arXiv subjects

Dong Liu

Publications and source records attributed to Dong Liu.

At least 343 records · Page 19Linked to original sources

arXiv2018

A Learning-based Approach to Joint Content Caching and Recommendation at Base Stations

Recommendation system is able to shape user demands, which can be used for boosting caching gain. In this paper, we jointly optimize content caching and recommendation at base stations to maximize the caching gain meanwhile not compromising the user preference. We first propose a model to capture the impact of recommendation on user demands, which is controlled by a user-specific psychological threshold. We then formulate a joint caching and recommendation problem maximizing the successful offloading probability, which is a mixed integer programming problem. We develop a hierarchical iterative algorithm to solve the problem when the threshold is known. Since the user threshold is unknown in practice, we proceed to propose an $\varepsilon$-greedy algorithm to find the solution by learning the threshold via interactions with users. Simulation results show that the proposed algorithms improve the successful offloading probability compared with prior works with/without recommendation. The $\varepsilon$-greedy algorithm learns the user threshold quickly, and achieves more than $1-\varepsilon$ of the performance obtained by the algorithm with known threshold.

cs.NI↗

arXiv2018

Caching at Base Stations with Heterogeneous User Demands and Spatial Locality

Existing proactive caching policies are designed by assuming that all users request contents with identical activity level at uniformly-distributed or known locations, among which most of the policies are optimized by assuming that user preference is identical to content popularity. However, these assumptions are not true based on recent data analysis. In this paper, we investigate what happens without these assumptions. To this end, we establish a framework to optimize caching policy for base stations exploiting heterogeneous user preference, activity level, and spatial locality. We derive success probability and average rate of each user as utility function, respectively, and obtain the optimal caching policy maximizing a weighted sum of average utility (reflecting network performance) and minimal utility of users (reflecting user fairness). To investigate the intertwined impact of individual user request behavior on caching, we provide an algorithm to synthesize user preference from given content popularity and activity level with controlled preference similarity, and validate the algorithm with real datasets. Analysis and simulation results show that exploiting individual user behavior can improve both network performance and user fairness, and the gain increases with the skewness of spatial locality, and the heterogeneity of user preference and activity level.

cs.IT↗

arXiv2018

Laser Streaming: Turning a Laser Beam into a Flow of Liquid

Transforming a laser beam into a mass flow has been a challenge both scientifically and technologically. Here we report the discovery of a new optofluidics principle and demonstrate the generation of a steady-state water flow by a pulsed laser beam through a glass window. In order to generate a flow or stream in the same path as the refracted laser beam in pure water from an arbitrary spot on the window, we first fill a glass cuvette with an aqueous solution of Au nanoparticles. A flow will emerge from the focused laser spot on the window after the laser is turned on for a few to tens of minutes, the flow remains after the colloidal solution is completely replaced by pure water. Microscopically, this transformation is made possible by an underlying plasmonic nanoparticle-decorated cavity which is self-fabricated on the glass by nanoparticle-assisted laser etching and exhibits size and shape uniquely tailored to the incident beam profile. Hydrophone signals indicate that the flow is driven via acoustic streaming by a long-lasting ultrasound wave that is resonantly generated by the laser and the cavity through the photoacoustic effect. The principle of this light-driven flow via ultrasound, i.e. photoacoustic streaming by coupling photoacoustics to acoustic streaming, is general and can be applied to any liquids, opening up new research and applications in optofluidics as well as traditional photoacoustics and acoustic streaming.

physics.flu-dyn↗

arXiv2018

Endowing Robots with Longer-term Autonomy by Recovering from External Disturbances in Manipulation through Grounded Anomaly Classification and Recovery Policies

Robot manipulation is increasingly poised to interact with humans in co-shared workspaces. Despite increasingly robust manipulation and control algorithms, failure modes continue to exist whenever models do not capture the dynamics of the unstructured environment. To obtain longer-term horizons in robot automation, robots must develop introspection and recovery abilities. We contribute a set of recovery policies to deal with anomalies produced by external disturbances as well as anomaly classification through the use of non-parametric statistics with memoized variational inference with scalable adaptation. A recovery critic stands atop of a tightly-integrated, graph-based online motion-generation and introspection system that resolves a wide range of anomalous situations. Policies, skills, and introspection models are learned incrementally and contextually in a task. Two task-level recovery policies: re-enactment and adaptation resolve accidental and persistent anomalies respectively. The introspection system uses non-parametric priors along with Markov jump linear systems and memoized variational inference with scalable adaptation to learn a model from the data. Extensive real-robot experimentation with various strenuous anomalous conditions is induced and resolved at different phases of a task and in different combinations. The system executes around-the-clock introspection and recovery and even elicited self-recovery when misclassifications occurred.

cs.RO↗

arXiv2018

DADA: Deep Adversarial Data Augmentation for Extremely Low Data Regime Classification

Deep learning has revolutionized the performance of classification, but meanwhile demands sufficient labeled data for training. Given insufficient data, while many techniques have been developed to help combat overfitting, the challenge remains if one tries to train deep networks, especially in the ill-posed extremely low data regimes: only a small set of labeled data are available, and nothing -- including unlabeled data -- else. Such regimes arise from practical situations where not only data labeling but also data collection itself is expensive. We propose a deep adversarial data augmentation (DADA) technique to address the problem, in which we elaborately formulate data augmentation as a problem of training a class-conditional and supervised generative adversarial network (GAN). Specifically, a new discriminator loss is proposed to fit the goal of data augmentation, through which both real and augmented samples are enforced to contribute to and be consistent in finding the decision boundaries. Tailored training techniques are developed accordingly. To quantitatively validate its effectiveness, we first perform extensive simulations to show that DADA substantially outperforms both traditional data augmentation and a few GAN-based options. We then extend experiments to three real-world small labeled datasets where existing data augmentation and/or transfer learning strategies are either less effective or infeasible. All results endorse the superior capability of DADA in enhancing the generalization ability of deep networks trained in practical extremely low data regimes. Source code is available at https://github.com/SchafferZhang/DADA.

cs.CV↗

arXiv2018

IGCV3: Interleaved Low-Rank Group Convolutions for Efficient Deep Neural Networks

In this paper, we are interested in building lightweight and efficient convolutional neural networks. Inspired by the success of two design patterns, composition of structured sparse kernels, e.g., interleaved group convolutions (IGC), and composition of low-rank kernels, e.g., bottle-neck modules, we study the combination of such two design patterns, using the composition of structured sparse low-rank kernels, to form a convolutional kernel. Rather than introducing a complementary condition over channels, we introduce a loose complementary condition, which is formulated by imposing the complementary condition over super-channels, to guide the design for generating a dense convolutional kernel. The resulting network is called IGCV3. We empirically demonstrate that the combination of low-rank and sparse kernels boosts the performance and the superiority of our proposed approach to the state-of-the-arts, IGCV2 and MobileNetV2 over image classification on CIFAR and ImageNet and object detection on COCO.

cs.CV↗

arXiv2018

Blockchain in the Eyes of Developers

The popularity of blockchain technology continues to grow rapidly in both industrial and academic fields. Most studies of blockchain focus on the improvements of security, usability, or efficiency of blockchain protocols, or the applications of blockchain in finance, Internet of Things, or public services. However, few of them could reveal the concerns of front-line developers and the situations of blockchain in practice. In this article, we investigate how developers use and discuss blockchain with a case study of Stack Overflow posts. We find blockchain is a relatively new topic in Stack Overflow but it is rising to popularity. We detect 13 types of questions that developers post in Stack Overflow and identify 45 blockchain relevant entities (e.g., frameworks, libraries, or tools) for building blockchain applications. These findings may help blockchain project communities to know where to improve and help novices to know where to start.

cs.SE↗

arXiv2018

Designing Planar, Ultra-Thin, Broad-Band and Material-Versatile Solar Absorbers via Bound-Electron and Exciton Absorption

Ultrathin planar absorbing layers, including semiconductor and metal films, and 2D materials, are promising building blocks for solar energy harvesting devices but poor light absorption has been a critical issue. Although interference in ultrathin absorbing layers has been studied to realize near perfect absorption at a specific wavelength, achieving high broadband absorption still remains challenging. Here, we both theoretically and experimentally demonstrated a method to tune not only reflection phase shift but also electromagnetic energy dissipation to design broadband solar absorber with simple planar structure consisting of an ultrathin absorbing layer separated from the metallic substrate by a transparent layer. We explicitly identified by deriving a new formulism that the absorbing material with refractive index proportional to the wavelength as well as extinction coefficient independent of the wavelength, is the ideal building block to create ultrathin planar broadband absorbers. To demonstrate the general strategy for naturally available absorbing materials in both high-loss (refractory metals) and weak-absorption (2D materials) regimes, we leveraged the bound-electron interband transition with a broad Lorentz oscillator peak to design a solar thermal absorber based on a ultrathin Cr layer; and leveraged the strong exciton attributed to the spin-orbit coupling for the spectrum near the band edge, and the bound-electron interband transition for shorter wavelengths, to design a solar photovoltaic absorber based on a atomically thin MoS2 layer. Furthermore, our designed ultrathin broadband solar absorbers with planar structures have comparable absorption properties compared to the absorbers with nanopatterns. Our proposed design strategies pave the way to novel nanometer thick energy harvesting and optoelectronic devices with simple planar structures.

physics.app-ph↗

arXiv2018

Fully Convolutional Adaptation Networks for Semantic Segmentation

The recent advances in deep neural networks have convincingly demonstrated high capability in learning vision models on large datasets. Nevertheless, collecting expert labeled datasets especially with pixel-level annotations is an extremely expensive process. An appealing alternative is to render synthetic data (e.g., computer games) and generate ground truth automatically. However, simply applying the models learnt on synthetic images may lead to high generalization error on real images due to domain shift. In this paper, we facilitate this issue from the perspectives of both visual appearance-level and representation-level domain adaptation. The former adapts source-domain images to appear as if drawn from the "style" in the target domain and the latter attempts to learn domain-invariant representations. Specifically, we present Fully Convolutional Adaptation Networks (FCAN), a novel deep architecture for semantic segmentation which combines Appearance Adaptation Networks (AAN) and Representation Adaptation Networks (RAN). AAN learns a transformation from one domain to the other in the pixel space and RAN is optimized in an adversarial learning manner to maximally fool the domain discriminator with the learnt source and target representations. Extensive experiments are conducted on the transfer from GTA5 (game videos) to Cityscapes (urban street scenes) on semantic segmentation and our proposal achieves superior results when comparing to state-of-the-art unsupervised adaptation techniques. More remarkably, we obtain a new record: mIoU of 47.5% on BDDS (drive-cam videos) in an unsupervised setting.

cs.CV↗

arXiv2018

When Exploiting Individual User Preference Is Beneficial for Caching at Base Stations

Most of prior works optimize caching policies based on the following assumptions: 1) every user initiates request according to content popularity, 2) all users are with the same active level, and 3) users are uniformly located in the considered region. In practice, these assumptions are often not true. In this paper, we explore the benefit of optimizing caching policies for base stations by exploiting user preference considering the spatial locality and different active level of users. We obtain optimal caching policies, respectively minimizing the download delay averaged over all file requests and user locations in the network (namely network average delay), and minimizing the maximal weighted download delay averaged over the file requests and location of each user (namely maximal weighted user average delay), as well as minimizing the weighted sum of both. The analysis and simulation results show that exploiting heterogeneous user preference and active level can improve user fairness, and can also improve network performance when users are with spatial locality.

cs.IT↗

arXiv2018

OpenQSEI: a MATLAB package for Quasi Static Elasticity Imaging

Quasi Static Elasticity Imaging (QSEI) aims to computationally reconstruct the inhomogeneous distribution of the elastic modulus using a measured displacement field. QSEI is a well-established imaging modality used in medical imaging for localizing tissue abnormalities. More recently, QSEI has shown promise in applications of structural health monitoring and materials characterization. Despite the growing usage of QSEI in multiple fields, fully-executable open source packages are not readily available. To address this, OpenQSEI is developed using a MATLAB platform and shared in an online community setting for continual algorithmic improvement. In this article, we describe the mathematical background of QSEI and demonstrate the basic functionalities of OpenQSEI with examples.

physics.comp-ph↗

arXiv2017

Human Pose Estimation using Global and Local Normalization

In this paper, we address the problem of estimating the positions of human joints, i.e., articulated pose estimation. Recent state-of-the-art solutions model two key issues, joint detection and spatial configuration refinement, together using convolutional neural networks. Our work mainly focuses on spatial configuration refinement by reducing variations of human poses statistically, which is motivated by the observation that the scattered distribution of the relative locations of joints e.g., the left wrist is distributed nearly uniformly in a circular area around the left shoulder) makes the learning of convolutional spatial models hard. We present a two-stage normalization scheme, human body normalization and limb normalization, to make the distribution of the relative joint locations compact, resulting in easier learning of convolutional spatial models and more accurate pose estimation. In addition, our empirical results show that incorporating multi-scale supervision and multi-scale fusion into the joint detection network is beneficial. Experiment results demonstrate that our method consistently outperforms state-of-the-art methods on the benchmarks.

cs.CV↗

arXiv2017

Neural network-based arithmetic coding of intra prediction modes in HEVC

In both H.264 and HEVC, context-adaptive binary arithmetic coding (CABAC) is adopted as the entropy coding method. CABAC relies on manually designed binarization processes as well as handcrafted context models, which may restrict the compression efficiency. In this paper, we propose an arithmetic coding strategy by training neural networks, and make preliminary studies on coding of the intra prediction modes in HEVC. Instead of binarization, we propose to directly estimate the probability distribution of the 35 intra prediction modes with the adoption of a multi-level arithmetic codec. Instead of handcrafted context models, we utilize convolutional neural network (CNN) to perform the probability estimation. Simulation results show that our proposed arithmetic coding leads to as high as 9.9% bits saving compared with CABAC.

cs.MM↗

arXiv2017

229 nm UV LEDs using p-type silicon for increased hole injection

Ultraviolet (UV) light emission at 229 nm wavelength from diode structures based on AlN/Al0.77Ga0.23N quantum wells and using p-type Si to significantly increase hole injection was reported. Both electrical and optical characteristics were measured. Owing to the large concentration of holes from p-Si and efficient hole injection, no efficiency droop was observed up to a current density of 76 A/cm2 under continuous wave operation and without external thermal management. An optical output power of 160 uW was obtained with corresponding external quantum efficiency of 0.027%. This study demonstrates that by adopting p-type Si nanomembrane contacts as hole injector, practical levels of hole injection can be realized in UV light-emitting diodes with very high Al composition AlGaN quantum wells, enabling emission wavelengths and power levels that were previously inaccessible using traditional p-i-n structures with poor hole injection efficiency.

physics.app-ph↗

arXiv2017

Convolutional Neural Network-Based Block Up-sampling for Intra Frame Coding

Inspired by the recent advances of image super-resolution using convolutional neural network (CNN), we propose a CNN-based block up-sampling scheme for intra frame coding. A block can be down-sampled before being compressed by normal intra coding, and then up-sampled to its original resolution. Different from previous studies on down/up-sampling-based coding, the up-sampling methods in our scheme have been designed by training CNN instead of hand-crafted. We explore a new CNN structure for up-sampling, which features deconvolution of feature maps, multi-scale fusion, and residue learning, making the network both compact and efficient. We also design different networks for the up-sampling of luma and chroma components, respectively, where the chroma up-sampling CNN utilizes the luma information to boost its performance. In addition, we design a two-stage up-sampling process, the first stage being within the block-by-block coding loop, and the second stage being performed on the entire frame, so as to refine block boundaries. We also empirically study how to set the coding parameters of down-sampled blocks for pursuing the frame-level rate-distortion optimization. Our proposed scheme is implemented into the High Efficiency Video Coding (HEVC) reference software, and a comprehensive set of experiments have been performed to evaluate our methods. Experimental results show that our scheme achieves significant bits saving compared with HEVC anchor especially at low bit rates, leading to on average 5.5% BD-rate reduction on common test sequences and on average 9.0% BD-rate reduction on ultra high definition (UHD) test sequences.

cs.MM↗

arXiv2017

UVC LEDs on Bulk AlN Substrates Using Silicon Nanomembranes for Efficient Hole Injection

As UV LEDs are explored at shorter wavelengths (< 280 nm) into the UVC spectral range, the crystalline quality of epitaxial AlGaN films with high Al compositions and inefficient hole injection from p-type AlGaN severely limit the LED performance and development. In this work, we report on 237 nm light emission with a record light output power of 265 uW from AlN/Al0.72Ga028N multiple quantum well UVC LEDs using bulk AlN substrates and p-type silicon nanomembrane contact layers for significantly improved AlGaN film quality and hole injection, respectively. No intensity degradation or efficiency droop was observed up to a current density of 245 A/cm2, which is attributed to the low dislocation density within AlGaN films, the large concentration of holes from p-Si, and efficient hole-transport to the active region. Additionally, the emission peak at 237 nm is dominant across the electroluminescence spectrum with no significant parasitic emissions observed. This study demonstrates the feasibility of using p-Si as a hole injector for UVC LEDs, which can be extended to even shorter wavelengths where hole injection from chemically doped AlGaN layers is not feasible.

physics.app-ph↗

arXiv2017

Offline software for the DAMPE experiment

A software system has been developed for the DArk Matter Particle Explorer (DAMPE) mission, a satellite-based experiment. The DAMPE software is mainly written in C++ and steered using Python script. This article presents an overview of the DAMPE offline software, including the major architecture design and specific implementation for simulation, calibration and reconstruction. The whole system has been successfully applied to DAMPE data analysis, based on which some results from simulation and beam test experiments are obtained and presented.

astro-ph.IM↗

arXiv2017

Caching Policy Toward Maximal Success Probability and Area Spectral Efficiency of Cache-enabled HetNets

In this paper, we investigate the optimal caching policy respectively maximizing the success probability and area spectral efficiency (ASE) in a cache-enabled heterogeneous network (HetNet) where a tier of multi-antenna macro base stations (MBSs) is overlaid with a tier of helpers with caches. Under the probabilistic caching framework, we resort to stochastic geometry theory to derive the success probability and ASE. After finding the optimal caching policies, we analyze the impact of critical system parameters and compare the ASE with traditional HetNet where the MBS tier is overlaid by a tier of pico BSs (PBSs) with limited-capacity backhaul. Analytical and numerical results show that the optimal caching probability is less skewed among helpers to maximize the success probability when the ratios of MBS-to-helper density, MBS-to-helper transmit power, user-to-helper density, or the rate requirement are small, but is more skewed to maximize the ASE in general. Compared with traditional HetNet, the helper density is much lower than the PBS density to achieve the same target ASE. The helper density can be reduced by increasing cache size. With given total cache size within an area, there exists an optimal helper node density that maximizes the ASE.

cs.IT↗