Search

The human parsing task is dedicated to allocating pixel-level fine-grained semantic labels. However, recent solutions are limited to fully utilize the prior information (poses and edges) and the potential information of the image data, remaining problems with ambiguous boundaries, incomplete human parts, and redundant labels. To solve the above problems, we propose a novel Multi-view Stack Network (MVSN), which is constructed by three stacks with multiple views of features included parts, edges, and pre-segmentation. Meanwhile, a channel correlator is developed to acquire the correlation between local and global information better. Comprehensive experiments and corresponding results on three public datasets show that the proposed MVSN performs favorably against the state-of-the-art methods. more...

New measures of test information, termed global information, quantify test information relative to the entire range of the trait being assessed. Estimating global information relative to a non-informative prior distribution results in a measure of how much information could be gained by administering the test to an unspecified examinee. Currently, such measures have been developed only for unidimensional tests. This study introduces measures of multidimensional global test information and validates them in simulated data. Then, the utility of global test information is tested in neuropsychological data collected as part of Rush University’s Memory and Aging Project. These measures allow for direct comparison of complex tests calibrated in different samples, facilitating test development and selection. more...

Existing methods that use document-level information for event detection ignore the dependencies between sentences and also have shortcomings in modeling the dependencies among words. In this paper, we propose a novel Hierarchical Graph Enhanced Event Detection (HGEED) framework to make full use of syntax and document information for the task of event detection. First, a sentence graph is used to model word-to-word dependencies, enriching the local information of words by incorporating syntactic features. Then, a document graph is built to model sentence-to-sentence dependencies, obtaining global semantic representations for word-level prediction. The experiment results on the widely used ACE 2005 and TAC KBP 2015 corpora show that our model can capture local and global information with dependencies and achieve significant improvements as compared to all baselines. more...

The article examines the security component of international strategic communications, in which Ukrainian state information and bibliographic centers participate. The main approaches to defining the concept of strategic communications elaborated by Ukrainian researchers, as well as their colleagues from the European Union, the United States and the Russian Federation, are analyzed. The own definition of the term strategic communications is offered. The Doctrine of Information Security of Ukraine is considered, in the part concerning strategic communications and strategic narratives. It has been determined that today the greatest danger is posed by destructive information influences directed from the outside to the information space of Ukraine, especially by international Islamist movements and the Russian Federation. In addition, it has been shown that the territory of Ukraine has been used since the 1970s to transport radical Islamists from the Middle East to Western Europe. It also inflicts additional reputational damage on Ukraine at the international level. Proponents of these movements may consider information and bibliographic centers to be one of the most attractive targets for the spread of radical Islamist literature. In this case, not only the possibility of replenishing the book collections with literature of the relevant orientation is taken into account, but also the active exhibition work carried out by the centers. It is noted that printed and electronic media containing strategic narratives beneficial to the Russian Federation can enter information and bibliographic centers in a covert form (as scientific publications from the Republic of Poland, the Czech Republic, Austria, Hungary, the Baltic States and other European Union countries), the ultimate beneficiary of which are organizations controlled by the Russian Federation, created to exercise destructive information influence. Measures to counteract negative information influences using the available capabilities of state information and bibliographic centers are proposed. more...

The legal implications of information activities in the context of armed conflict against the background of the digital transformation have so far received only scarce attention. This article aims to fill this gap by exposing some of the legal issues arising in relation to mis- and disinformation tactics during armed conflict in order to provide a starting point for further debate in this respect. Specifically, it explores the existence and content of existing limits imposed by international humanitarian law on (digital) information operations and inquires whether the current framework adequately captures the humanitarian protection needs that arise from such conduct. more...

Human-Object Interaction (HOI) Detection is a new genre of human-centric visual relationship detection task, which is significant to deep understanding of visual scenes. Due to the complexity of the visual scene in the image, HOI detection is still a challenging task, the most critical part of which is feature extraction and representation. Some existing approaches rely solely on local region information for HOI detection without using global contextual information, but global contextual information contributes to this task in some HOI categories. Other approaches incorporate global contextual information for HOI detection while losing local region information. In this work, we propose a multi-stream neural network architecture composed of three special module that employs both local region information and global contextual information for HOI detection. This model can detect not only the HOI categories based on local region information but also on global contextual information. Our model more fully considers all HOI categories in the dataset. Compared with other existing approaches, the proposed model shows improved performance on V-COCO and HICO-DET benchmark datasets, especially when predicting rare HOI categories. more...

It is still a hot research topic to identify node importance in complex networks. Recently many methods have been proposed to deal with this problem. However, most of the methods only focus on local or path information, they do not combine local and global information well. In this paper, a new model to identify node importance based on Decision-making Trial and Evaluation Laboratory (DEMATEL) is presented. DEMATEL method is based on graph theory which takes the global information into full consideration so that it can effectively identify the importance of one element in the whole complex system. Some experiments based on susceptible-infected (SI) model are used to compare the new model with other methods. The applications in three different networks illustrate the effectiveness of the new model. more...

Cybersecurity protects citizens and society from harm perpetrated through computer networks. Its task is made ever more complex by the diversity of actors—criminals, spies, militaries, hacktivists, firms—operating in global information networks, so that cybersecurity is intimately entangled with the so-called grey zone of conflict between war and peace in the early twenty-first century. To counter cyber threats from this environment, cybersecurity is turning to artificial intelligence and machine learning (AI) to mitigate anomalous behaviours in cyberspace. This article argues that AI algorithms create new modes and sites of cybersecurity knowledge production through hybrid assemblages of humans and nonhumans. It concludes by looking beyond ‘everyday’ cybersecurity to military and intelligence use of AI and asks what is at stake in the automation of cybersecurity. more...

The development of global information and transportation technology is accompanied by the development of criminal acts that no longer recognize jurisdictional boundaries, so that the response requires joint handling of world countries. Mutual assistance in criminal matters is one way to stop fraudulent acts of criminals who want to hide assets. The implementation of the Mutual Legal Assistance System was realized in the implementation of Law Number 1 of 2006 concerning Reciprocal Assistance in Criminal Issues, namely by establishing cooperation with countries that have the potential to be used as hiding places or to save income from the crime. This has become one of the efforts to prevent cross-jurisdictional cyber crime by implementing a Mutual Legal Assistance System more...

In this paper, we present robust variants of distributed clustering algorithms for large datasets distributed across multiple machines in the presence of Byzantines. We propose a redundant data assignment scheme that enables us to obtain global information about the entire dataset for clustering purposes even when some machines are adversarial in nature. Simulation results show that the distributed algorithms based on the proposed assignment scheme provide good-quality solutions for a variety of clustering problems. more...

Nowadays, most deep learning models ignore Chinese habits and global information when processing Chinese tasks. To solve this problem, we constructed the BERT-BiLSTM-Attention-CRF model. In the model, we embeded the BERT pre-training language model that adopts the Whole Word Mask strategy, and added a document-level attention. Experimental results show that our method achieves good results in the MSRA corpus, and F1 reaches 95.00%. more...

Extracting relational triples from unstructured text is essential for the construction of large-scale knowledge graphs, QA and other downstream tasks. The purpose of dialogue relation extraction is to extract the relations between entities from the multi-person dialogue texts. The existing dialogue relation extraction models only focused on coarse-grained global information and ignored fine-grained local information. In this paper, we propose a dialogue relation extraction model BERT-MG to capture the features on different granularity at different BERT layers to take advantage of the fine-grained dialogue features. Moreover, we design a type-confidence mechanism to use the entity type information to assist relation inference. Experimental results on the DialogRE dataset prove that our proposed model BERT-MG outperforms the SOTA baselines. more...

Existing Graph Convolution Networks (GCNs) mainly explore the depth structure with focusing on aggregating information from the local 1-hop neighbor, leading to the common issues of over-smoothing and gradient vanishing, and further hurting the model performance on downstream tasks. To alleviate these deficiencies, we tentatively explore the width of the graph structure and propose Wider Graph Convolution Network (WGCN), targeting to reason the hierarchical relations and aggregate the global information from multi-hops dilated neighbor in a wide but shallow framework. Meanwhile, Dynamic Graph Convolution (DGC) is adopted via the masking and attention mechanisms to distinguish and re-weight informative neighborhood nodes, for stabilizing the model optimization. We demonstrate the effectiveness of WGCN on three popular tasks, including Document classification, Scene text detection and Point cloud segmentation. On the basis of achieving state-of-the-art performance among presented tasks, we verify that exploration through width expansion of GCNs is more effective than stacking more layers for model improvements. more...

Tento článek podrobně popisuje inovativní techniku vkládání globální (nebo obecně rozsáhlé) míry kvality do hluboké neuronové sítě (DNN) s cílem kompenzovat tendenci DNN zakládat výslednou klasifikaci prakticky ze superpozice lokálních transformací a projekcí okolí. Jako globální míru kvality jsme použili rys podobný pravděpodobnosti stavu AQ1 a injektovali ji do klasifikátoru založeného na DNN nasazeného v konkrétní úloze určení, které části webové stránky jsou určitým zájmem pro další zpracování technikami NLP. Naším cílem bylo rozložit webové stránky různých internetových diskusních fór AQ2 na užitečný obsah, tj. příspěvky uživatelů, a neužitečný obsah, tj. grafiku fóra, menu, bannery, reklamy atd. This paper describes in details an innovative technique of injection of a global (or generally largescale) quality measure into a deep neural network (DNN) in order to compensate for the tendency of DNNs to found the resulting classification virtually from a superposition of local neighbourhood transformations and projections. We used a state AQ1 probability-like feature as the global quality measure and injected it into a DNN-based classifier deployed in a specific task of determining which parts of a web page are of certain interest for further processing by NLP techniques. Our goal was to decompose web sites of various internet AQ2 discussion forums to useful content, i.e. the posts of users, and useless content, i.e. forum graphics, menus, banners, advertisements, etc. more...

Recently, in order to address the unsupervised domain adaptation (UDA) problem, extensive studies have been proposed to achieve transferrable models. Among them, the most prevalent method is adversarial domain adaptation, which can shorten the distance between the source domain and the target domain. Although adversarial learning is very effective, it still leads to the instability of the network and the drawbacks of confusing category information. In this paper, we propose a Robust Ensembling Network (REN) for UDA, which applies a robust time ensembling teacher network to learn global information for domain transfer. Specifically, REN mainly includes a teacher network and a student network, which performs standard domain adaptation training and updates weights of the teacher network. In addition, we also propose a dual-network conditional adversarial loss to improve the ability of the discriminator. Finally, for the purpose of improving the basic ability of the student network, we utilize the consistency constraint to balance the error between the student network and the teacher network. Extensive experimental results on several UDA datasets have demonstrated the effectiveness of our model by comparing with other state-of-the-art UDA algorithms. more...

Fine-tuning with pre-trained models has achieved exceptional results for many language tasks. In this study, we focused on one such self-attention network model, namely BERT, which has performed well in terms of stacking layers across diverse language-understanding benchmarks. However, in many downstream tasks, information between layers is ignored by BERT for fine-tuning. In addition, although self-attention networks are well-known for their ability to capture global dependencies, room for improvement remains in terms of emphasizing the importance of local contexts. In light of these advantages and disadvantages, this paper proposes SesameBERT, a generalized fine-tuning method that (1) enables the extraction of global information among all layers through Squeeze and Excitation and (2) enriches local information by capturing neighboring contexts via Gaussian blurring. Furthermore, we demonstrated the effectiveness of our approach in the HANS dataset, which is used to determine whether models have adopted shallow heuristics instead of learning underlying generalizations. The experiments revealed that SesameBERT outperformed BERT with respect to GLUE benchmark and the HANS evaluation set. more...

As the community detection is able to facilitate the discovery of hidden information in complex networks, it has been drawn a lot of attention recently. However, due to the growth in computational power and data storage, the scale of these complex networks has grown dramatically. In order to detect communities by utilizing global approaches, it is required to have all the global information of the whole network; something which is impossible, because of the rapid growth in the size of the networks. In this paper, a local approach has been proposed based on the detection and expansion of core nodes. First, a community’s central node (core node) which has a high level of embeddedness is detected based on the similarity between graph’s nodes. By using this, the total weights of a weighted graph’s edges created. Following by that, the expansion of these nodes will be considered, by utilizing the concept of node’s membership based on the definition of strong community for weighted graphs. It can be seen that in detecting communities, the more accurate the weights of edges detected based on the node similarity, the more precise the local algorithm will be. In fact, the algorithm has the ability to detect all the graph’s communities in a network using local information as well as identifying various roles of nodes, either being (core or outlier). Test results on both real-world and artificial networks prove that the quality of the communities which are detected by the proposed algorithm is better than the results which are achieved by other state-of-the-art algorithms in the complex networks. more...

Urgency of the research. Further improvement of the security corporate networks in the conditions of massive influence of computer attacks requires an increase in the probability of detection of new computer attacks and a decrease in the recognition time for the signs of known attacks. Target setting. Analysis of the texts of the global information space reduces the time of detection of possible threats. Actual scientific researches and issues analysis. Recent publications about systems of defense from attacks and use of text analysis in detecting threats were considered. Uninvestigated parts of general matters defining. It is necessary to improve the methods of working out the data sets of the body of network packets, content of Internet pages, information of mass media and social networks, which in turn raises the problem of semantic and syntactic processing of natural language texts. The research objective. The aim of the paper is organization collective protection of corporate networks via the introduction of threat monitoring systems, active intelligence activities in the global information space in order to search collect and analyze data about attacks, abnormal behavior, and content of Internet resources. The statement of basic materials. The requirements of security systems to reduce the time of a threat detection lead to the need for active intelligence assessment aimed at continuous monitoring of the surrounding cyberspace that consists of a variety of individual users and organizations’ computer networks. The purpose of such monitoring is to determine the characteristics, interests, features of the security policy of a particular corporate network in the global information space. In this context, particular importance attaches to the analysis of text information from both fully and partially open digital sources. A rational solution to this task is the establishment of threat monitoring centers aimed at the organization of collective protection for corporate networks related to them. Conclusions. The proposed method of protection allows both to detect cyber threats in the global information space and to customize their own corporate network security systems in accordance with their characteristic threat vectors., Актуальность темы исследования. Дальнейшее усовершенствование защищенности корпоративных сетей в условиях массированного влияния компьютерных атак требует роста вероятности выявления новых компьютерных атак и уменьшения времени распознавания признаков известных атак. Постановка проблемы. Анализ текстов глобального информационного пространства позволяет сократить время обнаружения возможных угроз. Анализ последних исследований и публикаций. Были рассмотрены последние публикации по системам защиты от атак и использованию анализа текстов для обнаружения угроз. Выделение неисследованных частей общей проблемы. Совершенствование методов обработки массивов данных тела сетевых пакетов, содержания интернет-страниц, информации СМИ и социальных сетей, что в свою очередь поднимает проблему семантико-синтаксической обработки текстов естественного языка. Постановка задачи. Организация коллективной защиты корпоративных сетей путем внедрения систем мониторинга угроз, активной разведывательной деятельности в глобальном информационном пространстве с целью поиска, сбора и анализа данных об атаках, аномальном поведении, размещаемом контенте ресурсов сети Интернет. Изложение основного материала. Требования систем защиты относительно минимизации времени выявления угроз приводят к необходимости ведения активной разведывательной деятельности, направленной на ведение постоянного мониторинга окружающего киберпространства, состоящего из множества компьютерных сетей отдельных пользователей и организаций. Цель такого мониторинга - определение характеристик, интересов, особенностей политики безопасности конкретной корпоративной сети в глобальном информационном пространстве. Особое значение при этом приобретает анализ текстовой информации из открытых и условно-открытых электронных источников. Рациональным решением такой задачи является формирование центров мониторинга угроз, направленных на организацию коллективной защиты подчиненных корпоративных сетей. Выводы в соответствии со статьей. Предложенный метод защиты позволяет как выявлять киберугрозы в глобальном информационном пространстве, так и настраивать собственные системы защиты корпоративных сетей в соответствии с их характеристическими векторами угроз., Актуальність теми дослідження. Подальше вдосконалення захищеності корпоративних мереж в умовах масованого впливу комп’ютерних атак вимагає підвищення ймовірності виявлення нових комп'ютерних атак і зниженням часу розпізнавання ознак відомих атак. Постановка проблеми. Аналіз текстів глобального інформаційного простору дозволяє скоротити час виявлення можливих загроз. Аналіз останніх досліджень і публікацій. Були розглянуті останні публікації щодо систем захисту від атак та використання аналізу текстів у разі виявлення загроз. Виділення недосліджених частин загальної проблеми. Вдосконалення методів опрацювання масивів даних тіла мережевих пакетів, вмісту інтернет-сторінок, інформації СМІ та соціальних мереж, що у свою чергу порушує проблему семантико-синтаксичної обробки текстів природної мови. Постановка завдання. Організація колективного захисту корпоративних мереж шляхом впровадження систем моніторингу загроз, активної розвідувальної діяльності в глобальному інформаційному просторі з метою пошуку, збору та аналізу даних про атаки, аномальну поведінку, вміщуваний контент ресурсів мережі Інтернет . Виклад основного матеріалу. Вимоги систем захисту щодо скорочення часу виявлення загроз призводять до необхідності ведення активної розвідувальної діяльності, спрямованої на проведення постійного моніторингу оточуючого кіберпростору, що складається з множини комп’ютерних мереж окремих користувачів та організацій. Метою такого моніторингу є визначення характеристик, інтересів, особливостей політики безпеки конкретної корпоративної мережі в глобальному інформаційному просторі. Особливе значення при цьому набуває аналіз текстової інформації з відкритих та умовно-відкритих електронних джерел. Раціональним вирішенням такого завдання є формування центрів моніторингу загроз, спрямованих на організацію колективного захисту підлеглих корпоративних мереж. Висновки відповідно до статті. Запропонований метод захисту дозволяє як виявляти кіберзагрози в глобальному інформаційному просторі, так і налаштовувати власні системи захисту корпоративних мереж згідно з їх характеристичними векторами загроз. more...

With advances in reinforcement learning (RL), agents are now being developed in high-stakes application domains such as healthcare and transportation. Explaining the behavior of these agents is challenging, as the environments in which they act have large state spaces, and their decision-making can be affected by delayed rewards, making it difficult to analyze their behavior. To address this problem, several approaches have been developed. Some approaches attempt to convey the global behavior of the agent, describing the actions it takes in different states. Other approaches devised local explanations which provide information regarding the agent's decision-making in a particular state. In this paper, we combine global and local explanation methods, and evaluate their joint and separate contributions, providing (to the best of our knowledge) the first user study of combined local and global explanations for RL agents. Specifically, we augment strategy summaries that extract important trajectories of states from simulations of the agent with saliency maps which show what information the agent attends to. Our results show that the choice of what states to include in the summary (global information) strongly affects people's understanding of agents: participants shown summaries that included important states significantly outperformed participants who were presented with agent behavior in a set of world-states that are likely to appear during gameplay. We find mixed results with respect to augmenting demonstrations with saliency maps (local information), as the addition of saliency maps, in the form of raw heat maps, did not significantly improve performance in most cases. However, we do find some evidence that saliency maps can help users better understand what information the agent relies on during its decision-making, suggesting avenues for future work that can further improve explanations of RL agents. more...

This study uses a longitudinal dataset extracted from a mobile news application and adopts a multilevel design to examine the evolution of diversity of individuals’ news consumption and to identify the factors that underlie such evolution. A decreasing trend in news consumption diversity is observed among users. The news consumption diversity of individuals is positively related to global information diversity. Furthermore, the news consumption diversity of males exhibits a stronger tendency to be influenced by global information diversity than that of females. The decreasing trend in news consumption diversity is less remarkable among males than among females. This study complements traditional motivation-driven perspectives of news consumption by mining the structural antecedents of news consumption diversity and further emphasizes the social implications of mobile news technology. Lastly, practical implications and limitations are discussed. more...

Classification is one of the most popular branches of supervised learning algorithms. In the real-world problems, there are some situations in which distributions of the two classes are not the same. This situation is known as a class imbalanced problem. In the past years, several investigations have been done to find a way of handling imbalanced data, which most of them stay in one of two groups, including internal techniques and external techniques. The proposed gravitational density-based mass sharing method (GDMS) is an internal method that is designed based on k-nearest neighbor and fixed radius nearest neighbor (FRNN) rules. GDMS is a new technique that assigns masses to instances based on their local density while considering the global information too. In the labeling phase, GDMS decides based on the sum of gravitational forces coming from the candidates set, which are defined by FRNN rule. The GDMS does not need any parameters to be set in the whole procedure of classification, which is an advantage in comparison with the previous methods. To demonstrate the effectiveness of our proposed method, we use 40 standard datasets from the KEEL repository. The results show the effectiveness and superiority of GDMS compared to the competing methods. more...

Coreference resolution aims at recognizing different forms in a document which refer to the same entity in the real world. Although many models have been proposed and achieved success, there still exist some challenges. Recent models that use recurrent neural networks to obtain mention representations ignore dependencies between spans and their proceeding distant spans, which will lead to predicted clusters that are locally consistent but globally inconsistent. In addition, these models are trained only by maximizing the marginal likelihood of gold antecedent spans from coreference clusters, which will make some gold mentions undetectable and cause unsatisfactory coreference results. To address these challenges, we propose a neural coreference resolution model. It employs mutual attention to take into account the dependencies between spans and their proceeding spans directly (use attention mechanism to capture global information between spans and their proceeding spans). And our model is trained by jointly optimizing mention clustering and imbalanced mention detection, which enables it to detect more gold mentions in a document to make more accurate coreference decisions. Experimental results on the CoNLL-2012 English dataset show that our model can detect the most gold mentions and achieve the state-of-the-art coreference performance compared with baselines. more...

It is extremely challenging to acquire perceptually plausible images under low-light conditions due to low SNR. Most recently, U-Nets have shown promising results for low-light imaging. However, vanilla U-Nets generate images with artifacts such as color inconsistency due to the lack of global color information. In this paper, we propose a global information aware (GIA) module, which is capable of extracting and integrating the global information into the network to improve the performance of low-light imaging. The GIA module can be inserted into a vanilla U-Net with negligible extra learnable parameters or computational cost. Moreover, a GIA-Net is constructed, trained and evaluated on a large scale real-world low-light imaging dataset. Experimental results show that the proposed GIA-Net outperforms the state-of-the-art methods in terms of four metrics, including deep metrics that measure perceptual similarities. Extensive ablation studies have been conducted to verify the effectiveness of the proposed GIA-Net for low-light imaging by utilizing global information. more...

Autoregressive neural machine translation (NMT) models are often used to teach non-autoregressive models via knowledge distillation. However, there are few studies on improving the quality of autoregressive translation (AT) using non-autoregressive translation (NAT). In this work, we propose a novel Encoder-NAD-AD framework for NMT, aiming at boosting AT with global information produced by NAT model. Specifically, under the semantic guidance of source-side context captured by the encoder, the non-autoregressive decoder (NAD) first learns to generate target-side hidden state sequence in parallel. Then the autoregressive decoder (AD) performs translation from left to right, conditioned on source-side and target-side hidden states. Since AD has global information generated by low-latency NAD, it is more likely to produce a better translation with less time delay. Experiments on WMT14 En-De, WMT16 En-Ro, and IWSLT14 De-En translation tasks demonstrate that our framework achieves significant improvements with only 8% speed degeneration over the autoregressive NMT. more...

Multivariate Time Series Classification (MTSC) is believed to be a crucial task towards dynamic process recognition and has been widely studied. Recent years, end-to-end MTSC with Convolutional Neural Network (CNN) has gained increasing attention thanks to its ability to integrates local features. However, it remains a significant challenge for CNN to handle global information and long-range dependencies of time series. In this paper, we present a simple and feasible architecture for MTSC to address these problems. Our model benefits from self-attention, which can help CNN directly capture the relationships of time series between two random time steps or variables. Experimental results of the proposed model work on thirty five complex MTSC tasks show its effectiveness and universality that has to outperform existing state-of-the-art (SOTA) model overall. Besides, our model is computationally efficient, and the parsing speed is six hours faster than the current model. more...

nowadays information technologies influence on all spheres of human activity that ensure the spread of information flows in society, and form a global information space. So, an integral and important part of the informatization of society is the informatization of education thanks to computers, PDAs tablets or mobile phones. Currently, there are about one and a half billion mobile phones in the world, which is three times more than traditional computers. Mobile phones, smartphones, tablets are becoming a major part of human's digital life. Some of these apps are the mobile version of those websites that provide English learning materials for the students as well as other language learners. This article describes the role of educational mobile applications in learning a foreign language. The author focuses on the didactic characteristics of educational mobile applications and their capabilities. more...

Telematization of almost everything; the world-wide ubiquitous “electronical” presence (“hic et nunc et ubique”) will make the idea of a global information village come true - not only in passive..

In this paper, we propose a global-to-local constraint network (GLCN) to improve the robustness of face alignment. Especially in the challenging cases when faces undergo large variations including severe poses and occlusions. Unlike conventional cascade regression which based on average faces to conduct regressing, which does not take full advantage of face information. Our GLCN learns a custom shape for every input images through global information. Then local adjustments will be applied for these initializations for robust feature representation. To achieve this goal, we design a local constraint network, which models the global-to-local constraint mechanism. We employ an hourglass network to learn global feature representation, and typically incorporates with five subnets to optimize facial landmarks. To further reinforce the alignment performance, we integrate our designed model in the paradigm of RNN network. Experimental results demonstrate the robustness of our GLCN versus many challenging cases especially. more...

In this paper we evaluate our new approach based on the Continuous Bag-of-Words and Skip-gram models enriched with global context information on highly inflected Czech language and compare it with English results. As a source of information we use Wikipedia, where articles are organized in a hierarchy of categories. These categories provide useful topical information about each article. Both models are evaluated on standard word similarity and word analogy datasets. Proposed models outperform other word representation methods when similar size of training data is used. Model provide similar performance especially with methods trained on much larger datasets. more...

Human parsing has recently attracted considerable interests due to its wide application potentials. However, developing an accurate human parsing system is still a challenge for researchers. In this paper, we demonstrate that global information are critical for accurate prediction by applying a non-local operation for effectively extracting global information. Meanwhile several training data refinement methodologies are proposed to further boost the performance. Benefiting from all the approaches, the proposed single human parsing model NLGINet achieves the state-of-the-art segmentation accuracy on two human parsing benchmark datasets LIP and Pascal-Person-Parts. more...

Single‐cell Hi‐C technology is emerging and will provide unprecedented opportunities to elucidate chromosomal dynamics with high resolution. How to characterize pseudo time‐series of single cells using single‐cell Hi‐C maps is an essential and challenging topic. To this end, a powerful circular trajectory reconstruction tool CIRCLET is developed to resolve cell cycle phases of single cells by considering multiscale features of chromosomal architectures without specifying a starting cell. CIRCLET reveals its best superiority based on the combination of one feature set about global information and another two feature sets about local interactional information in terms of designed evaluation indexes and verification strategies from a collection of cell‐cycle Hi‐C maps of 1171 single cells. Further division of the reconstructed trajectory into 12 stages helps to accurately characterize the dynamics of chromosomal structures and explain the special regulatory events along cell‐cycle progression. Last but not the least, the reconstructed trajectory helps to uncover important regulatory genes related with dynamic substructures, providing a novel framework for discovering regulatory regions even cancer markers at single‐cell resolution., CIRCLET first extracts multiscale features of chromosomal architectures for each cell, and then reconstructs a circular trajectory of single cells. Further division of the reconstructed trajectory helps to characterize the dynamics of chromosomal structures and uncover important regulatory genes along cell‐cycle progression. CIRCLET is the first tool for accurate reconstruction of a pseudotime trajectory based on single‐cell 3D genome maps. more...

Utilization of multitype measurements including local and global information for structural health monitoring (SHM) has typically outperformed that using solo-type measurements. However, in...

To find the important nodes in complex networks is a fundamental issue. A number of methods have been recently proposed to address this problem but most previous studies have the limitations, and few of them considering both local and global information of the network. The location of node, which is a significant property of a node in the network, is seldom considered in identifying the importance of nodes before. To address this issue, we propose an improved gravity centrality measure on the basis of the k-shell algorithm named KSGC to identify influential nodes in the complex networks. Our method takes the location of nodes into consideration, which is more reasonable compared to original gravity centrality measure. Several experiments on real-world networks are conducted to show that our method can effectively evaluate the importance of nodes in complex networks. more...

General Game Playing (GGP) aims at creating computer programs able to play any arbitrary game at an expert level given only its rules. The lack of game-specific knowledge and the necessity of learning a strategy online have made Monte-Carlo Tree Search (MCTS) a suitable method to tackle the challenges of GGP. An efficient search-control mechanism can substantially increase the performance of MCTS. The RAVE strategy and its more recent variant, GRAVE, have been proposed for this reason. In this paper we further investigate the use of GRAVE for GGP and compare its performance with the more established RAVE strategy and with a new variant, called HRAVE, that uses more global information. Experiments show that for some games GRAVE and HRAVE perform better than RAVE, with GRAVE being the most promising one overall. more...

The World Society for Pediatric and Congenital Heart Surgery was founded with the mission to “promote the highest quality comprehensive cardiac care to all patients with congenital heart disease, from the fetus to the adult, regardless of the patient’s economic means, with an emphasis on excellence in teaching, research, and community service.” Early on, the Society’s members realized that a crucial step in meeting this goal was to establish a global database that would collect vital information, allowing cardiac surgical centers worldwide to benchmark their outcomes and improve the quality of congenital heart disease care. With tireless efforts from all corners of the globe and utilizing the vast experience and invaluable input of multiple international experts, such a platform of global information exchange was created: The World Database for Pediatric and Congenital Heart Disease went live on January 1, 2017. This database has been thoughtfully designed to produce meaningful performance and quality analyses of surgical outcomes extending beyond immediate hospital survival, allowing capture of important morbidities and mortalities for up to 1 year postoperatively. In order to advance the societal mission, this quality improvement program is available free of charge to WSPCHS members. In establishing the World Database, the Society has taken an essential step to further the process of global improvement in care for children with congenital heart disease. more...

In this article, we investigate the problem of cross-document person name disambiguation, which aimed at resolving ambiguities between person names and clustering web documents according to their association to different persons sharing the same name. The majority of previous work often formulated cross-document name disambiguation as a clustering problem. These methods employed various syntactic and semantic features either from the local corpus or distant knowledge bases to compute similarities between entities and group similar entities. However, these approaches show limitations regarding robustness and performance. We propose an unsupervised, graph-based name disambiguation approach to improve the performance and robustness of the state-of-the-art. Our approach exploits both local information extracted from the given corpus, and global information obtained from distant knowledge bases. We show the effectiveness of our approach by testing it on standard WePS datasets. The experimental results are encouraging and show that our proposed method outperforms several baseline methods and also its counterparts. The experiments show that our approach not only improves the performances, but also increases the robustness of name disambiguation. more...

Many neural models have been built to carry out reading comprehension tasks. However, these models mainly focus on formal passages like news and book stories. Although human dialog is the most important part of daily life, machine reading comprehension on dialogs (i.e., passage completion on dialogs) has not been sufficiently explored. Existing models show some weaknesses when comprehending dialogs and they are unable to capture global information over a distance and local detailed information at the same time. This paper introduces a neural network model DeepDial that aims at addressing the problems mentioned above. The model explores both word-level and utterance-level information in a dialog, and achieves the state-of-the-art performance on the benchmark dataset constructed from a TV series Friends. more...

The charge prediction task for multi-defendant cases is to determine appropriate charges for a specific defendant according to its name and its fact description. This task is not trivial since it is hard to recognize fact descriptions for different defendants. Therefore, we propose a multi-scale attention model for this problem. We employ local attention, which is highly related to the position of the specific defendant’s name appear in the fact description, to restrict our model to the description for a specific defendant and employ global attention, which is calculated by a charge prediction model for single-defendant cases, to supplement the model with global information of the case. We collect about 160,000 indictments for experiments. After data preprocessing, we choose the two most common charge pairs which are Theft with Concealment of Crime-related Income, and Open Casinos with Gamble for experiments. Experimental results show the effectiveness of our model, the multi-scale attention model does benefit from the global information from the complete case compared to the local attention model. more...

Graph Convolutional Networks (GCNs) are a class of spectral clustering techniques that leverage localized convolution filters to perform supervised classification directly on graphical structures. While such methods model nodes’ local pairwise importance, they lack the capability to model global importance relative to other nodes of the graph. This causes such models to miss critical information in tasks where global ranking is a key component for the task, such as in keyphrase extraction. We address this shortcoming by allowing the proper incorporation of global information into the GCN family of models through the use of scaled node weights. In the context of keyphrase extraction, incorporating global random walk scores obtained from TextRank boosts performance significantly. With our proposed method, we achieve state-of-the-art results, bettering a strong baseline by an absolute 2% increase in F1 score. more...

The performance of traditional part model in person retrieval is greatly affected by the quality of parts. The recent work[1] consider refining the hard partitioned part when training the network itself and got state-of-the-art performance, but during our experimentation, we found the masks which it generated contains the problems that misguide the networks with additional constraints, Targeting to solve above problem, we proposed a new networks called Global Refine Net. The backbone network focus on learning the local information which improve the ability of extract feature of details, Global Refine block introduce global information to adjust the hard-shaped part generated by the backbone network in an end-to-end manner. Also we modified the self-adversarial training mechanism in [1]. We employ an special loss function to prevent the incorrect convergence and adjust the degree of self-adversarial training, the new regularization term we added in the loss benefit both in stabilizing and speeding the training process. The performance of our model beat most previous soft partitioned works, improved about 2.3% rank-1 accuracy and 5.1% mAP to the PCB baseline on market-1501 dataset. more...

Relief based algorithms have often been claimed to uncover feature interactions. However, it is still unclear whether and how interaction terms will be differentiated from marginal effects. In this paper, we propose IMMIGRATE algorithm by including and training weights for interaction terms. Besides applying the large margin principle, we focus on the robustness of the contributors of margin and consider local and global information simultaneously. Moreover, IMMIGRATE has been shown to enjoy attractive properties, such as robustness and combination with Boosting. We evaluate our proposed method on several tasks, which achieves state-of-the-art results significantly., R package ('Immigrate') available on CRAN more...

Drug repositioning is an efficient and promising strategy to identify new indications for existing drugs, which can improve the productivity of traditional drug discovery and development. Rapid advances in high-throughput technologies have generated various types of biomedical data over the past decades, which lay the foundations for furthering the development of computational drug repositioning approaches. Although many researches have tried to improve the repositioning accuracy by integrating information from multiple sources and different levels, it is still appealing to further investigate how to efficiently exploit valuable data for drug repositioning. In this study, we propose an efficient approach, Random Walk on a Heterogeneous Network for Drug Repositioning (RWHNDR), to prioritize candidate drugs for diseases. First, an integrated heterogeneous network is constructed by combining multiple sources including drugs, drug targets, diseases and disease genes data. Then, a random walk model is developed to capture the global information of the heterogeneous network. RWHNDR takes advantage of drug targets and disease genes data more comprehensively for drug repositioning. The experiment results show that our approach can achieve better performance, compared with other state-of-the-art approaches which prioritized candidate drugs based on multi-source data. more...

Understanding the community structure of realworld networks is an important task to predict the dynamics of many complex systems. To this end, several optimization methods were developed to maximize the widely studied measure known as Modularity. Most of these methods use global information and, therefore, are computationally expensive to process large-scale networks. This paper proposes a genetic algorithm to detect communities in directed networks, named NGA-LP, that contains local genetic operators designed to have low computational cost. The primary advantage of NGA-LP is the local representation, where the vertices store the information of the individuals. This representation makes possible the use of local genetic operators which do not require global information. Moreover, NGA-LP combines a pair of crossover operators that are automatically chosen according to the characteristics of the network, guided by the quality of the solution. The goal of combining different crossover operators is to ensure the robustness and capability of handling with different networks in an adaptive fashion. In the computational tests carried out in this paper, the introduced algorithm achieved excellent results and outperformed the other benchmark algorithms, even for undirected networks. more...

Symbolic value partitioning is a knowledge reduction technique in the field of data min- ing. In this paper, we propose a granular computing approach for the partitioning task that includes granule construction and granule selection algorithms. The granule construction algorithm takes advantage of local information associated with each attribute. A binary attribute value taxonomy tree is built to merge these attribute values in a bottom-up manner using information-loss heuristics. The use of a balancing technique enables us to control different nodes in the same level to have approximately the same size. The granule selection algorithm uses global information about all of the attributes in the decision system. Hence, nodes across the taxonomy forest of all attributes are selected and expanded using information-gain heuristics. We present a series of experimental results that demonstrate the effectiveness of the proposed approach in terms of reducing the data size and improving the resulting classification accuracy. more...

Dropout is used to avoid overfitting by randomly dropping units from the neural networks during training. Inspired by dropout, this paper presents GI-Dropout, a novel dropout method integrating with global information to improve neural networks for text classification. Unlike the traditional dropout method in which the units are dropped randomly according to the same probability, we aim to use explicit instructions based on global information of the dataset to guide the training process. With GI-Dropout, the model is supposed to pay more attention to inapparent features or patterns. Experiments demonstrate the effectiveness of the dropout with global information on seven text classification tasks, including sentiment analysis and topic classification. more...

International audience; Relation extraction (RE) between a pair of entity mentions from text is an important and challenging task specially for open domain relations. Generally, relations are extracted based on the lexical and syntactical information at the sentence level. However, global information about known entities has not been explored yet for RE task. In this paper, we propose to extract a graph of entities from the overall corpus and to compute features on this graph that are able to capture some evidences of holding relationships between a pair of entities. The proposed features boost the RE performance significantly when these are combined with some linguistic features. more...

Clinical named entity recognition (NER) is a foundational technology to acquire the knowledge within the electronic medical records. Conventional clinical NER methods suffer from heavily feature engineering. Besides, these methods treat NER as a sentence-level task and ignore the long-range contextual dependencies. In this paper, we propose an attention-based neural network architecture to leverage document-level global information to alleviate the problem. The global information is obtained from document represented by pre-trained bidirectional language model (Bi-LM) with neural attention. The parameters of pre-trained Bi-LM which makes use of unlabeled data can be transferred to NER model to further improve the performance. We evaluate our model on 2010 i2b2/VA datasets to verify the effectiveness of leveraging global information and transfer strategy. Our model outperforms previous state-of-the-art method with less labeled data and no feature engineering. more...