arXiv subjects

Long Chen

Publications and source records attributed to Long Chen.

At least 613 records · Page 34Linked to original sources

arXiv2018

End-to-end driving simulation via angle branched network

Imitation learning for end-to-end autonomous driving has drawn attention from academic communities. Current methods either only use images as the input which is ambiguous when a car approaches an intersection, or use additional command information to navigate the vehicle but not automated enough. Focusing on making the vehicle drive along the given path, we propose a new navigation command that does not require human's participation and a novel model architecture called angle branched network. Both the new navigation command and the angle branched network are easy to understand and effective. Besides, we find that not only segmentation information but also depth information can boost the performance of the driving model. We conduct experiments in a 3D urban simulator and both qualitative and quantitative evaluation results show the effectiveness of our model.

cs.CV↗

arXiv2018

QUICK: QoS-guaranteed Efficient Cloudlet Placement in Wireless Metropolitan Area Networks

This article defines the QoS-guaranteed efficient cloudlet deploy problem in wireless metropolitan area network, which aims to minimize the average access delay of mobile users i.e. the average delay when service requests are successfully sent and being served by cloudlets. Meanwhile, we try to optimize total deploy cost represented by the total number of deployed cloudlets. For the first target, both un-designated capacity and constrained capacity cases are studied, and we have designed efficient heuristic and clustering algorithms respectively. We show our algorithms are more efficient than the existing algorithm. For the second target, we formulate an integer linear programming to minimize the number of used cloudlets with given average access delay requirement. A clustering algorithm is devised to guarantee the scalability. For a special case of the deploy cost optimization problem where all cloudlets' computing capabilities have been given, i.e., designated capacity, a minimal cloudlets efficient heuristic algorithm is further proposed. We finally evaluate the performance of proposed algorithms through extensive experimental simulations. Simulation results demonstrate the proposed algorithms are more than 46% efficient than existing algorithms on the average cloudlet access delay. Compared with existing algorithms, our proposed clustering and heuristic algorithms can reduce the number of deployed cloudlets by about 50% averagely.

cs.NI↗

arXiv2018

Equivalence of Weak Galerkin Methods and Virtual Element Methods for Elliptic Equations

We propose a modification of the weak Galerkin methods and show its equivalence to a new version of virtual element methods. We also show the original weak Galerkin method is equivalent to the non-conforming virtual element method. As a consequence, ideas and techniques used for one method can be transferred to another. The key of the connection is the degree of freedoms.

math.NA↗

arXiv2018

Programming of Finite Element Methods in MATLAB

We discuss how to implement the linear finite element method for solving the Poisson equation. We begin with the data structure to represent the triangulation and boundary conditions, introduce the sparse matrix, and then discuss the assembling process. We pay special attention to an efficient programming style using sparse matrices in MATLAB.

math.NA↗

arXiv2018

Convergence Analysis for A Class of Iterative Methods for Solving Saddle Point Systems

Convergence analysis of a nested iterative scheme proposed by Bank,Welfert and Yserentant (BWY) ([Numer. Math., 666: 645-666, 1990]) for solving saddle point system is presented. It is shown that this scheme converges under weaker conditions: the contraction rate for solving the $(1,1)$ block matrix is bound by $(\sqrt{5}-1)/2$. Similar convergence result is also obtained for a class of inexact Uzawa method with even weaker contraction bound $\sqrt{2}/2$. Preconditioned generalized minimal residual method using BWY method as a preconditioner is shown to converge with realistic assumptions.

math.NA↗

arXiv2018

Efficient Three-stage Auction Schemes for Cloudlets Deployment in Wireless Access Network

Cloudlet deployment and resource allocation for mobile users (MUs) have been extensively studied in existing works for computation resource scarcity. However, most of them failed to jointly consider the two techniques together, and the selfishness of cloudlet and access point (AP) are ignored. Inspired by the group-buying mechanism, this paper proposes three-stage auction schemes by combining cloudlet placement and resource assignment, to improve the social welfare subject to the economic properties. We first divide all MUs into some small groups according to the associated APs. Then the MUs in same group can trade with cloudlets in a group-buying way through the APs. Finally, the MUs pay for the cloudlets if they are the winners in the auction scheme. We prove that our auction schemes can work in polynomial time. We also provide the proofs for economic properties in theory. For the purpose of performance comparison, we compare the proposed schemes with HAF, which is a centralized cloudlet placement scheme without auction. Numerical results confirm the correctness and efficiency of the proposed schemes.

cs.NI↗

arXiv2018

Zero-Shot Visual Recognition using Semantics-Preserving Adversarial Embedding Networks

We propose a novel framework called Semantics-Preserving Adversarial Embedding Network (SP-AEN) for zero-shot visual recognition (ZSL), where test images and their classes are both unseen during training. SP-AEN aims to tackle the inherent problem --- semantic loss --- in the prevailing family of embedding-based ZSL, where some semantics would be discarded during training if they are non-discriminative for training classes, but could become critical for recognizing test classes. Specifically, SP-AEN prevents the semantic loss by introducing an independent visual-to-semantic space embedder which disentangles the semantic space into two subspaces for the two arguably conflicting objectives: classification and reconstruction. Through adversarial learning of the two subspaces, SP-AEN can transfer the semantics from the reconstructive subspace to the discriminative one, accomplishing the improved zero-shot recognition of unseen classes. Comparing with prior works, SP-AEN can not only improve classification but also generate photo-realistic images, demonstrating the effectiveness of semantic preservation. On four popular benchmarks: CUB, AWA, SUN and aPY, SP-AEN considerably outperforms other state-of-the-art methods by an absolute performance difference of 12.2\%, 9.3\%, 4.0\%, and 3.6\% in terms of harmonic mean values

cs.CV↗

arXiv2018

Semantic Augmented Reality Environment with Material-Aware Physical Interactions

In Augmented Reality (AR) environment, realistic interactions between the virtual and real objects play a crucial role in user experience. Much of recent advances in AR has been largely focused on developing geometry-aware environment, but little has been done in dealing with interactions at the semantic level. High-level scene understanding and semantic descriptions in AR would allow effective design of complex applications and enhanced user experience. In this paper, we present a novel approach and a prototype system that enables the deeper understanding of semantic properties of the real world environment, so that realistic physical interactions between the real and the virtual objects can be generated. A material-aware AR environment has been created based on the deep material learning using a fully convolutional network (FCN). The state-of-the-art dense Simultaneous Localisation and Mapping (SLAM) has been used for the semantic mapping. Together with efficient accelerated 3D ray casting, natural and realistic physical interactions are generated for interactive AR games. Our approach has significant impact on the future development of advanced AR systems and applications.

cs.CV↗

arXiv2018

Self-Supervised Monocular Image Depth Learning and Confidence Estimation

Convolutional Neural Networks (CNNs) need large amounts of data with ground truth annotation, which is a challenging problem that has limited the development and fast deployment of CNNs for many computer vision tasks. We propose a novel framework for depth estimation from monocular images with corresponding confidence in a self-supervised manner. A fully differential patch-based cost function is proposed by using the Zero-Mean Normalized Cross Correlation (ZNCC) that takes multi-scale patches as a matching strategy. This approach greatly increases the accuracy and robustness of the depth learning. In addition, the proposed patch-based cost function can provide a 0 to 1 confidence, which is then used to supervise the training of a parallel network for confidence map learning and estimation. Evaluation on KITTI dataset shows that our method outperforms the state-of-the-art results.

cs.CV↗

arXiv2018

Context-Aware Mixed Reality: A Framework for Ubiquitous Interaction

Mixed Reality (MR) is a powerful interactive technology that yields new types of user experience. We present a semantic based interactive MR framework that exceeds the current geometry level approaches, a step change in generating high-level context-aware interactions. Our key insight is to build semantic understanding in MR that not only can greatly enhance user experience through object-specific behaviours, but also pave the way for solving complex interaction design challenges. The framework generates semantic properties of the real world environment through dense scene reconstruction and deep image understanding. We demonstrate our approach with a material-aware prototype system for generating context-aware physical interactions between the real and the virtual objects. Quantitative and qualitative evaluations are carried out and the results show that the framework delivers accurate and fast semantic information in interactive MR environment, providing effective semantic level interactions.

cs.CV↗

arXiv2018

BRAINS: Joint Bandwidth-Relay Allocation in Multi-Homing Cooperative D2D Networks

Cooperative device to device (CD2D) communication has been considered to be a solution to capacity shortage problem. Combining multi-homing and CD2D techniques together can potentially improve network performance. We propose a novel multi-homing CD2D (MH-CD2D) network, in which multiple homing mobile devices (MMDs) act as relays for the cooperative communications of ordinary mobile devices (OMDs). We formulate such joint bandwidth-relay allocation problem as a two-stage game, in order to deal with two challenges: how to motivate MMDs to lease spare bandwidths and help OMDs to choose appropriate MMD relays. In the first stage, we use a non-cooperative game to model the competition between MMDs in terms of shared bandwidth and price. In the second stage, we model the behavior of OMDs selecting MMDs by an evolutionary game. We prove that there exists Nash equilibrium in the game and propose a distributed incentive scheme named IMES to solve the joint bandwidth-relay allocation problem. Extensive simulation results show that the equilibrium can be achieved and the best response price of one MMD increases with the other's best price in the Stackelberg game. The utility of MMDs increases with the number of OMDs in each OMD group at the evolutionary equilibrium. The proposed algorithms are able to reduce average service delay by more than 25% in comparison to the randomized scheme which is frequently used in existing works. On average, IMES outperforms existing scheme by about 20.37% in terms of utility of MMDs.

cs.NI↗

arXiv2017

Multigrid Methods for Hellan-Herrmann-Johnson Mixed Method of Kirchhoff Plate Bending Problems

A V-cycle multigrid method for the Hellan-Herrmann-Johnson (HHJ) discretization of the Kirchhoff plate bending problems is developed in this paper. It is shown that the contraction number of the V-cycle multigrid HHJ mixed method is bounded away from one uniformly with respect to the mesh size. The uniform convergence is achieved for the V-cycle multigrid method with only one smoothing step and without full elliptic regularity. The key is a stable decomposition of the kernel space which is derived from an exact sequence of the HHJ mixed method, and the strengthened Cauchy Schwarz inequality. Some numerical experiments are provided to confirm the proposed V-cycle multigrid method. The exact sequences of the HHJ mixed method and the corresponding commutative diagram is of some interest independent of the current context.

math.NA↗

arXiv2017

TARCO: Two-Stage Auction for D2D Relay Aided Computation Resource Allocation in Hetnet

In heterogeneous cellular network, task scheduling for computation offloading is one of the biggest challenges. Most works focus on alleviating heavy burden of macro base stations by moving the computation tasks on macro-cell user equipment (MUE) to remote cloud or small-cell base stations. But the selfishness of network users is seldom considered. Motivated by the cloud edge computing, this paper provides incentive for task transfer from macro cell users to small cell base stations. The proposed incentive scheme utilizes small cell user equipment to provide relay service. The problem of computation offloading is modelled as a two-stage auction, in which the remote MUEs with common social character can form a group and then buy the computation resource of small-cell base stations with the relay of small cell user equipment. A two-stage auction scheme named TARCO is contributed to maximize utilities for both sellers and buyers in the network. The truthful, individual rationality and budget balance of the TARCO are also proved in this paper. In addition, two algorithms are proposed to further refine TARCO on the social welfare of the network. Extensive simulation results demonstrate that, TARCO is better than random algorithm by about 104.90% in terms of average utility of MUEs, while the performance of TARCO is further improved up to 28.75% and 17.06% by the proposed two algorithms, respectively.

cs.NI↗

arXiv2017

Ballistic transport of long wavelength phonons and thermal conductivity accumulation in nanograined silicon-germanium alloys

Computationally efficient modeling of the thermal conductivity of materials is crucial to thorough experimental planning and theoretical understanding of thermal properties. We present a modeling approach in this work that utilizes frequency-dependent effective medium to calculate lattice thermal conductivity of nanostructured solids. The method accurately predicts a significant reduction in the thermal conductivity of nanostructured Si80Ge20 systems, along with previous reported thermal conductivities in nanowires and nanoparticles-in-matrix materials. We use our model to gain insight into the role of long wavelength phonons on the thermal conductivity of nanograined silicon-germanium alloys. Through thermal conductivity accumulation calculations with our modified effective medium model, we show that phonons with wavelengths much greater than the average grain size will not be impacted by grain boundary scattering, counter to the traditionally assumed notion that grain boundaries in solids will act as diffusive interfaces that will limit long wavelength phonon transport. This is further supported through a modulation frequency dependent thermal conductivity as measured with time-domain thermoreflectance.

cond-mat.mes-hall↗

arXiv2017

Convergence of Adaptive Mixed Finite Element Methods for the Hodge Laplacian Equation: without harmonic forms

Finite element exterior calculus (FEEC) has been developed as a systematical framework for constructing and analyzing stable and accurate numerical method for partial differential equations by employing differential complexes. This paper is devoted to analyze convergence of adaptive mixed finite element methods for Hodge Laplacian equations based on FEEC without considering harmonic forms. More precisely, a residual type posteriori error estimates is obtained by using the Hodge decomposition, the regular decomposition and bounded commuting quasi-interpolants. An additional marking strategy is added to ensure the quasi-orthogonality. Using this quasi-orthogonality, the uniform convergence of adaptive mixed finite element methods is obtained without assuming the initial mesh size is small enough.

math.NA↗

arXiv2017

Some Error Analysis on Virtual Element Methods

Some error analysis on virtual element methods including inverse inequalities, norm equivalence, and interpolation error estimates are presented for polygonal meshes which admits a virtual quasi-uniform triangulation.

math.NA↗

arXiv2017

Cascade one-vs-rest detection network for fine-grained recognition without part annotations

Fine-grained recognition is a challenging task due to the small intra-category variances. Most of top-performing fine-grained recognition methods leverage parts of objects for better performance. Therefore, part annotations which are extremely computationally expensive are required. In this paper, we propose a novel cascaded deep CNN detection framework for fine-grained recognition which is trained to detect the whole object without considering parts. Nevertheless, most of current top-performing detection networks use the N+1 class (N object categories plus background) softmax loss, and the background category with much more training samples dominates the feature learning progress so that the features are not good for object categories with fewer samples. To bridge this gap, we introduce a cascaded structure to eliminate background and exploit a one-vs-rest loss to capture more minute variances among different subordinate categories. Experiments show that our proposed recognition framework achieves comparable performance with state-of-the-art, part-free, fine-grained recognition methods on the CUB-200-2011 Bird dataset. Moreover, our method even outperforms most of part-based methods while does not need part annotations at the training stage and is free from any annotations at test stage.

cs.CV↗

arXiv2017

Recent Developments and Future Challenges in Medical Mixed Reality

Mixed Reality (MR) is of increasing interest within technology-driven modern medicine but is not yet used in everyday practice. This situation is changing rapidly, however, and this paper explores the emergence of MR technology and the importance of its utility within medical applications. A classification of medical MR has been obtained by applying an unbiased text mining method to a database of 1,403 relevant research papers published over the last two decades. The classification results reveal a taxonomy for the development of medical MR research during this period as well as suggesting future trends. We then use the classification to analyse the technology and applications developed in the last five years. Our objective is to aid researchers to focus on the areas where technology advancements in medical MR are most needed, as well as providing medical practitioners with a useful source of reference.

cs.CV↗