arXiv subjects

Kai Li

Publications and source records attributed to Kai Li.

At least 397 records · Page 22Linked to original sources

arXiv2019

Visual Semantic Reasoning for Image-Text Matching

Image-text matching has been a hot research topic bridging the vision and language areas. It remains challenging because the current representation of image usually lacks global semantic concepts as in its corresponding text caption. To address this issue, we propose a simple and interpretable reasoning model to generate visual representation that captures key objects and semantic concepts of a scene. Specifically, we first build up connections between image regions and perform reasoning with Graph Convolutional Networks to generate features with semantic relationships. Then, we propose to use the gate and memory mechanism to perform global semantic reasoning on these relationship-enhanced features, select the discriminative information and gradually generate the representation for the whole scene. Experiments validate that our method achieves a new state-of-the-art for the image-text matching on MS-COCO and Flickr30K datasets. It outperforms the current best method by 6.8% relatively for image retrieval and 4.8% relatively for caption retrieval on MS-COCO (Recall@1 using 1K test set). On Flickr30K, our model improves image retrieval by 12.6% relatively and caption retrieval by 5.8% relatively (Recall@1). Our code is available at https://github.com/KunpengLi1994/VSRN.

cs.CV↗

arXiv2019

Escape of α-particle in inertial confinement fusion

Escape of $α$-particles from a burning or an ignited burning deuterium-tritium (DT) fuel with temperature up to more than tens keV is very important in inertial confinement fusion, which can significantly influence not only the hot spot dynamics and the energy gain but also the shielding design in fusion devices. In this paper, we study the $α$-particle escape from a burning or an ignited burning DT fuel by considering the modifications including the $α$-particle stopping by both DT ions and electrons with their Maxwellian average stopping weights, the relativity effect on electron distribution, and the modified Coulomb logarithm of the DT-$α$ particle collisions. As a result of our studies, the escape-effect from our modified model is obviously stronger than those from the traditional models. A fitted expression is presented to calculate the escape factor in a DT fuel, which can be applied to a burning fuel with temperatures of 1 to 150 keV and areal densities of 0.04 to 3 g/cm$^2$ with an accuracy within $\pm0.02$. Finally, we discuss the $α$-particle escape-effect on the hot-spot dynamics and the thermonuclear energy gain by comparing the results with escape factors from different models.

physics.plasm-ph↗

arXiv2019

No Need of Data Pre-processing: A General Framework for Radio-Based Device-Free Context Awareness

Device-free context awareness is important to many applications. There are two broadly used approaches for device-free context awareness, i.e. video-based and radio-based. Video-based applications can deliver good performance, but privacy is a serious concern. Radio-based context awareness has drawn researchers attention instead because it does not violate privacy and radio signal can penetrate obstacles. Recently, deep learning has been introduced into radio-based device-free context awareness and helps boost the recognition accuracy. The present works design explicit methods for each radio based application. They also use one additional step to extract features before conducting classification and exploit deep learning as a classification tool. The additional initial data processing step introduces unnecessary noise and information loss. Without initial data processing, it is, however, challenging to explore patterns of raw signals. In this paper, we are the first to propose an innovative deep learning based general framework for both signal processing and classification. The key novelty of this paper is that the framework can be generalised for all the radio-based context awareness applications. We also eliminate the additional effort to extract features from raw radio signals. We conduct extensive evaluations to show the superior performance of our proposed method and its generalisation.

cs.NI↗

arXiv2019

Secure Consistency Verification for Untrusted Cloud Storage by Public Blockchains

This work presents ContractChecker, a Blockchain-based security protocol for verifying the storage consistency between the mutually distrusting cloud provider and clients. Unlike existing protocols, the ContractChecker uniquely delegates log auditing to the Blockchain, and has the advantages in reducing client cost and lowering requirements on client availability, lending itself to modern scenarios with mobile and web clients. The ContractChecker collects the logs from both clients and the cloud server, and verifies the consistency by cross-checking the logs. By this means, it does not only detects the attacks from malicious clients and server forging their logs, but also is able to mitigate those attacks and recover the system from them. In addition, we design new attacks against ContractChecker exploiting various limits in real Blockchain systems (e.g., write unavailability, Blockchain forks, contract race conditions). We analyze and harden the security of ContractChecker protocols against the proposed new attacks. For evaluating the cost, we build a functional prototype of the ContractChecker on Ethereum/Solidity. By experiments on private and public Ethereum testnets, we extensively evaluate the cost of the ContractChecker in comparison with that of existing client-based log auditing works. The result shows the ContractChecker can scale to hundreds of clients and save client costs by more than one order of magnitude.

cs.CR↗

arXiv2019

Quantum cluster approach to the topological invariants in correlated Chern insulators

We detect the topological properties of Chern insulators with strong Coulomb interactions by use of cluster perturbation theory and variational cluster approach. The common scheme in previous studies only involves the calculation of the interacting bulk Chern number within the natural unit cell by means of the so-called topological Hamiltonian. With close investigations on a prototype model, the half-filled Haldane Hubbard model, which is subject to both periodic and open boundary conditions, we uncover the unexpected failure of this scheme due to the explicit breaking of the translation symmetry. Instead, we assert that the faithful interacting bulk Chern number in the framework of quantum cluster approaches can be computed in the enlarged unit cell, which is free of the fault caused by the explicit translation symmetry breaking and consistent with the interacting bulk-edge correspondence.

cond-mat.str-el↗

arXiv2019

FPGA-based real-time autocorrelator and its application in dynamic light scattering

Digital correlators play a significant role in dynamic light scattering (DLS) technology, which characterizes particle size distribution. We present a field programmable gate array (FPGA)-based digital correlator that can be applied to process DLS data. To satisfy the DLS requirements in the FPGA logic with limited resources, a multiple lag time period (multi-τ ) method is employed that does not require storing the full data set in memory. Moreover, the device directly accepts the transistor-transistor logic (TTL) signal from the photon counting detector by measuring the time intervals between photon events and calculates the autocorrelation functions in real time. Furthermore, we derive estimates for the error arising from the use of the multi-τ correlator. We implement all the necessary operations in a single Xilinx FPGA chip with a lag time from 10 ns to 45 min, including a highly optimized photon counter.

eess.SP↗

arXiv2019

On-board Deep Q-Network for UAV-assisted Online Power Transfer and Data Collection

Unmanned Aerial Vehicles (UAVs) with Microwave Power Transfer (MPT) capability provide a practical means to deploy a large number of wireless powered sensing devices into areas with no access to persistent power supplies. The UAV can charge the sensing devices remotely and harvest their data. A key challenge is online MPT and data collection in the presence of on-board control of a UAV (e.g., patrolling velocity) for preventing battery drainage and data queue overflow of the sensing devices, while up-to-date knowledge on battery level and data queue of the devices is not available at the UAV. In this paper, an on-board deep Q-network is developed to minimize the overall data packet loss of the sensing devices, by optimally deciding the device to be charged and interrogated for data collection, and the instantaneous patrolling velocity of the UAV. Specifically, we formulate a Markov Decision Process (MDP) with the states of battery level and data queue length of sensing devices, channel conditions, and waypoints given the trajectory of the UAV; and solve it optimally with Q-learning. Furthermore, we propose the on-board deep Q-network that can enlarge the state space of the MDP, and a deep reinforcement learning based scheduling algorithm that asymptotically derives the optimal solution online, even when the UAV has only outdated knowledge on the MDP states. Numerical results demonstrate that the proposed deep reinforcement learning algorithm reduces the packet loss by at least 69.2%, as compared to existing non-learning greedy algorithms.

cs.NI↗

arXiv2019

Patient-Specific Effects of Medication Using Latent Force Models with Gaussian Processes

Multi-output Gaussian processes (GPs) are a flexible Bayesian nonparametric framework that has proven useful in jointly modeling the physiological states of patients in medical time series data. However, capturing the short-term effects of drugs and therapeutic interventions on patient physiological state remains challenging. We propose a novel approach that models the effect of interventions as a hybrid Gaussian process composed of a GP capturing patient physiology convolved with a latent force model capturing effects of treatments on specific physiological features. This convolution of a multi-output GP with a GP including a causal time-marked kernel leads to a well-characterized model of the patients' physiological state responding to interventions. We show that our model leads to analytically tractable cross-covariance functions, allowing scalable inference. Our hierarchical model includes estimates of patient-specific effects but allows sharing of support across patients. Our approach achieves competitive predictive performance on challenging hospital data, where we recover patient-specific response to the administration of three common drugs: one antihypertensive drug and two anticoagulants.

stat.ML↗

arXiv2019

Custom Edge-Element FEM Solver and its Application to Eddy-Current Simulation of Realistic 2M-Element Human Brain Phantom

Extensive research papers of three-dimensional computational techniques are widely used for the investigation of human brain pathophysiology. Eddy current analyzing could provide an indication of conductivity change within a biological body. A significant obstacle to current trend analyses is the development of a numerically stable and efficiency-finite element scheme that performs well at low frequency and does not require a large number of degrees of freedom. Here, a custom finite element method (FEM) solver based on edge elements is proposed using the weakly coupled theory, which separates the solution into two steps. First, the background field (the magnetic vector potential on each edge) is calculated and stored. Then, the electric scalar potential on each node is obtained by FEM based on Galerkin formulations. Consequently, the electric field and eddy current distribution in the object can be obtained. This solver is more efficient than typical commercial solvers since it reduces the vector eddy current equation to a scalar one, and reduces the meshing domain to just the eddy current region. It can therefore tackle complex eddy current calculations for models with much larger numbers of elements, such as those encountered in eddy current computation in biological tissues. An example is presented with a realistic human brain mesh of 2 million elements. In addition, with this solver, the equivalent magnetic field induced from the excitation coil is applied, and therefore there is no need to mesh the excitation coil. In combination, these significantly increase the efficiency of the solver.

physics.med-ph↗

arXiv2019

Photometric investigations on two totally eclipsing contact binaries: V342 UMa and V509 Cam

By analyzing two sets of complete $BVR_cI_c$ light curves of V342 UMa and three sets of complete $BVR_cI_c$ light curves of V509 Cam, we determined that the two systems are both W-subtype contact binaries and that V342 UMa shows a shallow contact configuration, while V509 Cam exhibits a medium contact configuration. Since both of them are totally eclipsing binaries, the physical parameters derived only by the photometric light curves are reliable.Meanwhile, the period changes of the two targets were analyzed based on all available eclipsing times. We discovered that V342 UMa shows long-term period decrease with a rate of $-1.02(\pm0.54)\times10^{-7}$ days/year and that V509 Cam exhibits long-term period increase with a rate of $3.96(\pm0.90)\times10^{-8}$ days/year. Both the conservative mass transfer and AML via magnetic stellar winds can interpret the long-term period decrease of V342 UMa. The long-term period increase of V509 Cam can be explained by mass transfer from the less massive star to the more massive one. The absolute parameters of the two binaries were estimated according to the Gaia distances and our derived photometric solution results. This method can be extended to other contact binaries without radial velocities but with reliable photometric solutions. The evolutionary states of them were discussed, we found that they reveal identical properties of other W-subtype contact systems.

astro-ph.SR↗

arXiv2019

FeatherNets: Convolutional Neural Networks as Light as Feather for Face Anti-spoofing

Face Anti-spoofing gains increased attentions recently in both academic and industrial fields. With the emergence of various CNN based solutions, the multi-modal(RGB, depth and IR) methods based CNN showed better performance than single modal classifiers. However, there is a need for improving the performance and reducing the complexity. Therefore, an extreme light network architecture(FeatherNet A/B) is proposed with a streaming module which fixes the weakness of Global Average Pooling and uses less parameters. Our single FeatherNet trained by depth image only, provides a higher baseline with 0.00168 ACER, 0.35M parameters and 83M FLOPS. Furthermore, a novel fusion procedure with ``ensemble + cascade'' structure is presented to satisfy the performance preferred use cases. Meanwhile, the MMFD dataset is collected to provide more attacks and diversity to gain better generalization. We use the fusion method in the Face Anti-spoofing Attack Detection Challenge@CVPR2019 and got the result of 0.0013(ACER), 0.999(TPR@FPR=10e-2), 0.998(TPR@FPR=10e-3) and 0.9814(TPR@FPR=10e-4).

cs.CV↗

arXiv2019

Residual Non-local Attention Networks for Image Restoration

In this paper, we propose a residual non-local attention network for high-quality image restoration. Without considering the uneven distribution of information in the corrupted images, previous methods are restricted by local convolutional operation and equal treatment of spatial- and channel-wise features. To address this issue, we design local and non-local attention blocks to extract features that capture the long-range dependencies between pixels and pay more attention to the challenging parts. Specifically, we design trunk branch and (non-)local mask branch in each (non-)local attention block. The trunk branch is used to extract hierarchical features. Local and non-local mask branches aim to adaptively rescale these hierarchical features with mixed attentions. The local mask branch concentrates on more local structures with convolutional operations, while non-local attention considers more about long-range dependencies in the whole feature map. Furthermore, we propose residual local and non-local attention learning to train the very deep network, which further enhance the representation ability of the network. Our proposed method can be generalized for various image restoration applications, such as image denoising, demosaicing, compression artifacts reduction, and super-resolution. Experiments demonstrate that our method obtains comparable or better results compared with recently leading methods quantitatively and visually.

cs.CV↗

arXiv2019

PZnet: Efficient 3D ConvNet Inference on Manycore CPUs

Convolutional nets have been shown to achieve state-of-the-art accuracy in many biomedical image analysis tasks. Many tasks within biomedical analysis domain involve analyzing volumetric (3D) data acquired by CT, MRI and Microscopy acquisition methods. To deploy convolutional nets in practical working systems, it is important to solve the efficient inference problem. Namely, one should be able to apply an already-trained convolutional network to many large images using limited computational resources. In this paper we present PZnet, a CPU-only engine that can be used to perform inference for a variety of 3D convolutional net architectures. PZNet outperforms MKL-based CPU implementations of PyTorch and Tensorflow by more than 3.5x for the popular U-net architecture. Moreover, for 3D convolutions with low featuremap numbers, cloud CPU inference with PZnet outperfroms cloud GPU inference in terms of cost efficiency.

cs.CV↗

arXiv2019

Quasiparticle Properties under Interactions in Weyl and Nodal Line Semimetals

The quasiparticle spectra of interacting Weyl and nodal-line semimetals on a cubic lattice are studied using the cluster perturbation theory. By tracking the spectral functions under interaction, we find that the Weyl points will move to and meet at a specific point in one Weyl semimetal model, while in the other Weyl semimetal model they are immobile. In the nodal-line semimetals, we find that the nodal line shrinks to a point and then disappears under interaction in one-nodal-line system. When we add another nodal line to this system, we find that the two nodal lines both shrink to specific points, but the disappearing processes of the two nodal lines are not synchronized. We argue that the nontrivial evolution of Weyl points and nodal lines under interaction is due to the presence of symmetry breaking order, e.g., a ferromagnetic moment, in the framework of mean field theory, whereas the stability of Weyl points under interaction is protected by symmetry. Among all these models, the spectral gap is finally opened when the interaction is strong enough.

cond-mat.str-el↗

arXiv2019

Data objects and documenting scientific processes: An analysis of data events in biodiversity data papers

The data paper, an emerging scholarly genre, describes research datasets and is intended to bridge the gap between the publication of research data and scientific articles. Research examining how data papers report data events, such as data transactions and manipulations, is limited. The research reported on in this paper addresses this limitation and investigated how data events are inscribed in data papers. A content analysis was conducted examining the full texts of 82 data papers, drawn from the curated list of data papers connected to the Global Biodiversity Information Facility (GBIF). Data events recorded for each paper were organized into a set of 17 categories. Many of these categories are described together in the same sentence, which indicates the messiness of data events in the laboratory space. The findings challenge the degrees to which data papers are a distinct genre compared to research papers and they describe data-centric research processes in a through way. This paper also discusses how our results could inform a better data publication ecosystem in the future.

cs.DL↗

arXiv2019

Contact binaries at the short period cut-off I. statistics and the first photometric investigations of ten totally eclipsing systems

The period distribution of contact binaries exhibits a very sharp short period cut-off at 0.22 days. In order to provide valuable information on this short period limit, we observed ten totally eclipsing contact binaries with orbital periods near this cut-off. By detailed analysis using W-D code, we determined that two of these systems are A-subtype contact binaries while the others are W-subtype contact binaries and all the targets show shallow contact configurations. Half of the targets exhibit stellar spot activity and four of them have third light. A statistical work on well studied USPCBs was carried out, and physical parameters of fifty-five USPCBs were obtained. Some common properties for these systems were derived. Due to the study of the period-color diagram of USPCBs, we found that the period-color relation of USPCBs is different from other W UMa type contact binaries and USPCBs are metal poor old stars. In addition, the evolutionary states of these systems were discussed by constructing the color-density diagram. We derived that the evolutionary states of the two components of USPCBs show identical characteristics of other contact binaries despite slower evolutionary status caused by smaller mass. We suggested that both the fully convective limit claimed by Rucinski (1992) and the AML theory recommended by Stepien (2006) can produce the short period cut-off and a tertiary companion is very important during the formation of the short period contact binaries by removing angular momentum from the host eclipsing pairs under certain circumstances.

astro-ph.SR↗

arXiv2019

Fast 3D Line Segment Detection From Unorganized Point Cloud

This paper presents a very simple but efficient algorithm for 3D line segment detection from large scale unorganized point cloud. Unlike traditional methods which usually extract 3D edge points first and then link them to fit for 3D line segments, we propose a very simple 3D line segment detection algorithm based on point cloud segmentation and 2D line detection. Given the input unorganized point cloud, three steps are performed to detect 3D line segments. Firstly, the point cloud is segmented into 3D planes via region growing and region merging. Secondly, for each 3D plane, all the points belonging to it are projected onto the plane itself to form a 2D image, which is followed by 2D contour extraction and Least Square Fitting to get the 2D line segments. Those 2D line segments are then re-projected onto the 3D plane to get the corresponding 3D line segments. Finally, a post-processing procedure is proposed to eliminate outliers and merge adjacent 3D line segments. Experiments on several public datasets demonstrate the efficiency and robustness of our method. More results and the C++ source code of the proposed algorithm are publicly available at https://github.com/xiaohulugo/3DLineDetection.

cs.CV↗

arXiv2018

A robust estimation for the extended t-process regression model

Robust estimation and variable selection procedure are developed for the extended t-process regression model with functional data. Statistical properties such as consistency of estimators and predictions are obtained. Numerical studies show that the proposed method performs well.

stat.AP↗