arXiv subjects

Ming Zhang

Publications and source records attributed to Ming Zhang.

At least 361 records · Page 20Linked to original sources

arXiv2020

New version of the gedanken experiments to test the weak cosmic censorship in charged dilaton-Lifshitz black holes

In this paper, based on the new version of the gedanken experiments proposed by Sorce and Wald, we examine the weak cosmic censorship in the perturbation process of accreting matter fields for the charged dilaton-Lifshitz black holes. In the investigation, we assume that the black hole is perturbed by some extra matter source satisfied the null energy condition and ultimately settle down to a static charged dilaton-Lifshitz black hole in the asymptotic future. Then, after applying the Noether charge method, we derive the first-order and second-order perturbation inequalities of the perturbation matter fields. As a result, we find that the nearly extremal charged dilaton-Lifshitz black hole cannot be destroyed under the second-order approximation of perturbation. This result implies that the weak cosmic censorship conjecture might be a general feature of the Einstein gravity, and it is independent of the asymptotic behaviors of the black holes.

gr-qc↗

arXiv2020

Can shadows reflect phase structures of black holes?

The relation between the black hole shadow and the black hole thermodynamics is investigated. We find that the phase structure can be reflected by the shadow radius for the spherically symmetric black hole. We also find that the shadow size gives correct information but the distortion of the shadow gives wrong information of the phase structure for the axially symmetric black hole.

gr-qc↗

arXiv2020

Association and Caching in Relay-Assisted mmWave Networks: From A Stochastic Geometry Perspective

Limited backhaul bandwidth and blockage effects are two main factors limiting the practical deployment of millimeter wave (mmWave) networks. To tackle these issues, we study the feasibility of relaying as well as caching in mmWave networks. A user association and relaying (UAR) criterion dependent on both caching status and maximum biased received power is proposed by considering the spatial correlation caused by the coexistence of base stations (BSs) and relay nodes (RNs). A joint UAR and caching placement problem is then formulated to maximize the backhaul offloading traffic. Using stochastic geometry tools, we decouple the joint UAR and caching placement problem by analyzing the relationship between UAR probabilities and caching placement probabilities. We then optimize the transformed caching placement problem based on polyblock outer approximation by exploiting the monotonic property in the general case and utilizing convex optimization in the noise-limited case. Accordingly, we propose a BS and RN selection algorithm where caching status at BSs and maximum biased received power are jointly considered. Experimental results demonstrate a significant enhancement of backhaul offloading using the proposed algorithms, and show that deploying more RNs and increasing cache size in mmWave networks is a more cost-effective alternative than increasing BS density to achieve similar backhaul offloading performance.

cs.NI↗

arXiv2020

Revisiting collisional Penrose processes in term of escape probabilities for spinning particles

We first study the escape probability of the spinning particle emitted from the Kerr black hole and find that the escape probability increases with the spin of the particle around the extreme Kerr black hole; in contrast, the escape probability decreases at the position near the horizon but increases at the position far away from the horizon with the increasing spin of the particle. We then probe the relation between the escape probabilities and the energy extraction efficiencies of collisional Penrose processes for the particles with varying spin. For the extreme Kerr black hole, the efficiency increases with the escape probability; for the non-extreme Kerr black hole, the near-horizon-efficiency decreases with the escape probability whilst the efficiency may increase with the escape probability in the ergosphere. In the event horizon limit, we also find that the average escape probability of the spinning particle produced in the collisional Penrose process decreases with the rotation parameter of the Kerr black hole.

gr-qc↗

arXiv2020

Testing the Weak Cosmic Censorship Conjecture in Lanczos-Lovelock gravity

In this paper, we test the weak cosmic censorship conjecture (WCCC) in the nearly extremal static charged black holes of Lanczos-Lovelock-Maxwell gravity based on the new version of gedanken experiments proposed by Sorce and Wald. After introducing the null energy condition of the matter fields, we show that the (nearly) extremal black holes can be destroyed under the first-order approximation for the case with $S'(r_h)\leq 0$, where $S(r_h)$ is the entropy of the background black hole geometry in Lanczos-Lovelock gravity. It implies that the WCCC is violated in these situations. For the case with $S'(r_h)>0$, the nearly extremal black holes cannot be overcharged by the new version of the gedanken experiments for both first- and second-order approximations of perturbation. These results indicate that the WCCC is satisfied in the Lanczos-Lovelock gravity with the condition $S'(r_h)>0$. Our work also implies that the WCCC will play a natural role to constrain the Lanczos-Lovelock gravities. Finally, we also show that the destroy condition $S'(r_h)<0$ implies that the nearly extremal black hole is thermodynamically unstable under the first-order approximation.

gr-qc↗

arXiv2020

Augmented Bi-path Network for Few-shot Learning

Few-shot Learning (FSL) which aims to learn from few labeled training data is becoming a popular research topic, due to the expensive labeling cost in many real-world applications. One kind of successful FSL method learns to compare the testing (query) image and training (support) image by simply concatenating the features of two images and feeding it into the neural network. However, with few labeled data in each class, the neural network has difficulty in learning or comparing the local features of two images. Such simple image-level comparison may cause serious mis-classification. To solve this problem, we propose Augmented Bi-path Network (ABNet) for learning to compare both global and local features on multi-scales. Specifically, the salient patches are extracted and embedded as the local features for every image. Then, the model learns to augment the features for better robustness. Finally, the model learns to compare global and local features separately, i.e., in two paths, before merging the similarities. Extensive experiments show that the proposed ABNet outperforms the state-of-the-art methods. Both quantitative and visual ablation studies are provided to verify that the proposed modules lead to more precise comparison results.

cs.CV↗

arXiv2020

EasyQuant: Post-training Quantization via Scale Optimization

The 8 bits quantization has been widely applied to accelerate network inference in various deep learning applications. There are two kinds of quantization methods, training-based quantization and post-training quantization. Training-based approach suffers from a cumbersome training process, while post-training quantization may lead to unacceptable accuracy drop. In this paper, we present an efficient and simple post-training method via scale optimization, named EasyQuant (EQ),that could obtain comparable accuracy with the training-based method.Specifically, we first alternately optimize scales of weights and activations for all layers target at convolutional outputs to further obtain the high quantization precision. Then, we lower down bit width to INT7 both for weights and activations, and adopt INT16 intermediate storage and integer Winograd convolution implementation to accelerate inference.Experimental results on various computer vision tasks show that EQ outperforms the TensorRT method and can achieve near INT8 accuracy in 7 bits width post-training.

cs.CV↗

arXiv2020

Escape probability of particle from Kerr-Sen black hole

Assuming that a particle source is at rest in a locally non-rotating frame on the equatorial plane of the Kerr-Sen black hole, we investigate the escape of the massless particle and massive particle from the black hole to spatial infinity. We calculate the escape probabilities of the emitted particles. We find that the angular momentum of the Kerr-Sen black hole boosts the escape probabilities near the horizon of the extreme Kerr-Sen black hole; however the angular momentum of the Kerr-Sen black hole suppresses the escape probabilities near the horizon of the non-extreme Kerr-Sen black hole. We also discover that the horizon limit of the Kerr-Sen black hole is not the most difficult position to escape for a particle in certain conditions.

gr-qc↗

arXiv2020

Relaxation rate of RNdS black hole

We investigate the relaxation rate of the RNdS black hole perturbed by neutral massless scalar field in the eikonal limit. We find that the fastest relaxation rate of the composed system increases with the cosmological constant for all spacetime dimensions. We also find that, when the cosmological constant decreases from maximum value to zero, the corresponding critical charge of the four-dimensional RNdS black hole maximizing the relaxation rate of the composed system gradually decreases monotonically to $\bar{Q}=0.726$. However, for the higher dimensional systems, this is not the case and the critical charges decrease from maximum to zero.

gr-qc↗

arXiv2020

Learning to Customize Model Structures for Few-shot Dialogue Generation Tasks

Training the generative models with minimal corpus is one of the critical challenges for building open-domain dialogue systems. Existing methods tend to use the meta-learning framework which pre-trains the parameters on all non-target tasks then fine-tunes on the target task. However, fine-tuning distinguishes tasks from the parameter perspective but ignores the model-structure perspective, resulting in similar dialogue models for different tasks. In this paper, we propose an algorithm that can customize a unique dialogue model for each task in the few-shot setting. In our approach, each dialogue model consists of a shared module, a gating module, and a private module. The first two modules are shared among all the tasks, while the third one will differentiate into different network structures to better capture the characteristics of the corresponding task. The extensive experiments on two datasets show that our method outperforms all the baselines in terms of task consistency, response quality, and diversity.

cs.CL↗

arXiv2020

Multi-task Learning via Adaptation to Similar Tasks for Mortality Prediction of Diverse Rare Diseases

Mortality prediction of diverse rare diseases using electronic health record (EHR) data is a crucial task for intelligent healthcare. However, data insufficiency and the clinical diversity of rare diseases make it hard for directly training deep learning models on individual disease data or all the data from different diseases. Mortality prediction for these patients with different diseases can be viewed as a multi-task learning problem with insufficient data and large task number. But the tasks with little training data also make it hard to train task-specific modules in multi-task learning models. To address the challenges of data insufficiency and task diversity, we propose an initialization-sharing multi-task learning method (Ada-Sit) which learns the parameter initialization for fast adaptation to dynamically measured similar tasks. We use Ada-Sit to train long short-term memory networks (LSTM) based prediction models on longitudinal EHR data. And experimental results demonstrate that the proposed model is effective for mortality prediction of diverse rare diseases.

cs.LG↗

arXiv2020

Impacts of dark energy on constraining neutrino mass after Planck 2018

Considering the mass splittings of three active neutrinos, we investigate how the nature of dark energy affects the cosmological constraints on the total neutrino mass $\sum m_ν$ using the latest cosmological observations. In this paper, some typical dark energy models, including $Λ$CDM, $w$CDM, CPL, and HDE models, are discussed. In the analysis, we also consider the effects from the neutrino mass hierarchies, i.e., the degenerate hierarchy (DH), the normal hierarchy (NH), and the inverted hierarchy (IH). We employ the current cosmological observations to do the analysis, including the Planck 2018 temperature and polarization power spectra, the baryon acoustic oscillations (BAO), the type Ia supernovae (SNe), and the Hubble constant $H_0$ measurement. In the $Λ$CDM+$\sum m_ν$ model, we obtain the upper limits of the neutrino mass $\sum m_ν< 0.123$ eV (DH), $\sum m_ν< 0.156$ eV (NH), and $\sum m_ν< 0.185$ eV (IH) at the $95\%$ C.L., using the Planck+BAO+SNe data combination. For the $w$CDM+$\sum m_ν$ model and the CPL+$\sum m_ν$ model, larger upper limits of $\sum m_ν$ are obtained compared to those of the $Λ$CDM+$\sum m_ν$ model. The most stringent constraint on the neutrino mass, $\sum m_ν<0.080$ eV (DH), is derived in the HDE+$\sum m_ν$ model. In addition, we find that the inclusion of the local measurement of the Hubble constant in the data combination leads to tighter constraints on the total neutrino mass in all these dark energy models.

astro-ph.CO↗

arXiv2020

Stable circular orbits of spinning test particles around accelerating Kerr black hole

We investigate the stable circular orbits of the spinning test particles around the accelerating Kerr black hole on the equatorial plane. To this end, we first calculate the equations of motion and analyze the parameter space for the particles. We study the effect of the particle's spin and the black hole's acceleration on the conserved angular momentum, conserved energy and radius of the spinning test particle on the innermost stable circular orbit. We find that the effect of the particle's spin on the orbit parameters is almost linear, the effect of the black hole's acceleration on those parameters is nonlinear. We also explore the effects of the particle's spin and the black hole's acceleration on the periastron shift for the spinning particle in the nearly stable circular orbit.

gr-qc↗

arXiv2020

Entropy increases at linear order in scalar-hairy Lovelock gravity

In this paper, we investigate the second law of the black holes in Lovelock gravity sourced by a conformally coupled scalar field under the first-order approximation when the perturbation matter fields satisfy the null energy condition. First of all, we show that the Wald entropy of this theory does not obey the linearized second law for the scalar-hairy Lovelock gravity which contains the higher curvature terms even if we replace the gravitational part of Wald entropy with Jacobson-Myers (JM) entropy. This implies that we cannot naively add the scalar field term of the Wald entropy to the JM entropy of the purely Lovelock gravity to get a valid linearized second law. By rescaling the metric, the action of the scalar field can be written as a purely Lovelock action with another metric. Using this property, by analogy with the JM entropy of the purely Lovelock gravity, we introduce a new formula of the entropy in the scalar-hairy Lovelock gravity. Then, we show that this new JM entropy increases along the event horizon for Vaidya-like black hole solutions and therefore it obeys a linearized second law. Moreover, we show that different from the entropy in $F($Riemann$)$ gravity, the difference between the JM entropy and Wald entropy also contains some additional corrections from the scalar field.

hep-th↗

arXiv2020

GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation

Molecular graph generation is a fundamental problem for drug discovery and has been attracting growing attention. The problem is challenging since it requires not only generating chemically valid molecular structures but also optimizing their chemical properties in the meantime. Inspired by the recent progress in deep generative models, in this paper we propose a flow-based autoregressive model for graph generation called GraphAF. GraphAF combines the advantages of both autoregressive and flow-based approaches and enjoys: (1) high model flexibility for data density estimation; (2) efficient parallel computation for training; (3) an iterative sampling process, which allows leveraging chemical domain knowledge for valency checking. Experimental results show that GraphAF is able to generate 68% chemically valid molecules even without chemical knowledge rules and 100% valid molecules with chemical rules. The training process of GraphAF is two times faster than the existing state-of-the-art approach GCPN. After fine-tuning the model for goal-directed property optimization with reinforcement learning, GraphAF achieves state-of-the-art performance on both chemical property optimization and constrained property optimization.

cs.LG↗

arXiv2020

A Lightweight and Accurate Localization Algorithm Using Multiple Inertial Measurement Units

This paper proposes a novel inertial-aided localization approach by fusing information from multiple inertial measurement units (IMUs) and exteroceptive sensors. IMU is a low-cost motion sensor which provides measurements on angular velocity and gravity compensated linear acceleration of a moving platform, and widely used in modern localization systems. To date, most existing inertial-aided localization methods exploit only one single IMU. While the single-IMU localization yields acceptable accuracy and robustness for different use cases, the overall performance can be further improved by using multiple IMUs. To this end, we propose a lightweight and accurate algorithm for fusing measurements from multiple IMUs and exteroceptive sensors, which is able to obtain noticeable performance gain without incurring additional computational cost. To achieve this, we first probabilistically map measurements from all IMUs onto a virtual IMU. This step is performed by stochastic estimation with least-square estimators and probabilistic marginalization of inter-IMU rotational accelerations. Subsequently, the propagation model for both state and error state of the virtual IMU is also derived, which enables the use of the classical filter-based or optimization-based sensor fusion algorithms for localization. Finally, results from both simulation and real-world tests are provided, which demonstrate that the proposed algorithm outperforms competing algorithms by noticeable margins.

cs.RO↗

arXiv2019

Learning to Answer Ambiguous Questions with Knowledge Graph

In the task of factoid question answering over knowledge base, many questions have more than one plausible interpretation. Previous works on SimpleQuestions assume only one interpretation as the ground truth for each question, so they lack the ability to answer ambiguous questions correctly. In this paper, we present a new way to utilize the dataset that takes into account the existence of ambiguous questions. Then we introduce a simple and effective model which combines local knowledge subgraph with attention mechanism. Our experimental results show that our approach achieves outstanding performance in this task.

cs.AI↗

arXiv2019

Effect of protein binding on exposure of unbound and total mycophenolic acid: a population pharmacokinetic analysis in Chinese adult kidney transplant recipients

AIMS A population pharmacokinetic (PK) analysis was performed to: (1) characterise the PK of unbound and total mycophenolic acid (MPA) and its 7-O-mycophenolic acid glucuronide (MPAG) metabolite, and (2) identify the clinically significant covariates that cause variability in the dose-exposure relationship to facilitate dose optimisation. METHODS A total of 740 unbound MPA (uMPA), 741 total MPA (tMPA) and 734 total MPAG (tMPAG) concentration-time data from 58 Chinese kidney transplant patients were analysed using a nonlinear mixed-effect model. The influence of covariates was tested using a stepwise procedure. RESULTS The PK of unbound MPA and MPAG were characterised by a two- and one-compartment model with first-order elimination, respectively. Apparent clearance of uMPA (CLuMPA/F) was estimated to be 852 L/h with a relative standard error (RSE) of 7.1%. The tMPA and uMPA were connected using a linear protein binding model, in which the protein binding rate constant (kB) increased non-linearly with the serum albumin (ALB) concentration. The estimated kB was 53.4 /h (RSE, 2.3%) for patients with ALB of 40 g/L. In addition, model-based simulation showed that changes in ALB substantially affected tMPA but not uMPA exposure. CONCLUSIONS The established model adequately described the population PK characteristics of the uMPA, tMPA, and MPAG. The estimated CLuMPA/F and unbound fraction of MPA (FUMPA) in Chinese kidney transplant recipients were comparable to those published previously in Caucasians. We recommend monitoring uMPA instead of tMPA to optimise mycophenolate mofetil (MMF) dosing for patients with lower ALB levels.

q-bio.TO↗