arXiv subjects

Jing Li

Publications and source records attributed to Jing Li.

At least 649 records · Page 36Linked to original sources

arXiv2020

High-order numerical methods for the Riesz space fractional advection-dispersion equations

In this paper, we propose high-order numerical methods for the Riesz space fractional advection-dispersion equations (RSFADE) on a {f}inite domain. The RSFADE is obtained from the standard advection-dispersion equation by replacing the first-order and second-order space derivative with the Riesz fractional derivatives of order $α\in(0,1)$ and $β\in(1,2]$, respectively. Firstly, we utilize the weighted and shifted Grünwald difference operators to approximate the Riesz fractional derivative and present the {f}inite difference method for the RSFADE. Specifically, we discuss the Crank-Nicolson scheme and solve it in matrix form. Secondly, we prove that the scheme is unconditionally stable and convergent with the accuracy of $\mathcal {O}(τ^2+h^2)$. Thirdly, we use the Richardson extrapolation method (REM) to improve the convergence order which can be $\mathcal {O}(τ^4+h^4)$. Finally, some numerical examples are given to show the effectiveness of the numerical method, and the results are excellent with the theoretical analysis.

math.NA↗

arXiv2020

Secure Metric Learning via Differential Pairwise Privacy

Distance Metric Learning (DML) has drawn much attention over the last two decades. A number of previous works have shown that it performs well in measuring the similarities of individuals given a set of correctly labeled pairwise data by domain experts. These important and precisely-labeled pairwise data are often highly sensitive in real world (e.g., patients similarity). This paper studies, for the first time, how pairwise information can be leaked to attackers during distance metric learning, and develops differential pairwise privacy (DPP), generalizing the definition of standard differential privacy, for secure metric learning. Unlike traditional differential privacy which only applies to independent samples, thus cannot be used for pairwise data, DPP successfully deals with this problem by reformulating the worst case. Specifically, given the pairwise data, we reveal all the involved correlations among pairs in the constructed undirected graph. DPP is then formalized that defines what kind of DML algorithm is private to preserve pairwise data. After that, a case study employing the contrastive loss is exhibited to clarify the details of implementing a DPP-DML algorithm. Particularly, the sensitivity reduction technique is proposed to enhance the utility of the output distance metric. Experiments both on a toy dataset and benchmarks demonstrate that the proposed scheme achieves pairwise data privacy without compromising the output performance much (Accuracy declines less than 0.01 throughout all benchmark datasets when the privacy budget is set at 4).

cs.LG↗

arXiv2020

The Effective Potential Originating from Swampland and the Non-trivial Brans-Dicke Coupling

The effective vacuum energy density contributed by the non-trivial contortion distribution and the bare vacuum energy density can be viewed as the energy density of the auxiliary quintessence field potential. We find that the negative bare vacuum energy density from string landscape leads to a monotonically decreasing quintessence potential while the positive one from swampland leads to the meta stable or stable de Sitter like potential. Moreover, the non-trivial Brans-Dicke like coupling between quintessence field and gravitation field is necessary in the latter case.

gr-qc↗

arXiv2020

A Survey on Deep Learning for Named Entity Recognition

Named entity recognition (NER) is the task to identify mentions of rigid designators from text belonging to predefined semantic types such as person, location, organization etc. NER always serves as the foundation for many natural language applications such as question answering, text summarization, and machine translation. Early NER systems got a huge success in achieving good performance with the cost of human engineering in designing domain-specific features and rules. In recent years, deep learning, empowered by continuous real-valued vector representations and semantic composition through nonlinear processing, has been employed in NER systems, yielding stat-of-the-art performance. In this paper, we provide a comprehensive review on existing deep learning techniques for NER. We first introduce NER resources, including tagged NER corpora and off-the-shelf NER tools. Then, we systematically categorize existing works based on a taxonomy along three axes: distributed representations for input, context encoder, and tag decoder. Next, we survey the most representative methods for recent applied techniques of deep learning in new NER problem settings and applications. Finally, we present readers with the challenges faced by NER systems and outline future directions in this area.

cs.CL↗

arXiv2020

Unusual Intralayer Ferromagnetism Between S = 5/2 ions in MnBi$_2$Te$_4$: Role of Empty Bi $p$ States

The layered magnetic topological insulator MnBi$_2$Te$_4$ is a promising platform to realize the quantum anomalous Hall effect because its layers possess intrinsic ferromagnetism. However, it is not well understood why the high-spin $d^5$ magnetic ions Mn$^{2+}$ forming the Mn-Te-Mn spin exchange paths prefer ferromagnetic (FM) coupling, contrary to the prediction of the Goodenough-Kanamori rule that a TM-L-TM spin exchange, where TM and L are a transition-metal magnetic cation and a main group ligand, respectively, is antiferromagnetic (AFM) even when the bond angle of the exchange path is 90$^{\circ}$. Using density functional theory (DFT) calculations, we show that the presence of Bi$^{3+}$ ions is essential for the FM coupling in MnBi$_2$Te$_4$. Then, using a tight-binding model Hamiltonian, we find that high-spin $d^5$ ions (S = 5/2) in TM-L-TM spin exchange paths prefer FM coupling if the empty p-orbitals of a nonmagnetic cation M (e.g., Bi$^{3+}$ ion) hybridize strongly with those of the bridging ligand L, but AFM coupling otherwise.

physics.comp-ph↗

arXiv2020

Hubble Space Telescope Search for Activity in High Perihelion Objects

Solar system objects with perihelia beyond the orbit of Jupiter ($q >$ 5 AU) are too cold for water ice to generate an appreciable coma via sublimation. Despite this, numerous high perihelion objects (HPOs) including many comets and recently escaped Kuiper belt objects (``Centaurs'') are observed to be active out at least to the orbit of Saturn ($q \sim$ 10 AU). Peak equilibrium temperatures at 10 AU ($\sim$125 K), while far too low to sublimate water ice, are sufficient to sublimate super-volatiles such as CO and CO$_2$ ice. Temperatures at 10 AU are also high enough to trigger the rapid crystallization of exposed amorphous ice, thus constituting another possible driver of distant activity. While supervolatile ices can sublimate strongly (as $r_H^{-2}$) to at least Kuiper belt (30 AU) distances, crystallization is an exponential function of temperature that cannot be sustained much beyond $\sim$10 AU. The heliocentric dependence of the activity thus suggests an observational test. If activity in high perihelion objects is triggered by crystallization, then no examples of activity should be found with perihelia $q >>$ 10 AU. If, on the other hand, activity is due to free sublimation of exposed supervolatile ices, or another cause, then distant activity might be detected. We obtained sensitive, high resolution Hubble Space Telescope observations of HPOs to search for activity beyond the crystallization zone. No examples of activity were detected in 53 objects with $q >$ 15 AU, consistent with the crystallization trigger hypothesis. However, sensitivity limits are such that we cannot reject the alternative hypothesis that mass loss is driven by the sublimation of supervolatile ices. We also searched for binary companions in our sample, finding none and setting an empirical 3$σ$ limit to the binary fraction of $<8$\%.

astro-ph.EP↗

arXiv2020

Global existence, asymptotic behavior, and pattern formation driven by the parametrization of a nonlocal Fisher-KPP problem

The global boundedness and the hair trigger effect of solutions for the nonlinear nonlocal reaction-diffusion equation \begin{align*} u_t=Δu+μu^α(1-κJ*u^β),\quad\hbox{in} \;\mathbb R^N\times(0,\infty),\; N\geq 1 \end{align*} with $α\geq1$, $β,μ,κ>0$ and $u(x,0)=u_0(x)$ are investigated. Under appropriate assumptions on $J$, it is proved that for any nonnegative and bounded initial condition, if $α\in[1,α^*)$ with $α^*=1+β$ for $N=1,2$ and $α^*=1+\frac{2β}{N}$ for $N>2$, then the problem has a global bounded classical solution. Under further assumptions on the initial datum, the solutions satisfying $0\leq u(x,t)\leqκ^{-\frac1β}$ for any $(x,t)\in\mathbb R^N\times[0,+\infty)$ are shown to converge to $κ^{-\frac1β}$ uniformly on any compact subset of $\mathbb R^N$, which is known as the hair trigger effect. 1D numerical simulations of the above nonlocal reaction-diffusion equation are performed and the effect of several combinations of parameters and convolution kernels on the solution behavior is investigated. The results motivate a discussion about some conjectures arising from this model and further issues to be studied in this context. A formal deduction of the model from a mesoscopic formulation is provided as well.

math.AP↗

arXiv2020

UFTR: A Unified Framework for Ticket Routing

Corporations today face increasing demands for the timely and effective delivery of customer service. This creates the need for a robust and accurate automated solution to what is formally known as the ticket routing problem. This task is to match each unresolved service incident, or "ticket", to the right group of service experts. Existing studies divide the task into two independent subproblems - initial group assignment and inter-group transfer. However, our study addresses both subproblems jointly using an end-to-end modeling approach. We first performed a preliminary analysis of half a million archived tickets to uncover relevant features. Then, we devised the UFTR, a Unified Framework for Ticket Routing using four types of features (derived from tickets, groups, and their interactions). In our experiments, we implemented two ranking models with the UFTR. Our models outperform baselines on three routing metrics. Furthermore, a post-hoc analysis reveals that this superior performance can largely be attributed to the features that capture the associations between ticket assignment and group assignment. In short, our results demonstrate that the UFTR is a superior solution to the ticket routing problem because it takes into account previously unexploited interrelationships between the group assignment and group transfer problems.

cs.AI↗

arXiv2020

GPM: A Generic Probabilistic Model to Recover Annotator's Behavior and Ground Truth Labeling

In the big data era, data labeling can be obtained through crowdsourcing. Nevertheless, the obtained labels are generally noisy, unreliable or even adversarial. In this paper, we propose a probabilistic graphical annotation model to infer the underlying ground truth and annotator's behavior. To accommodate both discrete and continuous application scenarios (e.g., classifying scenes vs. rating videos on a Likert scale), the underlying ground truth is considered following a distribution rather than a single value. In this way, the reliable but potentially divergent opinions from "good" annotators can be recovered. The proposed model is able to identify whether an annotator has worked diligently towards the task during the labeling procedure, which could be used for further selection of qualified annotators. Our model has been tested on both simulated data and real-world data, where it always shows superior performance than the other state-of-the-art models in terms of accuracy and robustness.

cs.AI↗

arXiv2020

Efficient implementation of immersed boundary-lattice Boltzmann method for massive particle-laden flows Part I: Serial computing

Immersed boundary-lattice Boltzmann method (IB-LBM) has been widely used for simulation of particle-laden flows recently. However, it was limited to small-scale simulations with no more than O(103) particles. Here, we expand IB-LBM for massive particle-laden flows with more than O(104) particles by two sequential works. First is the Part I: serial computing on a single CPU core and following the Part II: parallel computing on many CPU cores. In this Part I paper, a highly efficient and localized implementation of IB-LBM is proposed for serial computing. We optimize in three main aspects: swap algorithm for incompressible LBM, local grid-to-point algorithm for IBM and improved grid search algorithm for particle pair short-range interaction. In addition, symmetry algorithm is proposed for the half-calculation of LB collision and external force term. The computational performance on a single CPU core is analyzed. Different scales of two dimensional (2D) and three-dimensional (3D) particles settling in closed cavities are used for testing. The solid volume fraction is varied from 0 to 0.40. Simulation results demonstrate that all calculation parts are dramatically decreased by the improved algorithm. For the particle-free flows, the Mega Lattice Site Update per Second (MLUPS) can be achieved up to 36 (2D) and 12 (3D) using the improved algorithm. For the particle-laden flows, MLUPS can be achieved no lower than 15 (2D) and 7 (3D) in the simulations of dense flows. At last, we discuss the potential of the new algorithms for the high-performance computation of the large-scale systems of particle-laden flows with MPI parallel technique.

physics.comp-ph↗

arXiv2020

What Changed Your Mind: The Roles of Dynamic Topics and Discourse in Argumentation Process

In our world with full of uncertainty, debates and argumentation contribute to the progress of science and society. Despite of the increasing attention to characterize human arguments, most progress made so far focus on the debate outcome, largely ignoring the dynamic patterns in argumentation processes. This paper presents a study that automatically analyzes the key factors in argument persuasiveness, beyond simply predicting who will persuade whom. Specifically, we propose a novel neural model that is able to dynamically track the changes of latent topics and discourse in argumentative conversations, allowing the investigation of their roles in influencing the outcomes of persuasion. Extensive experiments have been conducted on argumentative conversations on both social media and supreme court. The results show that our model outperforms state-of-the-art models in identifying persuasive arguments via explicitly exploring dynamic factors of topic and discourse. We further analyze the effects of topics and discourse on persuasiveness, and find that they are both useful - topics provide concrete evidence while superior discourse styles may bias participants, especially in social media arguments. In addition, we draw some findings from our empirical results, which will help people better engage in future persuasive conversations.

cs.CL↗

arXiv2020

Ground-state correlation energy of beryllium dimer by the Bethe-Salpeter equation

Since the '30s the interatomic potential of the beryllium dimer Be$_2$ has been both an experimental and a theoretical challenge. Calculating the ground-state correlation energy of Be$_2$ along its dissociation path is a difficult problem for theory. We present ab initio many-body perturbation theory calculations of the Be$_2$ interatomic potential using the GW approximation and the Bethe-Salpeter equation (BSE). The ground-state correlation energy is calculated by the trace formula with checks against the adiabatic-connection fluctuation-dissipation theorem formula. We show that inclusion of GW corrections already improves the energy even at the level of the random-phase approximation. At the level of the BSE on top of the GW approximation, our calculation is in surprising agreement with the most accurate theories and with experiment. It even reproduces an experimentally observed flattening of the interatomic potential due to a delicate correlations balance from a competition between covalent and van der Waals bonding.

physics.chem-ph↗

arXiv2020

Molecular dynamics simulation for cross-linking processes and material properties of epoxy resins with the first principle calculation combined with global reaction route mapping algorithms

Herein, epoxy resin is cured by coupling quantum chemical (QC) calculations with molecular dynamics (MD) simulations that enable parameter-free prediction of material characteristics. A polymer network is formed by the reaction between base resin and curing agent. The reaction uses activation energy and heat of formation data obtained by first-principle calculations coupled with global reaction route mapping (GRRM) algorithms. Density, glass transition temperature, Young's modulus, and curing conversion is used to validate the procedure. Experimental and simulation results indicate that base resin with multi-functional reaction groups increases glass-transition temperature and Young's modulus because of cross-linked formations at the molecular scale.

cond-mat.soft↗

arXiv2020

Pose-Assisted Multi-Camera Collaboration for Active Object Tracking

Active Object Tracking (AOT) is crucial to many visionbased applications, e.g., mobile robot, intelligent surveillance. However, there are a number of challenges when deploying active tracking in complex scenarios, e.g., target is frequently occluded by obstacles. In this paper, we extend the single-camera AOT to a multi-camera setting, where cameras tracking a target in a collaborative fashion. To achieve effective collaboration among cameras, we propose a novel Pose-Assisted Multi-Camera Collaboration System, which enables a camera to cooperate with the others by sharing camera poses for active object tracking. In the system, each camera is equipped with two controllers and a switcher: The vision-based controller tracks targets based on observed images. The pose-based controller moves the camera in accordance to the poses of the other cameras. At each step, the switcher decides which action to take from the two controllers according to the visibility of the target. The experimental results demonstrate that our system outperforms all the baselines and is capable of generalizing to unseen environments. The code and demo videos are available on our website https://sites.google.com/view/pose-assistedcollaboration.

cs.CV↗

arXiv2020

Large Batch Optimization for Deep Learning: Training BERT in 76 minutes

Training large deep neural networks on massive datasets is computationally very challenging. There has been recent surge in interest in using large batch stochastic optimization methods to tackle this issue. The most prominent algorithm in this line of research is LARS, which by employing layerwise adaptive learning rates trains ResNet on ImageNet in a few minutes. However, LARS performs poorly for attention models like BERT, indicating that its performance gains are not consistent across tasks. In this paper, we first study a principled layerwise adaptation strategy to accelerate training of deep neural networks using large mini-batches. Using this strategy, we develop a new layerwise adaptive large batch optimization technique called LAMB; we then provide convergence analysis of LAMB as well as LARS, showing convergence to a stationary point in general nonconvex settings. Our empirical results demonstrate the superior performance of LAMB across various tasks such as BERT and ResNet-50 training with very little hyperparameter tuning. In particular, for BERT training, our optimizer enables use of very large batch sizes of 32868 without any degradation of performance. By increasing the batch size to the memory limit of a TPUv3 Pod, BERT training time can be reduced from 3 days to just 76 minutes (Table 1). The LAMB implementation is available at https://github.com/tensorflow/addons/blob/master/tensorflow_addons/optimizers/lamb.py

cs.LG↗

arXiv2019

Interleaved Composite Quantization for High-Dimensional Similarity Search

Similarity search retrieves the nearest neighbors of a query vector from a dataset of high-dimensional vectors. As the size of the dataset grows, the cost of performing the distance computations needed to implement a query can become prohibitive. A method often used to reduce this computational cost is quantization of the vector space and location-based encoding of the dataset vectors. These encodings can be used during query processing to find approximate nearest neighbors of the query point quickly. Search speed can be improved by using shorter codes, but shorter codes have higher quantization error, leading to degraded precision. In this work, we propose the Interleaved Composite Quantization (ICQ) which achieves fast similarity search without using shorter codes. In ICQ, a small subset of the code is used to approximate the distances, with complete codes being used only when necessary. Our method effectively reduces both code length and quantization error. Furthermore, ICQ is compatible with several recently proposed techniques for reducing quantization error and can be used in conjunction with these other techniques to improve results. We confirm these claims and show strong empirical performance of ICQ using several synthetic and real-word datasets.

cs.LG↗

arXiv2019

TOCO: A Framework for Compressing Neural Network Models Based on Tolerance Analysis

Neural network compression methods have enabled deploying large models on emerging edge devices with little cost, by adapting already-trained models to the constraints of these devices. The rapid development of AI-capable edge devices with limited computation and storage requires streamlined methodologies that can efficiently satisfy the constraints of different devices. In contrast, existing methods often rely on heuristic and manual adjustments to maintain accuracy, support only coarse compression policies, or target specific device constraints that limit their applicability. We address these limitations by proposing the TOlerance-based COmpression (TOCO) framework. TOCO uses an in-depth analysis of the model, to maintain the accuracy, in an active learning system. The results of the analysis are tolerances that can be used to perform compression in a fine-grained manner. Finally, by decoupling compression from the tolerance analysis, TOCO allows flexibility to changes in the hardware.

cs.LG↗

arXiv2019

Model reduction for a power grid model

We apply model reduction techniques to the DeMarco power grid model. The DeMarco model, when augmented by an appropriate line failure mechanism, can be used to study cascade failures. Here we examine the DeMarco model without the line failure mechanism and we investigate how to construct reduced order models for subsets of the state variables. We show that due to the oscillating nature of the solutions and the absence of timescale separation between resolved and unresolved variables, the construction of accurate reduced models becomes highly non-trivial since one has to account for long memory effects. In addition, we show that a reduced model which includes even a short memory is drastically better than a memoryless model.

eess.SP↗