Despite rapid advances in face recognition, there remains a clear gap between the performance of still image-based face recognition and video-based face recognition, due to the vast difference in visual quality between the domains and the difficulty of curating diverse large-scale video datasets. This paper addresses both of those challenges, through an image to video feature-level domain adaptation approach, to learn discriminative video frame representations. The framework utilizes large-scale unlabeled video data to reduce the gap between different domains while transferring discriminative knowledge from large-scale labeled still images. Given a face recognition network that is pretrained in the image domain, the adaptation is achieved by (i) distilling knowledge from the network to a video adaptation network through feature matching, (ii) performing feature restoration through synthetic data augmentation and (iii) learning a domain-invariant feature through a domain adversarial discriminator. We further improve performance through a discriminator-guided feature fusion that boosts high-quality frames while eliminating those degraded by video domain-specific factors. Experiments on the YouTube Faces and IJB-A datasets demonstrate that each module contributes to our feature-level domain adaptation framework and substantially improves video face recognition performance to achieve state-of-the-art accuracy. We demonstrate qualitatively that the network learns to suppress diverse artifacts in videos such as pose, illumination or occlusion without being explicitly trained for them.
Hybrid precoding has been recently proposed as a cost-effective transceiver solution for millimeter wave (mm-wave) systems. The analog component in such precoders, which is composed of a phase shifter network, is the key differentiating element in contrast to conventional fully digital precoders. While a large number of phase shifters with unquantized phases are commonly assumed in existing works, in practice the phase shifters should be discretized with a coarse quantization, and their number should be reduced to a minimum due to cost and power consideration. In this paper, we propose a new hybrid precoder implementation using a small number of phase shifters with quantized and fixed phases, i.e., a fixed phase shifter (FPS) implementation, which significantly reduces the cost and hardware complexity. In addition, a dynamic switch network is proposed to enhance the spectral efficiency. Based on the proposed FPS implementation, an effective alternating minimization (AltMin) algorithm is developed with closed-form solutions in each iteration. Simulation results show that the proposed algorithm with the FPS implementation outperforms existing ones. More importantly, it needs much fewer phase shifters than existing hybrid precoder proposals, e.g., $\sim$10 fixed phase shifters are sufficient for practically relevant system settings.
Materials design and development typically takes several decades from the initial discovery to commercialization with the traditional trial and error development approach. With the accumulation of data from both experimental and computational results, data based machine learning becomes an emerging field in materials discovery, design and property prediction. This manuscript reviews the history of materials science as a disciplinary the most common machine learning method used in materials science, and specifically how they are used in materials discovery, design, synthesis and even failure detection and analysis after materials are deployed in real application. Finally, the limitations of machine learning for application in materials science and challenges in this emerging field is discussed.
Jun 07 2017 cs.CL
We present a general-purpose tagger based on convolutional neural networks (CNN), used for both composing word vectors and encoding context information. The CNN tagger is robust across different tagging tasks: without task-specific tuning of hyper-parameters, it achieves state-of-the-art results in part-of-speech tagging, morphological tagging and supertagging. The CNN tagger is also robust against the out-of-vocabulary problem, it performs well on artificially unnormalized texts.
Jun 01 2017 cs.CL
We present a transition-based dependency parser that uses a convolutional neural network to compose word representations from characters. The character composition model shows great improvement over the word-lookup model, especially for parsing agglutinative languages. These improvements are even better than using pre-trained word embeddings from extra data. On the SPMRL data sets, our system outperforms the previous best greedy parser (Ballesteros et al., 2015) by a margin of 3% on average.
Hybrid precoding is a cost-effective approach to support directional transmissions for millimeter wave (mm-wave) communications, and its design challenge mainly lies in the analog component which consists of a network of phase shifters. The partially-connected structure employs a small number of phase shifters and therefore serves as an energy efficient solution for hybrid precoding. In this paper, we propose a double phase shifter (DPS) implementation for the phase shifter network in the partially-connected structure, which allows more tractable and flexible hybrid precoder design. In particular, the hybrid precoder design is identified as an eigenvalue problem. To further enhance the performance, dynamic mapping from radio frequency (RF) chains to antennas is proposed, for which a greedy algorithm and a modified K-means algorithm are developed. Simulation results demonstrate the performance gains of the proposed hybrid precoding algorithms with the DPS implementation over existing ones. Given its low hardware complexity and high spectral efficiency, the proposed structure is a promising candidate for 5G mm-wave systems.
Apr 21 2017 cs.CV
Despite recent advances in face recognition using deep learning, severe accuracy drops are observed for large pose variations in unconstrained environments. Learning pose-invariant features is one solution, but needs expensively labeled large-scale data and carefully designed feature learning algorithms. In this work, we focus on frontalizing faces in the wild under various head poses, including extreme profile views. We propose a novel deep 3D Morphable Model (3DMM) conditioned Face Frontalization Generative Adversarial Network (GAN), termed as FF-GAN, to generate neutral head pose face images. Our framework differs from both traditional GANs and 3DMM based modeling. Incorporating 3DMM into the GAN structure provides shape and appearance priors for fast convergence with less training data, while also supporting end-to-end training. The 3DMM-conditioned GAN employs not only the discriminator and generator loss but also a new masked symmetry loss to retain visual quality under occlusions, besides an identity loss to recover high frequency information. Experiments on face recognition, landmark localization and 3D reconstruction consistently show the advantage of our frontalization method on faces in the wild datasets.
Apr 11 2017 cs.AR
Putting the DRAM on the same package with a processor enables several times higher memory bandwidth than conventional off-package DRAM. Yet, the latency of in-package DRAM is not appreciably lower than that of off-package DRAM. A promising use of in-package DRAM is as a large cache. Unfortunately, most previous DRAM cache designs mainly optimize for hit latency and do not consider off-chip bandwidth efficiency as a first-class design constraint. Hence, as we show in this paper, these designs are suboptimal for use with in-package DRAM. We propose a new DRAM cache design, Banshee, that optimizes for both in- and off-package DRAM bandwidth efficiency without degrading access latency. The key ideas are to eliminate the in-package DRAM bandwidth overheads due to costly tag accesses through virtual memory mechanism and to incorporate a bandwidth-aware frequency-based replacement policy that is biased to reduce unnecessary traffic to off-package DRAM. Our extensive evaluation shows that Banshee provides significant performance improvement and traffic reduction over state-of-the-art latency-optimized DRAM cache designs.
Researchers often summarize their work in the form of scientific posters. Posters provide a coherent and efficient way to convey core ideas expressed in scientific papers. Generating a good scientific poster, however, is a complex and time consuming cognitive task, since such posters need to be readable, informative, and visually aesthetic. In this paper, for the first time, we study the challenging problem of learning to generate posters from scientific papers. To this end, a data-driven framework, that utilizes graphical models, is proposed. Specifically, given content to display, the key elements of a good poster, including attributes of each panel and arrangements of graphical elements are learned and inferred from data. During the inference stage, an MAP inference framework is employed to incorporate some design principles. In order to bridge the gap between panel attributes and the composition within each panel, we also propose a recursive page splitting algorithm to generate the panel layout for a poster. To learn and validate our model, we collect and release a new benchmark dataset, called NJU-Fudan Paper-Poster dataset, which consists of scientific papers and corresponding posters with exhaustively labelled panels and attributes. Qualitative and quantitative results indicate the effectiveness of our approach.
Millimeter wave (mm-wave) communications is considered a promising technology for 5G networks. Exploiting beamforming gains with large-scale antenna arrays to combat the increased path loss at mm-wave bands is one of its defining features. However, previous works on mm-wave network analysis usually adopted oversimplified antenna patterns for tractability, which can lead to significant deviation from the performance with actual antenna patterns. In this paper, using tools from stochastic geometry, we carry out a comprehensive investigation on the impact of directional antenna arrays in mm-wave networks. We first present a general and tractable framework for coverage analysis with arbitrary distributions for interference power and arbitrary antenna patterns. It is then applied to mm-wave ad hoc and cellular networks, where two sophisticated antenna patterns with desirable accuracy and analytical tractability are proposed to approximate the actual antenna pattern. Compared with previous works, the proposed approximate antenna patterns help to obtain more insights on the role of directional antenna arrays in mm-wave networks. In particular, it is shown that the coverage probabilities of both types of networks increase as a non-decreasing concave function with the antenna array size. The analytical results are verified to be effective and reliable through simulations, and numerical results also show that large-scale antenna arrays are required for satisfactory coverage in mm-wave networks.
Densifying the network and deploying more antennas at each access point are two principal ways to boost the capacity of wireless networks. However, due to the complicated distributions of random signal and interference channel gains, largely induced by various space-time processing techniques, it is highly challenging to quantitatively characterize the performance of dense multi-antenna networks. In this paper, using tools from stochastic geometry, a tractable framework is proposed for the analytical evaluation of such networks. The major result is an innovative representation of the coverage probability, as an induced $\ell_1$-norm of a Toeplitz matrix. This compact representation incorporates lots of existing analytical results on single- and multi-antenna networks as special cases, and its evaluation is almost as simple as the single-antenna case with Rayleigh fading. To illustrate its effectiveness, we apply the proposed framework to investigate two kinds of prevalent dense wireless networks, i.e., physical layer security aware networks and millimeter-wave networks. In both examples, in addition to tractable analytical results of relevant performance metrics, insightful design guidelines are also analytically obtained.
Feb 13 2017 cs.CV
Deep neural networks (DNNs) trained on large-scale datasets have recently achieved impressive improvements in face recognition. But a persistent challenge remains to develop methods capable of handling large pose variations that are relatively underrepresented in training data. This paper presents a method for learning a feature representation that is invariant to pose, without requiring extensive pose coverage in training data. We first propose to generate non-frontal views from a single frontal face, in order to increase the diversity of training data while preserving accurate facial details that are critical for identity discrimination. Our next contribution is to seek a rich embedding that encodes identity features, as well as non-identity ones such as pose and landmark locations. Finally, we propose a new feature reconstruction metric learning to explicitly disentangle identity and pose, by demanding alignment between the feature reconstructions through various combinations of identity and pose features, which is obtained from two images of the same subject. Experiments on both controlled and in-the-wild face datasets, such as MultiPIE, 300WLP and the profile view database CFP, show that our method consistently outperforms the state-of-the-art, especially on images with large head pose variations. Detail results and resource are referred to https://sites.google.com/site/xipengcshomepage/iccv2017
Hybrid precoding is a cost-effective approach to support directional transmissions for millimeter wave (mmWave) communications. While existing works on hybrid precoding mainly focus on single-user single-carrier transmission, in practice multicarrier transmission is needed to combat the much increased bandwidth, and multiuser MIMO can provide additional spatial multiplexing gains. In this paper, we propose a new hybrid precoding structure for multiuser OFDM mmWave systems, which greatly simplifies the hybrid precoder design and is able to approach the performance of the fully digital precoder. In particular, two groups of phase shifters are combined to map the signals from radio frequency (RF) chains to antennas. Then an effective hybrid precoding algorithm based on alternating minimization (AltMin) is proposed, which will alternately optimize the digital and analog precoders. A major algorithmic innovation is a LASSO formulation for the analog precoder, which yields computationally efficient algorithms. Simulation results will show the performance gain of the proposed algorithm. Moreover, it will reveal that canceling the interuser interference is critical in multiuser OFDM hybrid precoding systems.
Dec 09 2016 cs.CV
Monocular 3D object parsing is highly desirable in various scenarios including occlusion reasoning and holistic scene interpretation. We present a deep convolutional neural network (CNN) architecture to localize semantic parts in 2D image and 3D space while inferring their visibility states, given a single RGB image. Our key insight is to exploit domain knowledge to regularize the network by deeply supervising its hidden layers, in order to sequentially infer intermediate concepts associated with the final task. To acquire training data in desired quantities with ground truth 3D shape and relevant concepts, we render 3D object CAD models to generate large-scale synthetic data and simulate challenging occlusion configurations between objects. We train the network only on synthetic data and demonstrate state-of-the-art performances on real image benchmarks including an extended version of KITTI, PASCAL VOC, PASCAL3D+ and IKEA for 2D and 3D keypoint localization and instance segmentation. The empirical results substantiate the utility of our deep supervision scheme by demonstrating effective transfer of knowledge from synthetic data to real images, resulting in less overfitting compared to standard end-to-end training.
Visible light communication (VLC) could provide short-range optical wireless communication together with illumination using LED lighting. However, conventional forward error correction (FEC) codes for reliable communication do not have the features for dimming support and flicker mitigation which are required in VLC for the main functionality of lighting. Therefore, auxiliary coding techniques are usually needed, which eventually reduce the coding efficiency and increase the complexity. In this paper, a polar codes-based FEC coding scheme for dimmable VLC is proposed to increase the coding efficiency and simplify the coding structure. Experimental results show that the proposed scheme has the following advantages: 1) equal probability of 1's and 0's in codewords, which is inherently supporting 50% dimming balance; 2) short run length property (about 90% bits have runs shorter than 5) which can avoid flickers and additional run-length limited line coding; 3) higher coding efficiency about twofold than that of other coding schemes; 4) capacity achieving error correction performance with low-complexity encoding and decoding, which is about 3 dB higher coding gain than that of RS(64,32) in IEEE standard for dimming ratio 50% and about 1 dB higher coding gain than that of LDPC codes for dimming ratio 25% (or 75%).
Money laundering is a major global problem, enabling criminal organisations to hide their ill-gotten gains and to finance further operations. Prevention of money laundering is seen as a high priority by many governments, however detection of money laundering without prior knowledge of predicate crimes remains a significant challenge. Previous detection systems have tended to focus on individuals, considering transaction histories and applying anomaly detection to identify suspicious behaviour. However, money laundering involves groups of collaborating individuals, and evidence of money laundering may only be apparent when the collective behaviour of these groups is considered. In this paper we describe a detection system that is capable of analysing group behaviour, using a combination of network analysis and supervised learning. This system is designed for real-world application and operates on networks consisting of millions of interacting parties. Evaluation of the system using real-world data indicates that suspicious activity is successfully detected. Importantly, the system exhibits a low rate of false positives, and is therefore suitable for use in a live intelligence environment.
Aug 03 2016 cs.SI
The publication of fake reviews by parties with vested interests has become a severe problem for consumers who use online product reviews in their decision making. To counter this problem a number of methods for detecting these fake reviews, termed opinion spam, have been proposed. However, to date, many of these methods focus on analysis of review text, making them unsuitable for many review systems where accom-panying text is optional, or not possible. Moreover, these approaches are often computationally expensive, requiring extensive resources to handle text analysis over the scale of data typically involved. In this paper, we consider opinion spammers manipulation of average ratings for products, focusing on dif-ferences between spammer ratings and the majority opinion of honest reviewers. We propose a lightweight, effective method for detecting opinion spammers based on these differences. This method uses binomial regression to identify reviewers having an anomalous proportion of ratings that deviate from the majority opinion. Experiments on real-world and synthetic data show that our approach is able to successfully iden-tify opinion spammers. Comparison with the current state-of-the-art approach, also based only on ratings, shows that our method is able to achieve similar detection accuracy while removing the need for assump-tions regarding probabilities of spam and non-spam reviews and reducing the heavy computation required for learning.
Anomalies in online social networks can signify irregular, and often illegal behaviour. Anomalies in online social networks can signify irregular, and often illegal behaviour. Detection of such anomalies has been used to identify malicious individuals, including spammers, sexual predators, and online fraudsters. In this paper we survey existing computational techniques for detecting anomalies in online social networks. We characterise anomalies as being either static or dynamic, and as being labelled or unlabelled, and survey methods for detecting these different types of anomalies. We suggest that the detection of anomalies in online social networks is composed of two sub-processes; the selection and calculation of network features, and the classification of observations from this feature space. In addition, this paper provides an overview of the types of problems that anomaly detection can address and identifies key areas of future research.
Here we study non-convex composite optimization: first, a finite-sum of smooth but non-convex functions, and second, a general function that admits a simple proximal mapping. Most research on stochastic methods for composite optimization assumes convexity or strong convexity of each function. In this paper, we extend this problem into the non-convex setting using variance reduction techniques, such as prox-SVRG and prox-SAGA. We prove that, with a constant step size, both prox-SVRG and prox-SAGA are suitable for non-convex composite optimization, and help the problem converge to a stationary point within $O(1/\epsilon)$ iterations. That is similar to the convergence rate seen with the state-of-the-art RSAG method and faster than stochastic gradient descent. Our analysis is also extended into the min-batch setting, which linearly accelerates the convergence. To the best of our knowledge, this is the first analysis of convergence rate of variance-reduced proximal stochastic gradient for non-convex composite optimization.
May 04 2016 cs.CV
We propose a novel cascaded framework, namely deep deformation network (DDN), for localizing landmarks in non-rigid objects. The hallmarks of DDN are its incorporation of geometric constraints within a convolutional neural network (CNN) framework, ease and efficiency of training, as well as generality of application. A novel shape basis network (SBN) forms the first stage of the cascade, whereby landmarks are initialized by combining the benefits of CNN features and a learned shape basis to reduce the complexity of the highly nonlinear pose manifold. In the second stage, a point transformer network (PTN) estimates local deformation parameterized as thin-plate spline transformation for a finer refinement. Our framework does not incorporate either handcrafted features or part connectivity, which enables an end-to-end shape prediction pipeline during both training and testing. In contrast to prior cascaded networks for landmark localization that learn a mapping from feature space to landmark locations, we demonstrate that the regularization induced through geometric priors in the DDN makes it easier to train, yet produces superior results. The efficacy and generality of the architecture is demonstrated through state-of-the-art performances on several benchmarks for multiple tasks such as facial landmark localization, human body pose estimation and bird part localization.
Feb 22 2016 cs.DC
Hierarchical ring networks, which hierarchically connect multiple levels of rings, have been proposed in the past to improve the scalability of ring interconnects, but past hierarchical ring designs sacrifice some of the key benefits of rings by introducing more complex in-ring buffering and buffered flow control. Our goal in this paper is to design a new hierarchical ring interconnect that can maintain most of the simplicity of traditional ring designs (no in-ring buffering or buffered flow control) while achieving high scalability as more complex buffered hierarchical ring designs. Our design, called HiRD (Hierarchical Rings with Deflection), includes features that allow us to mostly maintain the simplicity of traditional simple ring topologies while providing higher energy efficiency and scalability. First, HiRD does not have any buffering or buffered flow control within individual rings, and requires only a small amount of buffering between the ring hierarchy levels. When inter-ring buffers are full, our design simply deflects flits so that they circle the ring and try again, which eliminates the need for in-ring buffering. Second, we introduce two simple mechanisms that provides an end-to-end delivery guarantee within the entire network without impacting the critical path or latency of the vast majority of network traffic. HiRD attains equal or better performance at better energy efficiency than multiple versions of both a previous hierarchical ring design and a traditional single ring design. We also analyze our design's characteristics and injection and delivery guarantees. We conclude that HiRD can be a compelling design point that allows higher energy efficiency and scalability while retaining the simplicity and appeal of conventional ring-based designs.
Millimeter wave (mmWave) communications has been regarded as a key enabling technology for 5G networks. In contrast to conventional multiple-input-multiple-output (MIMO) systems, precoding in mmWave MIMO cannot be performed entirely at baseband using digital precoders, as only a limited number of signal mixers and analog-to-digital converters (ADCs) can be supported considering their cost and power consumption. As a cost-effective alternative, a hybrid precoding transceiver architecture, combining a digital precoder and an analog precoder, has recently received considerable attention. However, the optimal design of such hybrid precoders has not been fully understood. In this paper, treating the hybrid precoder design as a matrix factorization problem, effective alternating minimization (AltMin) algorithms will be proposed for two different hybrid precoding structures, i.e., the fully-connected and partially-connected structures. In particular, for the fully-connected structure, an AltMin algorithm based on manifold optimization is proposed to approach the performance of the fully digital precoder, which, however, has a high complexity. Thus, a low-complexity AltMin algorithm is then proposed, by enforcing an orthogonal constraint on the digital precoder. Furthermore, for the partially-connected structure, an AltMin algorithm is also developed with the help of semidefinite relaxation. For practical implementation, the proposed AltMin algorithms are further extended to the broadband setting with orthogonal frequency division multiplexing (OFDM) modulation. Simulation results will demonstrate significant performance gains of the proposed AltMin algorithms over existing hybrid precoding algorithms. Moreover, based on the proposed algorithms, simulation comparisons between the two hybrid precoding structures will provide valuable design insights.
Artificial companion agents are defined as hardware or software entities designed to provide companionship to a person. The senior population are facing a special demand for companionship. Artificial companion agents have been demonstrated to be useful in therapy, offering emotional companionship and facilitating socialization. However, there is lack of empirical studies on what the artificial agents should do and how they can communicate with human beings better. To address these functional research problems, we attempt to establish a model to guide artificial companion designers to meet the emotional needs of the elderly through fulfilling absent roles in their social interactions. We call this model the Role Fulfilling Model. This model aims to use role as a key concept to analyse the demands from the elderly for functionalities from an emotional perspective in artificial companion agent designs and technologies. To evaluate the effectiveness of this model, we proposed a serious game platform named Happily Aging in Place. This game will help us to involve a large scale of senior users through crowdsourcing to test our model and hypothesis. To improve the emotional communication between artificial companion agents and users, This book draft addresses an important but largely overlooked aspect of affective computing: how to enable companion agents to express mixed emotions with facial expressions? And furthermore, for different users, do individual heterogeneity affects the perception of the same facial expressions? Some preliminary results about gender differences have been found. The perception of facial expressions between different age groups or cultural backgrounds will be held in future study.
Nov 30 2015 cs.AR
Cache coherence scalability is a big challenge in shared memory systems. Traditional protocols do not scale due to the storage and traffic overhead of cache invalidation. Tardis, a recently proposed coherence protocol, removes cache invalidation using logical timestamps and achieves excellent scalability. The original Tardis protocol, however, only supports the Sequential Consistency (SC) memory model, limiting its applicability. Tardis also incurs extra network traffic on some benchmarks due to renew messages, and has suboptimal performance when the program uses spinning to communicate between threads. In this paper, we address these downsides of Tardis protocol and make it significantly more practical. Specifically, we discuss the architectural, memory system and protocol changes required in order to implement the TSO consistency model on Tardis, and prove that the modified protocol satisfies TSO. We also describe modifications for Partial Store Order (PSO) and Release Consistency (RC). Finally, we propose optimizations for better leasing policies and to handle program spinning. On a set of benchmarks, optimized Tardis improves on a full-map directory protocol in the metrics of performance, storage and network traffic, while being simpler to implement.
Nov 24 2015 cs.NI
Interference coupling in heterogeneous networks introduces the inherent non-convexity to the network resource optimization problem, hindering the development of effective solutions. A new framework based on multi-pattern formulation has been proposed in this paper to study the energy efficient strategy for joint cell activation, user association and multicell multiuser channel allocation. One key feature of this interference pattern formulation is that the patterns remain fixed and independent of the optimization process. This creates a favorable opportunity for a linear programming formulation while still taking interference coupling into account. A tailored algorithm is developed to solve the formulated network energy saving problem in the dual domain by exploiting the problem structure, which gives a significant complexity saving compared to using standard solvers. Numerical results show a huge improvement in energy saving achieved by the proposed scheme.
Polynomial systems occur in many fields of science and engineering. Polynomial homotopy continuation methods apply symbolic-numeric algorithms to solve polynomial systems. We describe the design and implementation of our web interface and reflect on the application of polynomial homotopy continuation methods to solve polynomial systems in the cloud. Via the graph isomorphism problem we organize and classify the polynomial systems we solved. The classification with the canonical form of a graph identifies newly submitted systems with systems that have already been solved.
May 26 2015 cs.DC
We prove the correctness of a recently-proposed cache coherence protocol, Tardis, which is simple, yet scalable to high processor counts, because it only requires O(logN) storage per cacheline for an N-processor system. We prove that Tardis follows the sequential consistency model and is both deadlock- and livelock-free. Our proof is based on simple and intuitive invariants of the system and thus applies to any system scale and many variants of Tardis.
Polynomial systems occur in many areas of science and engineering. Unlike general nonlinear systems, the algebraic structure enables to compute all solutions of a polynomial system. We describe our massive parallel predictor-corrector algorithms to track many solution paths of a polynomial homotopy. The data parallelism that provides the speedups stems from the evaluation and differentiation of the monomials in the same polynomial system at different data points, which are the points on the solution paths. Polynomial homotopies that have tens of thousands of solution paths can keep a sufficiently large amount of threads occupied. Our accelerated code combines the reverse mode of algorithmic differentiation with double double and quad double precision to compute more accurate results faster.
Cities comprise various functional zones, including residential, educational, commercial zones, etc. It is important for urban planners to identify different functional zones and understand their spatial structure within the city in order to make better urban plans. In this research, we used 77976010 bus smart card records of Beijing City in one week in April 2008 and converted them into two-dimensional time series data of each bus platform, Then, through data mining in the big database system and previous studies on citizens' trip behavior, we established the DZoF (discovering zones of different functions) model based on SCD (smart card Data) and POIs (points of interest), and pooled the results at the TAZ (traffic analysis zone) level. The results suggested that DzoF model and cluster analysis based on dimensionality reduction and EM (expectation-maximization) algorithm can identify functional zones that well match the actual land uses in Beijing. The methodology in the present research can help urban planners and the public understand the complex urban spatial structure and contribute to the academia of urban geography and urban planning.
Enterprises have put more and more emphasis on data analysis so as to obtain effective management advices. Managers and researchers are trying to dig out the major factors that lead to employees' promotion and resignation. Most previous analyses were based on questionnaire survey, which usually consists of a small fraction of samples and contains biases caused by psychological defense. In this paper, we successfully collect a data set consisting of all the employees' work-related interactions (action network, AN for short) and online social connections (social network, SN for short) of a company, which inspires us to reveal the correlations between structural features and employees' career development, namely promotion and resignation. Through statistical analysis and prediction, we show that the structural features of both AN and SN are correlated and predictive to employees' promotion and resignation, and the AN has higher correlation and predictability. More specifically, the in-degree in AN is the most relevant indicator for promotion; while the k-shell index in AN and in-degree in SN are both very predictive to resignation. Our results provide a novel and actionable understanding of enterprise management and suggest that to enhance the interplays among employees, no matter work-related or social interplays, can largely improve the loyalty of employees.
Numerical continuation methods track a solution path defined by a homotopy. The systems we consider are defined by polynomials in several variables with complex coefficients. For larger dimensions and degrees, the numerical conditioning worsens and hardware double precision becomes often insufficient to reach the end of the solution path. With double double and quad double arithmetic, we can solve larger problems that we could not solve with hardware double arithmetic, but at a higher computational cost. This cost overhead can be compensated by acceleration on a Graphics Processing Unit (GPU). We describe our implementation and report on computational results on benchmark polynomial systems.
Jan 20 2015 cs.DC
A new memory coherence protocol, Tardis, is proposed. Tardis uses timestamp counters representing logical time as well as physical time to order memory operations and enforce sequential consistency in any type of shared memory system. Tardis is unique in that as compared to the widely-adopted directory coherence protocol, and its variants, it completely avoids multicasting and only requires O(log N) storage per cache block for an N-core system rather than O(N) sharer information. Tardis is simpler and easier to reason about, yet achieves similar performance to directory protocols on a wide range of benchmarks run on 16, 64 and 256 cores.
Jan 09 2015 cs.CR
We live in a world where our personal data are both valuable and vulnerable to misappropriation through exploitation of security vulnerabilities in online services. For instance, Dropbox, a popular cloud storage tool, has certain security flaws that can be exploited to compromise a user's data, one of which being that a user's access pattern is unprotected. We have thus created an implementation of Path Oblivious RAM (Path ORAM) for Dropbox users to obfuscate path access information to patch this vulnerability. This implementation differs significantly from the standard usage of Path ORAM, in that we introduce several innovations, including a dynamically growing and shrinking tree architecture, multi-block fetching, block packing and the possibility for multi-client use. Our optimizations together produce about a 77% throughput increase and a 60% reduction in necessary tree size; these numbers vary with file size distribution.
The logistic normal distribution has recently been adapted via the transformation of multivariate Gaus- sian variables to model the topical distribution of documents in the presence of correlations among topics. In this paper, we propose a probit normal alternative approach to modelling correlated topical structures. Our use of the probit model in the context of topic discovery is novel, as many authors have so far con- centrated solely of the logistic model partly due to the formidable inefficiency of the multinomial probit model even in the case of very small topical spaces. We herein circumvent the inefficiency of multinomial probit estimation by using an adaptation of the diagonal orthant multinomial probit in the topic models context, resulting in the ability of our topic modelling scheme to handle corpuses with a large number of latent topics. An additional and very important benefit of our method lies in the fact that unlike with the logistic normal model whose non-conjugacy leads to the need for sophisticated sampling schemes, our ap- proach exploits the natural conjugacy inherent in the auxiliary formulation of the probit model to achieve greater simplicity. The application of our proposed scheme to a well known Associated Press corpus not only helps discover a large number of meaningful topics but also reveals the capturing of compellingly intuitive correlations among certain topics. Besides, our proposed approach lends itself to even further scalability thanks to various existing high performance algorithms and architectures capable of handling millions of documents.
May 29 2014 cs.NE
For learning problem of Radial Basis Function Process Neural Network (RBF-PNN), an optimization training method based on GA combined with SA is proposed in this paper. Through building generalized Fréchet distance to measure similarity between time-varying function samples, the learning problem of radial basis centre functions and connection weights is converted into the training on corresponding discrete sequence coefficients. Network training objective function is constructed according to the least square error criterion, and global optimization solving of network parameters is implemented in feasible solution space by use of global optimization feature of GA and probabilistic jumping property of SA . The experiment results illustrate that the training algorithm improves the network training efficiency and stability.
In order to compensate for the higher cost of double double and quad double arithmetic when solving large polynomial systems, we investigate the application of NVIDIA Tesla K20C general purpose graphics processing unit. The focus on this paper is on Newton's method, which requires the evaluation of the polynomials, their derivatives, and the solution of a linear system to compute the update to the current approximation for the solution. The reverse mode of algorithmic differentiation for a product of variables is rewritten in a binary tree fashion so all threads in a block can collaborate in the computation. For double arithmetic, the evaluation and differentiation problem is memory bound, whereas for complex quad double arithmetic the problem is compute bound. With acceleration we can double the dimension and get results that are twice as accurate in about the same time.
Feb 11 2014 cs.CV
We propose a foreground segmentation algorithm that does foreground extraction under different scales and refines the result by matting. First, the input image is filtered and resampled to 5 different resolutions. Then each of them is segmented by adaptive figure-ground classification and the best segmentation is automatically selected by an evaluation score that maximizes the difference between foreground and background. This segmentation is upsampled to the original size, and a corresponding trimap is built. Closed-form matting is employed to label the boundary region, and the result is refined by a final figure-ground classification. Experiments show the success of our method in treating challenging images with cluttered background and adapting to loose initial bounding-box.
Geometric constraint systems underly popular Computer Aided Design soft- ware. Automated approaches for detecting dependencies in a design are critical for developing robust solvers and providing informative user feedback, and we provide algorithms for two types of dependencies. First, we give a pebble game algorithm for detecting generic dependencies. Then, we focus on identifying the "special positions" of a design in which generically independent constraints become dependent. We present combinatorial algorithms for identifying subgraphs associated to factors of a particular polynomial, whose vanishing indicates a special position and resulting dependency. Further factoring in the Grassmann- Cayley algebra may allow a geometric interpretation giving conditions (e.g., "these two lines being parallel cause a dependency") determining the special position.
Feb 24 2012 cs.CR
We present Path ORAM, an extremely simple Oblivious RAM protocol with a small amount of client storage. Partly due to its simplicity, Path ORAM is the most practical ORAM scheme known to date with small client storage. We formally prove that Path ORAM has a O(log N) bandwidth cost for blocks of size B = Omega(log^2 N) bits. For such block sizes, Path ORAM is asymptotically better than the best known ORAM schemes with small client storage. Due to its practicality, Path ORAM has been adopted in the design of secure processors since its proposal.
Jan 21 2010 cs.CC
Random Number Generators play a critical role in a number of important applications. In practice, statistical testing is employed to gather evidence that a generator indeed produces numbers that appear to be random. In this paper, we reports on the studies that were conducted on the compressed data using 8 compression algorithms or compressors. The test results suggest that the output of compression algorithms or compressors has bad randomness, the compression algorithms or compressors are not suitable as random number generator. We also found that, for the same compression algorithm, there exists positive correlation relationship between compression ratio and randomness, increasing the compression ratio increases randomness of compressed data. As time permits, additional randomness testing efforts will be conducted.
Nov 07 2001 cs.NI
The machines and beamlines controlled by VME industrial networks are very popular in accelerator faculties. Recently new software technology, among of which are Internet/Intranet application, Java language, and distributed calculating environment, changes the control manner rapidly. A program based on DCOM is composed to control of a variable included angle spherical grating monochromator beamline at National Synchrotron Radiation Laboratory (NSRL) in China. The control computer with a residential DCOM program is connected to Intranet by LAN, over which the user-end-operating program located in another computer sends driving beamline units' commands to the control computer. And also a web page coded in Java, published by the WWW service running in the control computer, is simply illustrated how to use web browser to query the states of or to control the beamline units.