement, we then review several pioneer studies that explore the potential of machine learning technologies in this field, and point out that the future development of Autonomous CRS is inseparable from machine learning.
In this paper, we prove a Liouville theorem for holomorphic functions on a class of complete Gauduchon manifolds. This generalizes a result of Yau for complete Kähler manifolds to the complete non-Kähler case.
The investment on the stock market is prone to be affected by the Internet. For the purpose of improving the prediction accuracy, we propose a multi-task stock prediction model that not only considers the stock correlations but also supports multi-source data fusion. Our proposed model first utilizes tensor to integrate the multi-sourced data, including financial Web news, investors' sentiments extracted from the social network and some quantitative data on stocks. In this way, the intrinsic relationships among different information sources can be captured, and meanwhile, multi-sourced information can be complemented to solve the data sparsity problem. Secondly, we propose an improved sub-mode coordinate algorithm (SMC). SMC is based on the stock similarity, aiming to reduce the variance of their subspace in each dimension produced by the tensor decomposition. The algorithm is able to improve the quality of the input features, and thus improves the prediction accuracy. And the paper utilizes the Long Short-Term Memory (LSTM) neural network model to predict the stock fluctuation trends. Finally, the experiments on 78 A-share stocks in CSI 100 and thirteen popular HK stocks in the year 2015 and 2016 are conducted. The results demonstrate the improvement on the prediction accuracy and the effectiveness of the proposed model.
Using a layered representation for motion estimation has the advantage of being able to cope with discontinuities and occlusions. In this paper, we learn to estimate optical flow by combining a layered motion representation with deep learning. Instead of pre-segmenting the image to layers, the proposed approach automatically generates a layered representation of optical flow using the proposed soft-mask module. The essential components of the soft-mask module are maxout and fuse operations, which enable a disjoint layered representation of optical flow and more accurate flow estimation. We show that by using masks the motion estimate results in a quadratic function of input features in the output layer. The proposed soft-mask module can be added to any existing optical flow estimation networks by replacing their flow output layer. In this work, we use FlowNet as the base network to which we add the soft-mask module. The resulting network is tested on three well-known benchmarks with both supervised and unsupervised flow estimation tasks. Evaluation results show that the proposed network achieve better results compared with the original FlowNet.
We present the first application of a bin-scheme microphysical and vertical transport model to determine the size distribution of titanium and silicate cloud particles in the atmospheres of hot Jupiters. We predict particle size distributions from first principles for a grid of planets at four representative equatorial longitudes, and investigate how observed cloud properties depend on the atmospheric thermal structure and vertical mixing. The predicted size distributions are frequently bimodal and irregular in shape. There is a negative correlation between total cloud mass and equilibrium temperature as well as a positive correlation between total cloud mass and atmospheric mixing. The cloud properties on the east and west limbs show distinct differences that increase with increasing equilibrium temperature. Cloud opacities are roughly constant across a broad wavelength range with the exception of features in the mid-infrared. Forward scattering is found to be important across the same wavelength range. Using the fully resolved size distribution of cloud particles as opposed to a mean particle size has a distinct impact on the resultant cloud opacities. The particle size that contributes the most to the cloud opacity depends strongly on the cloud particle size distribution. We predict that it is unlikely that silicate or titanium clouds are responsible for the optical Rayleigh scattering slope seen in many hot Jupiters. We suggest that cloud opacities in emission may serve as sensitive tracers of the thermal state of a planet's deep interior through the existence or lack of a cold trap in the deep atmosphere.
In this paper, we prove a generalized Donaldson-Uhlenbeck-Yau theorem on Higgs bundles over a class of non-compact Gauduchon manifolds.
Recent works have shown that social media platforms are able to influence the trends of stock price movements. However, existing works have majorly focused on the U.S. stock market and lacked attention to certain emerging countries such as China, where retail investors dominate the market. In this regard, as retail investors are prone to be influenced by news or other social media, psychological and behavioral features extracted from social media platforms are thought to well predict stock price movements in the China's market. Recent advances in the investor social network in China enables the extraction of such features from web-scale data. In this paper, on the basis of tweets from Xueqiu, a popular Chinese Twitter-like social platform specialized for investors, we analyze features with regard to collective sentiment and perception on stock relatedness and predict stock price movements by employing nonlinear models. The features of interest prove to be effective in our experiments.
Traditional stock market prediction approaches commonly utilize the historical price-related data of the stocks to forecast their future trends. As the Web information grows, recently some works try to explore financial news to improve the prediction. Effective indicators, e.g., the events related to the stocks and the people's sentiments towards the market and stocks, have been proved to play important roles in the stocks' volatility, and are extracted to feed into the prediction models for improving the prediction accuracy. However, a major limitation of previous methods is that the indicators are obtained from only a single source whose reliability might be low, or from several data sources but their interactions and correlations among the multi-sourced data are largely ignored. In this work, we extract the events from Web news and the users' sentiments from social media, and investigate their joint impacts on the stock price movements via a coupled matrix and tensor factorization framework. Specifically, a tensor is firstly constructed to fuse heterogeneous data and capture the intrinsic relations among the events and the investors' sentiments. Due to the sparsity of the tensor, two auxiliary matrices, the stock quantitative feature matrix and the stock correlation matrix, are constructed and incorporated to assist the tensor decomposition. The intuition behind is that stocks that are highly correlated with each other tend to be affected by the same event. Thus, instead of conducting each stock prediction task separately and independently, we predict multiple correlated stocks simultaneously through their commonalities, which are enabled via sharing the collaboratively factorized low rank matrices between matrices and the tensor. Evaluations on the China A-share stock data and the HK stock data in the year 2015 demonstrate the effectiveness of the proposed model.
We demonstrate that hybrid laser wakefield and direct acceleration (LWDA) can be significantly improved by using two laser pulses with different polarizations or frequencies. The improvement entails higher energy and charge of the accelerated electrons, as well as weaker sensitivity to the time delay between the leading (pump) pulse responsible for generating the wakefield, and the trailing laser pulse responsible for direct laser acceleration (DLA). When the two laser pulses have the same frequency, it is shown that it is advantageous to have their polarization states orthogonal to each other. A specific scenario of a frequency-doubled ($λ_{\rm DLA}=λ_{\rm pump}/2$) DLA laser pulse is also considered, and found to be an improvement over the single-frequency ($λ_{\rm DLA} = λ_{\rm pump}$) scenario.
This paper presents a novel approach for lecture video indexing using a boosted deep convolutional neural network system. The indexing is performed by matching high quality slide images, for which text is either known or extracted, to lower resolution video frames with possible noise, perspective distortion, and occlusions. We propose a deep neural network integrated with a boosting framework composed of two sub-networks targeting feature extraction and similarity determination to perform the matching. The trained network is given as input a pair of slide image and a candidate video frame image and produces the similarity between them. A boosting framework is integrated into our proposed network during the training process. Experimental results show that the proposed approach is much more capable of handling occlusion, spatial transformations, and other types of noises when compared with known approaches.
As the rapid growth of multi-modal data, hashing methods for cross-modal retrieval have received considerable attention. Deep-networks-based cross-modal hashing methods are appealing as they can integrate feature learning and hash coding into end-to-end trainable frameworks. However, it is still challenging to find content similarities between different modalities of data due to the heterogeneity gap. To further address this problem, we propose an adversarial hashing network with attention mechanism to enhance the measurement of content similarities by selectively focusing on informative parts of multi-modal data. The proposed new adversarial network, HashGAN, consists of three building blocks: 1) the feature learning module to obtain feature representations, 2) the generative attention module to generate an attention mask, which is used to obtain the attended (foreground) and the unattended (background) feature representations, 3) the discriminative hash coding module to learn hash functions that preserve the similarities between different modalities. In our framework, the generative module and the discriminative module are trained in an adversarial way: the generator is learned to make the discriminator cannot preserve the similarities of multi-modal data w.r.t. the background feature representations, while the discriminator aims to preserve the similarities of multi-modal data w.r.t. both the foreground and the background feature representations. Extensive evaluations on several benchmark datasets demonstrate that the proposed HashGAN brings substantial improvements over other state-of-the-art cross-modal hashing methods.
This paper presents Aicyber's system for NLPCC 2017 shared task 2. It is formed by a voting of three deep learning based system trained on character-enhanced word vectors and a well known bag-of-word model.
In this paper, we prove a mean value formula for bounded subharmonic Hermitian matrix valued function on a complete Riemannian manifold with nonnegative Ricci curvature. As its application, we obtain a Liouville type theorem for the complex Monge-Ampère equation on product manifolds.
Energy harvesting emerges as a potential solution for prolonging the lifetime of the energy-constrained mobile wireless devices. In this paper, we focus on Radio Frequency (RF) energy harvesting for multiuser multicarrier mobile wireless networks. Specifically, we propose joint data and energy transfer optimization frameworks for powering mobile wireless devices through RF energy harvesting. We introduce a power utility that captures the power consumption cost at the base station (BS) and the used power from the users' batteries, and determine optimal power resource allocations that meet data rate requirements of downlink and uplink communications. Two types of harvesting capabilities are considered at each user: harvesting only from dedicated RF signals and hybrid harvesting from both dedicated and ambient RF signals. The developed frameworks increase the end users' battery lifetime at the cost of a slight increase in the BS power consumption. Several evaluation studies are conducted in order to validate our proposed frameworks.