arXiv subjects

Liang Wang

Publications and source records attributed to Liang Wang.

At least 577 records · Page 32Linked to original sources

arXiv2019

Semi-supervised Compatibility Learning Across Categories for Clothing Matching

Learning the compatibility between fashion items across categories is a key task in fashion analysis, which can decode the secret of clothing matching. The main idea of this task is to map items into a latent style space where compatible items stay close. Previous works try to build such a transformation by minimizing the distances between annotated compatible items, which require massive item-level supervision. However, these annotated data are expensive to obtain and hard to cover the numerous items with various styles in real applications. In such cases, these supervised methods fail to achieve satisfactory performances. In this work, we propose a semi-supervised method to learn the compatibility across categories. We observe that the distributions of different categories have intrinsic similar structures. Accordingly, the better distributions align, the closer compatible items across these categories become. To achieve the alignment, we minimize the distances between distributions with unsupervised adversarial learning, and also the distances between some annotated compatible items which play the role of anchor points to help align. Experimental results on two real-world datasets demonstrate the effectiveness of our method.

cs.IR↗

arXiv2019

Threshold Voltage Improvement and Leakage Reduction of AlGaN/GaN HEMTs Using Dual-Layer SiNx Stressors

In this work, AlGaN/GaN HEMTs with dual-layer SiNx stressors (composed of a low-stress layer and a high-stress layer) were investigated. The low-stress padding layer solved the surface damage problem caused during the deposition of the high-stress SiNx, and provided a good passivated interface. The HEMTs with dual-layer stressors showed a 1 V increase in the threshold voltage (Vth) with comparable on-current and RF current gain to those without stressors. Moreover, the off-current (I_off) was shown to be reduced by one to three orders of magnitude in the strained devices as a result of the lower electric field in AlGaN, which suppressed the gate injection current. The dual-layer stressor scheme supports strain engineering as an effective approach in the pursuit of the normally-off operation of AlGaN/GaN HEMTs.

physics.app-ph↗

arXiv2019

A Method to Determine Broadband Complex Permittivity of Thin Film Dielectric Materials up to 60 GHz

This paper describes a method to determine the complex permittivity of a thin dielectric film from finite element analysis and microstrip line measurements. Two transmission line equivalent circuit models were used for the cases of an air-filled line and a lossless line, whose distributed elements can be calculated from full wave finite element simulations. With these calculated distributed elements and microstrip line measurements, the complex permittivity was extracted. The technique utilizes a simple way to separate the dielectric loss from the measured total loss and the complex permittivity was extracted using the measured propagation constant. A rational dielectric model was employed to fit the extracted complex permittivity, which ensures causality of the final solution. The best fitting results obtained through this procedure are considered as the final permittivity results, which have shown excellent match to the data sheet values. Moreover, simulations using the fitted permittivity exhibit good agreement with the experimental propagation constant data of microstrip lines up to 60 GHz. The proposed method was demonstrated on polyimide, and it can be applied to other thin film materials.

physics.app-ph↗

arXiv2019

Learn to Allocate Resources in Vehicular Networks

Resource allocation has a direct and profound impact on the performance of vehicle-to-everything (V2X) networks. Considering the dynamic nature of vehicular environments, it is appealing to devise a decentralized strategy to perform effective resource sharing. In this paper, we exploit deep learning to promote coordination among multiple vehicles and propose a hybrid architecture consisting of centralized decision making and distributed resource sharing to maximize the long-term sum rate of all vehicles. To reduce the network signaling overhead, each vehicle uses a deep neural network to compress its own observed information that is thereafter fed back to the centralized decision-making unit, which employs a deep Q-network to allocate resources and then sends the decision results to all vehicles. We further adopt a quantization layer for each vehicle that learns to quantize the continuous feedback. Extensive simulation results demonstrate that the proposed hybrid architecture can achieve near-optimal performance. Meanwhile, there exists an optimal number of continuous feedback and binary feedback, respectively. Besides, this architecture is robust to different feedback intervals, input noise, and feedback noise.

cs.NI↗

arXiv2019

Drivers of disc tilting I: Correlations and possible drivers for Milky Way analogues

The direction of the spin vectors of disk galaxies change over time. We present the tilting rate of a sample of galaxies in the NIHAO suite of cosmological hydrodynamical simulations. Galaxies in our sample have been selected to be isolated and to have well determined spins. We compare the tilting rates to the predicted observing limit of Gaia, finding that our entire sample lies above the limit, in agreement with previous work. To test the role of dark matter and of gas we compare the weighted Pearson's correlation coefficients between the tilting rates and various properties. We find no correlation between the dark halo's tilting rate, shape, or misalignment with respect to the disc, and the tilting rate of the stellar disc. Therefore, we argue that, in the presence of gas, the dark halo plays a negligible role in the tilting of the stellar disc. On the other hand, we find a strong correlation between the tilting rate of the stellar disc and the misalignment of the cold gas warp. Adding the stellar mass fraction improves the correlation, while none of the dark matter's properties together with the cold gas misalignment improves the correlation to any significant extent. This implies that the gas cooling onto the disc is the principal driver of disc tilting.

astro-ph.GA↗

arXiv2019

Discrete effect on the anti-bounce-back boundary condition of lattice Bhatnagar-Gross-Krook model for convection-diffusion equations

The discrete effect on the boundary condition has been a fundamental topic for the lattice Boltzmann method in simulating heat and mass transfer problems. In previous works based on the halfway anti-bounce-back (ABB) boundary condition for convection-diffusion equations (CDEs), it is reported that the discrete effect cannot be commonly removed in the Bhatnagar-Gross-Krook (BGK) model except for a special value of relaxation time. Targeting this point in the present paper, we still proceed within the framework of BGK model for two-dimensional CDEs, and analyze the discrete effect on a non-halfway ABB boundary condition which incorporates the effect of the distance ratio. By analyzing an unidirectional diffusion problem with a parabolic distribution, the theoretical derivations with three different discrete velocity models show that the numerical slip is a combined function of the relaxation time and the distance ratio. Different from previous works, we definitely find that the relaxation time can be freely adjusted by the distance ratio in a proper range to eliminate the numerical slip. Some numerical simulations are carried out to validate the theoretical derivations, and the numerical results for the cases of straight and curved boundaries confirm our theoretical analysis. Finally, it should be noted that the present analysis can be extended from the BGK model to other lattice Boltzmann (LB) collision models for CDEs, which can broaden the parameter range of the relaxation time to approach 0.5.

physics.comp-ph↗

arXiv2019

Improving Description-based Person Re-identification by Multi-granularity Image-text Alignments

Description-based person re-identification (Re-id) is an important task in video surveillance that requires discriminative cross-modal representations to distinguish different people. It is difficult to directly measure the similarity between images and descriptions due to the modality heterogeneity (the cross-modal problem). And all samples belonging to a single category (the fine-grained problem) makes this task even harder than the conventional image-description matching task. In this paper, we propose a Multi-granularity Image-text Alignments (MIA) model to alleviate the cross-modal fine-grained problem for better similarity evaluation in description-based person Re-id. Specifically, three different granularities, i.e., global-global, global-local and local-local alignments are carried out hierarchically. Firstly, the global-global alignment in the Global Contrast (GC) module is for matching the global contexts of images and descriptions. Secondly, the global-local alignment employs the potential relations between local components and global contexts to highlight the distinguishable components while eliminating the uninvolved ones adaptively in the Relation-guided Global-local Alignment (RGA) module. Thirdly, as for the local-local alignment, we match visual human parts with noun phrases in the Bi-directional Fine-grained Matching (BFM) module. The whole network combining multiple granularities can be end-to-end trained without complex pre-processing. To address the difficulties in training the combination of multiple granularities, an effective step training strategy is proposed to train these granularities step-by-step. Extensive experiments and analysis have shown that our method obtains the state-of-the-art performance on the CUHK-PEDES dataset and outperforms the previous methods by a significant margin.

cs.CV↗

arXiv2019

Diffusing Your Mobile Apps: Extending In-Network Function Virtualization to Mobile Function Offloading

Motivated by the huge disparity between the limited battery capacity of user devices and the ever-growing energy demands of modern mobile apps, we propose INFv. It is the first offloading system able to cache, migrate and dynamically execute on demand functionality from mobile devices in ISP networks. It aims to bridge this gap by extending the promising NFV paradigm to mobile applications in order to exploit in-network resources. In this paper, we present the overall design, state-of-the-art technologies adopted, and various engineering details in the INFv system. We also carefully study the deployment configurations by investigating over 20K Google Play apps, as well as thorough evaluations with realistic settings. In addition to a significant improvement in battery life (up to 6.9x energy reduction) and execution time (up to 4x faster), INFv has two distinct advantages over previous systems: 1) a non-intrusive offloading mechanism transparent to existing apps; 2) an inherent framework support to effectively balance computation load and exploit the proximity of in-network resources. Both advantages together enable a scalable and incremental deployment of computation offloading framework in practical ISPs' networks.

cs.DC↗

arXiv2019

Improving Grammatical Error Correction via Pre-Training a Copy-Augmented Architecture with Unlabeled Data

Neural machine translation systems have become state-of-the-art approaches for Grammatical Error Correction (GEC) task. In this paper, we propose a copy-augmented architecture for the GEC task by copying the unchanged words from the source sentence to the target sentence. Since the GEC suffers from not having enough labeled training data to achieve high accuracy. We pre-train the copy-augmented architecture with a denoising auto-encoder using the unlabeled One Billion Benchmark and make comparisons between the fully pre-trained model and a partially pre-trained model. It is the first time copying words from the source context and fully pre-training a sequence to sequence model are experimented on the GEC task. Moreover, We add token-level and sentence-level multi-task learning for the GEC task. The evaluation results on the CoNLL-2014 test set show that our approach outperforms all recently published state-of-the-art results by a large margin. The code and pre-trained models are released at https://github.com/zhawe01/fairseq-gec.

cs.CL↗

arXiv2019

Hierarchical Graph Convolutional Networks for Semi-supervised Node Classification

Graph convolutional networks (GCNs) have been successfully applied in node classification tasks of network mining. However, most of these models based on neighborhood aggregation are usually shallow and lack the "graph pooling" mechanism, which prevents the model from obtaining adequate global information. In order to increase the receptive field, we propose a novel deep Hierarchical Graph Convolutional Network (H-GCN) for semi-supervised node classification. H-GCN first repeatedly aggregates structurally similar nodes to hyper-nodes and then refines the coarsened graph to the original to restore the representation for each node. Instead of merely aggregating one- or two-hop neighborhood information, the proposed coarsening procedure enlarges the receptive field for each node, hence more global information can be captured. The proposed H-GCN model shows strong empirical performance on various public benchmark graph datasets, outperforming state-of-the-art methods and acquiring up to 5.9% performance improvement in terms of accuracy. In addition, when only a few labeled samples are provided, our model gains substantial improvements.

cs.SI↗

arXiv2019

Progressive Cluster Purification for Transductive Few-shot Learning

Few-shot learning aims to learn to generalize a classifier to novel classes with limited labeled data. Transductive inference that utilizes unlabeled test set to deal with low-data problem has been employed for few-shot learning in recent literature. Yet, these methods do not explicitly exploit the manifold structures of semantic clusters, which is inefficient for transductive inference. In this paper, we propose a novel Progressive Cluster Purification (PCP) method for transductive few-shot learning. The PCP can progressively purify the cluster by exploring the semantic interdependency in the individual cluster space. Specifically, the PCP consists of two-level operations: inter-class classification and intra-class transduction. The inter-class classification partitions all the test samples into several clusters by comparing the test samples with the prototypes. The intra-class transduction effectively explores trustworthy test samples for each cluster by modeling data relations within a cluster as well as among different clusters. Then, it refines the prototypes to better represent the real distribution of semantic clusters. The refined prototypes are used to remeasure all the test instances and purify each cluster. Furthermore, the inter-class classification and the intra-class transduction are extremely flexible to be repeated several times to progressively purify the clusters. Experimental results are provided on two datasets: miniImageNet dataset and tieredImageNet dataset. The comparison results demonstrate the effectiveness of our approach and show that our approach outperforms the state-of-the-art methods on both datasets.

cs.CV↗

arXiv2019

NIHAO XVI: The properties and evolution of kinematically selected discs, bulges and stellar haloes

We use 25 simulated galaxies from the NIHAO project to define and characterize a variety of kinematic stellar structures: thin and thick discs, large scale single discs, classical and pseudo bulges, spheroids, inner discs, and stellar haloes. These structures have masses, spins, shapes and rotational support in good agreement with theoretical expectations and observational data. Above a dark matter halo mass of $2.5\times10^{\rm~11}M_{\rm\odot}$, all galaxies have a classical bulge and 70\% have a thin and thick disc. The kinematic (thin) discs follow a power-law relation between angular momentum and stellar mass $J_{\rm *}=3.4M_{\rm *}^{\rm1.26\pm0.06}$, in very good agreement with the prediction based on the empirical stellar-to-halo mass relation in the same mass range, and show a strong correlation between maximum `observed' rotation velocity and dark matter halo circular velocity $v_{\rm c}=6.4v_{\rm max}^{0.64\pm0.04}$. Tracing back in time these structures' progenitors, we find all to lose a fraction $1-f_j$ of their maximum angular momentum. Thin discs are significantly better at retaining their high-redshift spins ($f_j\sim0.70$) than thick ones ($f_j\sim0.40$). Stellar haloes have their progenitor baryons assembled the latest ($z_{\rm~1/2}\sim1.1$) and over the longest timescales ($τ\sim6.2$~Gyr), and have the smallest fraction of stars born in-situ ($f_{\rm in-situ}=0.35\pm0.14$). All other structures have $1.5\lesssim z_{\rm1/2}\lesssim3$, $τ=4\pm2$~Gyr and $f_{\rm in-situ}\gtrsim0.9$.

astro-ph.GA↗

arXiv2019

Richly Activated Graph Convolutional Network for Action Recognition with Incomplete Skeletons

Current methods for skeleton-based human action recognition usually work with completely observed skeletons. However, in real scenarios, it is prone to capture incomplete and noisy skeletons, which will deteriorate the performance of traditional models. To enhance the robustness of action recognition models to incomplete skeletons, we propose a multi-stream graph convolutional network (GCN) for exploring sufficient discriminative features distributed over all skeleton joints. Here, each stream of the network is only responsible for learning features from currently unactivated joints, which are distinguished by the class activation maps (CAM) obtained by preceding streams, so that the activated joints of the proposed method are obviously more than traditional methods. Thus, the proposed method is termed richly activated GCN (RA-GCN), where the richly discovered features will improve the robustness of the model. Compared to the state-of-the-art methods, the RA-GCN achieves comparable performance on the NTU RGB+D dataset. Moreover, on a synthetic occlusion dataset, the performance deterioration can be alleviated by the RA-GCN significantly.

cs.CV↗

arXiv2019

NIHAO IV: Core creation and destruction in dark matter density profiles across cosmic time

We use the NIHAO simulations to investigate the effects of baryonic physics on the time evolution of Dark Matter central density profiles. The sample is made of $\approx 70$ independent high resolution hydrodynamical simulations of galaxy formation and covers a wide mass range: 1e10< Mhalo <1e12, i.e., from dwarfs to L* . We confirm previous results on the dependence of the inner dark matter density slope, $α$, on the ratio between stellar-to-halo mass. We show that this relation holds approximately at all redshifts (with an intrinsic scatter of ~0.18 in $α$). This implies that in practically all haloes the shape of their inner density profile changes quite substantially over cosmic time, as they grow in stellar and total mass. Thus, depending on their final stellar-to-halo mass ratio, haloes can either form and keep a substantial density core (size~1 kpc), or form and then destroy the core and re-contract the halo, going back to a cuspy profile, which is even steeper than CDM predictions for massive galaxies (~1e12 Msun). We show that results from the NIHAO suite are in good agreement with recent observational measurements of $α$ in dwarf galaxies. Overall our results suggest that the notion of a universal density profile for dark matter haloes is no longer valid in the presence of galaxy formation.

astro-ph.GA↗

arXiv2019

Box-driven Class-wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation

Semantic segmentation has achieved huge progress via adopting deep Fully Convolutional Networks (FCN). However, the performance of FCN based models severely rely on the amounts of pixel-level annotations which are expensive and time-consuming. To address this problem, it is a good choice to learn to segment with weak supervision from bounding boxes. How to make full use of the class-level and region-level supervisions from bounding boxes is the critical challenge for the weakly supervised learning task. In this paper, we first introduce a box-driven class-wise masking model (BCM) to remove irrelevant regions of each class. Moreover, based on the pixel-level segment proposal generated from the bounding box supervision, we could calculate the mean filling rates of each class to serve as an important prior cue, then we propose a filling rate guided adaptive loss (FR-Loss) to help the model ignore the wrongly labeled pixels in proposals. Unlike previous methods directly training models with the fixed individual segment proposals, our method can adjust the model learning with global statistical information. Thus it can help reduce the negative impacts from wrongly labeled proposals. We evaluate the proposed method on the challenging PASCAL VOC 2012 benchmark and compare with other methods. Extensive experimental results show that the proposed method is effective and achieves the state-of-the-art results.

cs.CV↗

arXiv2019

Metal-poor stars observed with the automated planet finder telescope. I. Discovery of five carbon-enhanced metal-poor stars from LAMOST

We report on the discovery of five carbon-enhanced metal-poor (CEMP) stars in the metallicity range of $-3.3<$ [Fe/H] $<-2.4$. These stars were selected from the LAMOST DR3 low-resolution (R$ \sim$ 2,000) spectroscopic database as metal-poor candidates and followed-up with high-resolution spectroscopy (R$ \sim$110,000) with the LICK/APF. Stellar parameters and individual abundances for 25 chemical elements (from Li to Eu) are presented for the first time. These stars exhibit chemical abundance patterns that are similar to those reported in other literature studies of very and extremely metal-poor stars. One of our targets, J2114$-$0616, shows high enhancement in carbon ([C/Fe]=1.37), nitrogen ([N/Fe]= 1.88), barium ([Ba/Fe]=1.00), and europium ([Eu/Fe]=0.84). Such chemical abundance pattern suggests that J2114$-$0616 can be classified as CEMP-r/s star. In addition, the star J1054+0528 can be classified as a CEMP-rI star, with [Eu/Fe]=0.44 and [Ba/Fe]=$-$0.52. The other stars in our sample show no enhancements in neutron-capture elements and can be classified as CEMP-no stars. We also performed a kinematic and dynamical analysis of the sample stars based on Gaia DR2 data. The kinematic parameters, orbits, and binding energy of these stars, show that J2114$-$0616 is member of the outer halo population, while the remaining stars belong to the inner halo population but with an accreted origin. Collectively, these results add important constraints on the origin and evolution of CEMP stars as well as on their possible formation scenarios.

astro-ph.GA↗

arXiv2019

Relational Network for Skeleton-Based Action Recognition

With the fast development of effective and low-cost human skeleton capture systems, skeleton-based action recognition has attracted much attention recently. Most existing methods use Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) to extract spatio-temporal information embedded in the skeleton sequences for action recognition. However, these approaches are limited in the ability of relational modeling in a single skeleton, due to the loss of important structural information when converting the raw skeleton data to adapt to the input format of CNN or RNN. In this paper, we propose an Attentional Recurrent Relational Network-LSTM (ARRN-LSTM) to simultaneously model spatial configurations and temporal dynamics in skeletons for action recognition. We introduce the Recurrent Relational Network to learn the spatial features in a single skeleton, followed by a multi-layer LSTM to learn the temporal features in the skeleton sequences. Between the two modules, we design an adaptive attentional module to focus attention on the most discriminative parts in the single skeleton. To exploit the complementarity from different geometries in the skeleton for sufficient relational modeling, we design a two-stream architecture to learn the structural features among joints and lines simultaneously. Extensive experiments are conducted on several popular skeleton datasets and the results show that the proposed approach achieves better results than most mainstream methods.

cs.CV↗

arXiv2019

Augmented LiDAR Simulator for Autonomous Driving

In Autonomous Driving (AD), detection and tracking of obstacles on the roads is a critical task. Deep-learning based methods using annotated LiDAR data have been the most widely adopted approach for this. Unfortunately, annotating 3D point cloud is a very challenging, time- and money-consuming task. In this paper, we propose a novel LiDAR simulator that augments real point cloud with synthetic obstacles (e.g., cars, pedestrians, and other movable objects). Unlike previous simulators that entirely rely on CG models and game engines, our augmented simulator bypasses the requirement to create high-fidelity background CAD models. Instead, we can simply deploy a vehicle with a LiDAR scanner to sweep the street of interests to obtain the background point cloud, based on which annotated point cloud can be automatically generated. This unique "scan-and-simulate" capability makes our approach scalable and practical, ready for large-scale industrial applications. In this paper, we describe our simulator in detail, in particular the placement of obstacles that is critical for performance enhancement. We show that detectors with our simulated LiDAR point cloud alone can perform comparably (within two percentage points) with these trained with real data. Mixing real and simulated data can achieve over 95% accuracy.

cs.CV↗