arXiv subjects

Liang Wang

Publications and source records attributed to Liang Wang.

At least 631 records · Page 35Linked to original sources

arXiv2017

Irregular Convolutional Neural Networks

Convolutional kernels are basic and vital components of deep Convolutional Neural Networks (CNN). In this paper, we equip convolutional kernels with shape attributes to generate the deep Irregular Convolutional Neural Networks (ICNN). Compared to traditional CNN applying regular convolutional kernels like ${3\times3}$, our approach trains irregular kernel shapes to better fit the geometric variations of input features. In other words, shapes are learnable parameters in addition to weights. The kernel shapes and weights are learned simultaneously during end-to-end training with the standard back-propagation algorithm. Experiments for semantic segmentation are implemented to validate the effectiveness of our proposed ICNN.

cs.CV↗

arXiv2017

Mining Significant Microblogs for Misinformation Identification: An Attention-based Approach

With the rapid growth of social media, massive misinformation is also spreading widely on social media, such as microblog, and bring negative effects to human life. Nowadays, automatic misinformation identification has drawn attention from academic and industrial communities. For an event on social media usually consists of multiple microblogs, current methods are mainly based on global statistical features. However, information on social media is full of noisy and outliers, which should be alleviated. Moreover, most of microblogs about an event have little contribution to the identification of misinformation, where useful information can be easily overwhelmed by useless information. Thus, it is important to mine significant microblogs for a reliable misinformation identification method. In this paper, we propose an Attention-based approach for Identification of Misinformation (AIM). Based on the attention mechanism, AIM can select microblogs with largest attention values for misinformation identification. The attention mechanism in AIM contains two parts: content attention and dynamic attention. Content attention is calculated based textual features of each microblog. Dynamic attention is related to the time interval between the posting time of a microblog and the beginning of the event. To evaluate AIM, we conduct a series of experiments on the Weibo dataset and the Twitter dataset, and the experimental results show that the proposed AIM model outperforms the state-of-the-art methods.

cs.IR↗

arXiv2017

Cross-modal Subspace Learning for Fine-grained Sketch-based Image Retrieval

Sketch-based image retrieval (SBIR) is challenging due to the inherent domain-gap between sketch and photo. Compared with pixel-perfect depictions of photos, sketches are iconic renderings of the real world with highly abstract. Therefore, matching sketch and photo directly using low-level visual clues are unsufficient, since a common low-level subspace that traverses semantically across the two modalities is non-trivial to establish. Most existing SBIR studies do not directly tackle this cross-modal problem. This naturally motivates us to explore the effectiveness of cross-modal retrieval methods in SBIR, which have been applied in the image-text matching successfully. In this paper, we introduce and compare a series of state-of-the-art cross-modal subspace learning methods and benchmark them on two recently released fine-grained SBIR datasets. Through thorough examination of the experimental results, we have demonstrated that the subspace learning can effectively model the sketch-photo domain-gap. In addition we draw a few key insights to drive future research.

cs.CV↗

arXiv2017

FairCache: Introducing Fairness to ICN Caching - Technical Report

Information-centric networking extensively uses universal in-network caching. However, developing an efficient and fair collaborative caching algorithm for selfish caches is still an open question. In addition, the communication overhead induced by collaboration is especially poorly understood in a general network setting such as realistic ISP and Autonomous System networks. In this paper, we address these two problems by modeling the in-network caching problem as a Nash bargaining game. We show that the game is a convex optimization problem and further derive the corresponding distributed algorithm. We analytically investigate the collaboration overhead on general graph topologies, and theoretically show that collaboration has to be constrained within a small neighborhood due to its cost growing exponentially. Our proposed algorithm achieves at least 16% performance gain over its competitors on different network topologies in the evaluation, and guarantees provable convergence, Pareto efficiency and proportional fairness.

cs.NI↗

arXiv2017

Forward stimulated Brillouin scattering in silicon microring resonators

Stimulated Brillouin scattering (SBS) has been demonstrated in silicon waveguides in recent years. However, due to the weak interaction between photons and acoustic phonons in these waveguides, long interaction length is typically necessary. Here, we experimentally show that forward stimulated Brillouin scattering in a short interaction length of a 20 um radius silicon microring resonator could give 1.2 dB peak gain at only 10mW coupled pump power. The experimental results demonstrate that both optical and acoustic modes can have efficient interaction in a short interaction length. The observed Brillouin gain varies with coupled pump power in good agreement with theoretical prediction. The work shows the potential of SBS in silicon for moving the demonstrated fiber SBS applications to the integrated silicon photonics platform.

physics.optics↗

arXiv2017

Modeling Temporal Dynamics and Spatial Configurations of Actions Using Two-Stream Recurrent Neural Networks

Recently, skeleton based action recognition gains more popularity due to cost-effective depth sensors coupled with real-time skeleton estimation algorithms. Traditional approaches based on handcrafted features are limited to represent the complexity of motion patterns. Recent methods that use Recurrent Neural Networks (RNN) to handle raw skeletons only focus on the contextual dependency in the temporal domain and neglect the spatial configurations of articulated skeletons. In this paper, we propose a novel two-stream RNN architecture to model both temporal dynamics and spatial configurations for skeleton based action recognition. We explore two different structures for the temporal stream: stacked RNN and hierarchical RNN. Hierarchical RNN is designed according to human body kinematics. We also propose two effective methods to model the spatial structure by converting the spatial graph into a sequence of joints. To improve generalization of our model, we further exploit 3D transformation based data augmentation techniques including rotation and scaling transformation to transform the 3D coordinates of skeletons during training. Experiments on 3D action recognition benchmark datasets show that our method brings a considerable improvement for a variety of actions, i.e., generic actions, interaction activities and gestures.

cs.CV↗

arXiv2017

An in-depth characterisation of Bots and Humans on Twitter

Recent research has shown a substantial active presence of bots in online social networks (OSNs). In this paper we utilise our past work on studying bots (Stweeler) to comparatively analyse the usage and impact of bots and humans on Twitter, one of the largest OSNs in the world. We collect a large-scale Twitter dataset and define various metrics based on tweet metadata. We divide and filter the dataset in four popularity groups in terms of number of followers. Using a human annotation task we assign 'bot' and 'human' ground-truth labels to the dataset, and compare the annotations against an online bot detection tool for evaluation. We then ask a series of questions to discern important behavioural bot and human characteristics using metrics within and among four popularity groups. From the comparative analysis we draw important differences as well as surprising similarities between the two entities, thus paving the way for reliable classification of automated political infiltration, advertisement campaigns, and general bot detection.

cs.SI↗

arXiv2017

NIHAO XII: galactic uniformity in a ΛCDM universe

We use a sample of 83 high-resolution cosmological zoom-in simulations and a semi-analytic model (SAM) to study the stochasticity of galaxy formation in haloes ranging from dwarf to Milky Way masses. Our simulated galaxies reproduce the observed inefficiency of galaxy formation as expressed through the stellar, gas and baryonic Tully-Fisher relations. For HI velocities in the range (70 < V < 220 km/s), the scatter is just 0.08 to 0.14 dex, consistent with the observed intrinsic scatter at these scales. At low velocities (20 < V < 70 km/s), the simulated scatter is 0.2-0.25 dex, which could be tested with future observations. The scatter in the stellar mass versus dark halo velocity relation is constant for 30 < V < 180 km/s, and smaller (~0.17 dex) when using the maximum circular velocity of the dark matter only simulation, Vmax, compared to the virial velocity ($V_{200}$ or $V_{200}^{DMO}$). The scatter in stellar mass is correlated with halo concentration, and is minimized when using a circular velocity at a fixed fraction of the virial radius $\simeq 0.4 R_{200}$ or with $V_α=V_{200}^{DMO} (V_{max}^{DMO} / V_{200}^{DMO})^α$ with α~ 0.7, consistent with constraints from halo clustering. Using the SAM we show the correlation between halo formation time and concentration is essential in order to reproduce this result. This uniformity in galaxy formation efficiency we see in our hydrodynamical simulations and a semi-analytic model proves the simplicity and self-regulating nature of galaxy formation in a ΛCold Dark Matter universe.

astro-ph.GA↗

arXiv2017

BOTDA using OFDM channel estimation

A novel Brillouin optical time-domain analysis (BOTDA) system is proposed using intensity-modulated optical orthogonal frequency division multiplexing probe signal and direct detection (IM-DD-OOFDM) without frequency sweep operation. The influence of peak to average power ratio (PAPR) of OFDM probe signal on the recovery of Brillouin gain spectrum (BGS) is analyzed in theory and experiment. The complex BGS is reconstructed by channel estimation algorithm and Brillouin frequency shift (BFS) is located by curve fitting of intensity spectrum. The IM-DD-OOFDM BOTDA is demonstrated experimentally with 25m spatial resolution over 2 km standard single mode fiber.

physics.ins-det↗

arXiv2017

Multi-behavioral Sequential Prediction with Recurrent Log-bilinear Model

With the rapid growth of Internet applications, sequential prediction in collaborative filtering has become an emerging and crucial task. Given the behavioral history of a specific user, predicting his or her next choice plays a key role in improving various online services. Meanwhile, there are more and more scenarios with multiple types of behaviors, while existing works mainly study sequences with a single type of behavior. As a widely used approach, Markov chain based models are based on a strong independence assumption. As two classical neural network methods for modeling sequences, recurrent neural networks cannot well model short-term contexts, and the log-bilinear model is not suitable for long-term contexts. In this paper, we propose a Recurrent Log-BiLinear (RLBL) model. It can model multiple types of behaviors in historical sequences with behavior-specific transition matrices. RLBL applies a recurrent structure for modeling long-term contexts. It models several items in each hidden layer and employs position-specific transition matrices for modeling short-term contexts. Moreover, considering continuous time difference in behavioral history is a key factor for dynamic prediction, we further extend RLBL and replace position-specific transition matrices with time-specific transition matrices, and accordingly propose a Time-Aware Recurrent Log-BiLinear (TA-RLBL) model. Experimental results show that the proposed RLBL model and TA-RLBL model yield significant improvements over the competitive compared methods on three datasets, i.e., Movielens-1M dataset, Global Terrorism Database and Tmall dataset with different numbers of behavior types.

cs.IR↗

arXiv2017

NIHAO VII: Predictions for the galactic baryon budget in dwarf to Milky Way mass haloes

We use the NIHAO galaxy formation simulations to make predictions for the baryonic budget in present day galaxies ranging from dwarf to Milky Way masses. The sample is made of 88 independent high resolution cosmological zoom-in simulations. NIHAO galaxies reproduce key properties of observed galaxies, such as the stellar mass vs halo mass and cold gas vs stellar mass relations. Thus they make plausible predictions for the baryon budget. We present the mass fractions of stars, cold gas ($T<10^4$K), cool gas ($10^4 < T < 10^5$K), warm-hot gas ($10^5 < T < 5\times10^6$K), and hot gas (T$> 5\times10^6$K) inside the virial radius, $R_{200}$. Compared to the predicted baryon mass, using the dark halo mass and the universal baryon fraction, $f_{\rm b}\equiv Ω_{\rm b}/Ω_{\rm m}=0.15$, we find that all of our haloes are missing baryons. The missing mass has been relocated past 2 virial radii, and cool gas dominates the corona at low mass (M$_{200} < 3 \times 10^{11} M_\odot$) while the warm-hot gas dominates at high mass (M$_{200} > 3 \times 10^{11} M_\odot$). Haloes of mass $\sim 10^{10} M_\odot$ are missing $\sim 90\%$ of their baryons. More massive haloes ($\sim 10^{12} M_\odot$) retain a higher fraction of their baryons, with $\sim 30\%$ missing, consistent with recent observational estimates. Moreover, these more massive haloes reproduce the observed fraction of cold, warm-hot and hot gas. The fraction of cool gas we predict ($0.11\pm0.06$) is significantly lower than the observation from COS-HALOs (0.3-0.47), but agrees with the alternative analysis of Stern et al. 2016.

astro-ph.GA↗

arXiv2016

Millimeter Wave Communications with OAM-SM Scheme for Future Mobile Networks

The orbital angular momentum (OAM) technique provides a new degree of freedom for information transmissions in millimeter wave communications. Considering the spatial distribution characteristics of OAM beams, a new OAM spatial modulation (OAM-SM) millimeter wave communication system is first proposed for future mobile networks. Furthermore, the capacity, average bit error probability and energy efficiency of OAM-SM millimeter wave communication systems are analytically derived for performance analysis. Compared with the conventional multi-input multi-output (MIMO) millimeter wave communication systems, the maximum capacity and energy efficiency of OAM-SM millimeter wave communication systems are improved by 36% and 472.3%, respectively. Moreover, numerical results indicate that the proposed OAM-SM millimeter wave communication systems are more robust to path-loss attenuations than the conventional MIMO millimeter wave communication systems, which makes it suitable for long-range transmissions. Therefore, OAM-SM millimeter wave communication systems provide a great growth space for future mobile networks.

cs.NI↗

arXiv2016

The Pan-Pacific Planet Search VI: Giant planets orbiting HD 86950 and HD 222076

We report the detection of two new planets orbiting the K giants HD 86950 and HD 222076, based on precise radial velocities obtained with three instruments: AAT/UCLES, FEROS, and CHIRON. HD 86950b has a period of 1270$\pm$57 days at $a=2.72\pm$0.08 AU, and m sin $i=3.6\pm$0.7 Mjup. HD 222076b has $P=871\pm$19 days at $a=1.83\pm$0.03 AU, and m sin $i=1.56\pm$0.11 Mjup. These two giant planets are typical of the population of planets known to orbit evolved stars. In addition, we find a high-amplitude periodic velocity signal ($K\sim$50 m/s) in HD 29399, and show that it is due to stellar variability rather than Keplerian reflex motion. We also investigate the relation between planet occurrence and host-star metallicity for the 164-star Pan-Pacific Planet Search sample of evolved stars. In spite of the small sample of PPPS detections, we confirm the trend of increasing planet occurrence as a function of metallicity found by other studies of planets orbiting evolved stars.

astro-ph.EP↗

arXiv2016

Instance-aware Image and Sentence Matching with Selective Multimodal LSTM

Effective image and sentence matching depends on how to well measure their global visual-semantic similarity. Based on the observation that such a global similarity arises from a complex aggregation of multiple local similarities between pairwise instances of image (objects) and sentence (words), we propose a selective multimodal Long Short-Term Memory network (sm-LSTM) for instance-aware image and sentence matching. The sm-LSTM includes a multimodal context-modulated attention scheme at each timestep that can selectively attend to a pair of instances of image and sentence, by predicting pairwise instance-aware saliency maps for image and sentence. For selected pairwise instances, their representations are obtained based on the predicted saliency maps, and then compared to measure their local similarity. By similarly measuring multiple local similarities within a few timesteps, the sm-LSTM sequentially aggregates them with hidden states to obtain a final matching score as the desired global similarity. Extensive experiments show that our model can well match image and sentence with complex content, and achieve the state-of-the-art results on two public benchmark datasets.

cs.CV↗

arXiv2016

Multimodal Memory Modelling for Video Captioning

Video captioning which automatically translates video clips into natural language sentences is a very important task in computer vision. By virtue of recent deep learning technologies, e.g., convolutional neural networks (CNNs) and recurrent neural networks (RNNs), video captioning has made great progress. However, learning an effective mapping from visual sequence space to language space is still a challenging problem. In this paper, we propose a Multimodal Memory Model (M3) to describe videos, which builds a visual and textual shared memory to model the long-term visual-textual dependency and further guide global visual attention on described targets. Specifically, the proposed M3 attaches an external memory to store and retrieve both visual and textual contents by interacting with video and sentence with multiple read and write operations. First, text representation in the Long Short-Term Memory (LSTM) based text decoder is written into the memory, and the memory contents will be read out to guide an attention to select related visual targets. Then, the selected visual information is written into the memory, which will be further read out to the text decoder. To evaluate the proposed model, we perform experiments on two publicly benchmark datasets: MSVD and MSR-VTT. The experimental results demonstrate that our method outperforms the state-of-theart methods in terms of BLEU and METEOR.

cs.CV↗

arXiv2016

ICE: Information Credibility Evaluation on Social Media via Representation Learning

With the rapid growth of social media, rumors are also spreading widely on social media and bring harm to people's daily life. Nowadays, information credibility evaluation has drawn attention from academic and industrial communities. Current methods mainly focus on feature engineering and achieve some success. However, feature engineering based methods require a lot of labor and cannot fully reveal the underlying relations among data. In our viewpoint, the key elements of user behaviors for evaluating credibility are concluded as "who", "what", "when", and "how". These existing methods cannot model the correlation among different key elements during the spreading of microblogs. In this paper, we propose a novel representation learning method, Information Credibility Evaluation (ICE), to learn representations of information credibility on social media. In ICE, latent representations are learnt for modeling user credibility, behavior types, temporal properties, and comment attitudes. The aggregation of these factors in the microblog spreading process yields the representation of a user's behavior, and the aggregation of these dynamic representations generates the credibility representation of an event spreading on social media. Moreover, a pairwise learning method is applied to maximize the credibility difference between rumors and non-rumors. To evaluate the performance of ICE, we conduct experiments on a Sina Weibo data set, and the experimental results show that our ICE model outperforms the state-of-the-art methods.

cs.SI↗

arXiv2016

NIHAO VIII: Circum-galactic medium and outflows - The puzzles of HI and OVI gas distributions

We study the hot and cold circum-galactic medium (CGM) of 86 galaxies of the cosmological, hydrodynamical simulation suite NIHAO. NIHAO allows a study of how the $z=0$ CGM varies across 5 orders of magnitude of stellar mass using OVI and HI as proxies for hot and cold gas. The cool HI covering fraction and column density profiles match observations well, particularly in the inner CGM. OVI shows increasing column densities with mass, a trend seemingly echoed in the observations. As in multiple previous simulations, the OVI column densities in simulations are lower than observed and optically thick HI does not extend as far out as in observations. We take a look at the collisional ionisation fraction of OVI as a function of halo mass. We make observable predictions of the bipolarity of outflows and their effect on the general shape of the CGM. Bipolar outflows can be seen out to around 40 kpc in intermediate and low mass halos ($M_{\mathrm{halo}}<10^{11}M_{\mathrm{sun}}$), but outside that radius, the CGM is too well mixed to detect an elongated shape. Larger halos have extended gas discs beyond the stellar disc that dominate the shape of the inner CGM. The simulated CGM is remarkably spherical even in low mass simulations. The chemical enrichment of both halo and disc gas follow expected increasing trends as a function of halo mass that are well fit with power laws. These relations can be used in non-hydrodynamic models, such as semi-analytic models.

astro-ph.GA↗

arXiv2016

Context-aware Sequential Recommendation

Since sequential information plays an important role in modeling user behaviors, various sequential recommendation methods have been proposed. Methods based on Markov assumption are widely-used, but independently combine several most recent components. Recently, Recurrent Neural Networks (RNN) based methods have been successfully applied in several sequential modeling tasks. However, for real-world applications, these methods have difficulty in modeling the contextual information, which has been proved to be very important for behavior modeling. In this paper, we propose a novel model, named Context-Aware Recurrent Neural Networks (CA-RNN). Instead of using the constant input matrix and transition matrix in conventional RNN models, CA-RNN employs adaptive context-specific input matrices and adaptive context-specific transition matrices. The adaptive context-specific input matrices capture external situations where user behaviors happen, such as time, location, weather and so on. And the adaptive context-specific transition matrices capture how lengths of time intervals between adjacent behaviors in historical sequences affect the transition of global sequential features. Experimental results show that the proposed CA-RNN model yields significant improvements over state-of-the-art sequential recommendation methods and context-aware recommendation methods on two public datasets, i.e., the Taobao dataset and the Movielens-1M dataset.

cs.IR↗