Abstract: Taxonomy matching aims to discover categories alignments between two taxonomies, which is an important operation of knowledge sharing task to benefit many applications. The existing methods for taxonomy matching mostly depend on string lexical features and domain-specific information. In this paper, we consider the method of representation learning of taxonomies, which projects categories and relationships into low-dimensional vector spaces. We propose a method to takes advantages of category hierarchies and siblings, which exploits a low-dimensional semantic space to modeling categories relations by translating operations in the semantic space. We take advantage of maximum weight matching problem on bipartite graphs to model taxonomy matching problem, which runs in polynomial time to generate optimal categories alignments for two taxonomies in a global manner. Experimental results on OAEI benchmark datasets show that our method significantly outperforms the baseline methods in taxonomy matching.

Abstract: Scientific Named Entity Referent Extraction is often more complicated than traditional Named Entity Recognition (NER). For ex- ample, in polymer science, chemical structure may be encoded in a variety of nonstandard naming conventions, and authors may refer to polymers with conventional names, commonly used names, labels (in lieu of longer names), synonyms, and acronyms. As a result, accurate scientific NER methods are often based on task-specific rules, which are difficult to develop and maintain, and are not easily generalized to other tasks and fields. Machine learning models require substantial expert-annotated data for training. Here we propose polyNER: a semi-automated system for efficient identification of scientific entities in text. PolyNER applies word embedding models to generate entity-rich corpora for productive expert labeling, and then uses the resulting labeled data to bootstrap a context-based word vector classifier. Evaluation on materials science publications shows that polyNER’s combination of automated analysis with minimal expert input enables noticeably improved precision or re- call relative to a state-of-the-art chemical entity extraction system. This remarkable result highlights the potential for human-computer partner- ship for constructing domain-specific scientific NER systems.

Abstract: The convergence of Big Data applications with High-Performance Computing requires new methodologies to store, manage and process large amounts of information. Traditional storage solutions are unable to scale and that results in complex coding strategies. For example, the brain atlas of the Human Brain Project has the challenge to process large amounts of high-resolution brain images. Given the computing needs, we study the effects of replacing a traditional storage system with a distributed key-value database on a cell segmentation application. The original code uses HDF5 files on GPFS through a complex interface and imposes synchronizations. On the other hand, by using Apache Cassandra or ScyllaDB through Hecuba, the application code is greatly simplified. Also, thanks to the key-value data model the number of synchronizations is reduced and the time dedicated to I/O scales when increasing the number of nodes.

Abstract: The recurrent neural network (RNN) models require longer training time with larger datasets and bigger number of parameters. Distributed training with large mini-batch size is a potential solution to accelerate the whole training process. This paper proposes a framework for large-scale training RNN/LSTM on the Sunway TaihuLight (SW) supercomputer. We take series of architecture-oriented optimizations for the memory-intensive kernels in RNN models to improve the computing performance. The lazy communication scheme with improved communication implementation and the distributed training and testing scheme are proposed to achieve high scalability for distributed training. Furthermore, we explore the training algorithm with large mini-batch size, in order to improve convergence speed without losing accuracy. The framework supports training RNN models with large size of parameters with at most 800 training nodes. The evaluation results show that, compared to training with single computing node, training based on proposed framework can achieve a 100-fold convergence rate with 8,000 mini-batch size.

Abstract: When a high population immunity already exists for a dis- ease, heterogeneities become more important to understand the spread of this disease. Individual-based models are suited to investigate the ef- fects of these heterogeneities. Measles is a disease for which, in many regions, high population immunity exists. However, different levels of immunity are observed for different age groups. For example, the gen- eration born between 1985 and 1995 in Flanders is incompletely vacci- nated, and thus has a higher level of susceptibility. As time progresses, this peak in susceptibility will shift to an older age category. Simultane- ously, susceptibility will increase due to the waning of vaccine-induced immunity. Older generations, with a high degree of natural immunity, will, on the other hand, eventually disappear from the population. Us- ing an individual-based model, we investigate the impact of changing age-dependent immunity levels (projected for Flanders, for years 2013 to 2040) on the risk for measles outbreaks. We find that, as time pro- gresses, the risk for measles outbreaks increases, and outbreaks tend to be larger. As such, it is important to not only consider infants when designing strategies for measles elimination, but to also take other age categories into account.

Abstract: Finding the regions where people appear plays a key role in many fields like user behavior analysis, urban planning, etc. Therefore, as the first step, how to partition the world, especially the urban areas where people are crowd and active, into regions is very crucial. In this paper, we propose a novel method called Restricted Spatial-Temporal DBSCAN(RST-DBSCAN). The key idea is to partition busy urban areas based on spatial-temporal information. Arbitrary and separated shapes of regions in urban areas would be then obtained. Besides, we would further get busier region earlier by RST-DBSCAN. Experimental results show that our approach yields significant improvements over existing methods on a real-world dataset extracted from Gowalla, a location-based social network.

Abstract: Although most of wealth and innovation have been the result of human interaction and cooperation, we are not yet able to quantitatively predict the spatial distributions of three main elements of cities: population, roads, and socioeconomic interactions. By a simple model mainly based on spatial attraction and matching growth mechanisms, we reveal that the spatial scaling rules of these three elements are in a consistent framework, which allows us to use any single observation to infer the others. All numerical and theoretical results are consistent with empirical data from ten representative cities. In addition, our model can also provide a general explanation of the origins of the universal super- and sub-linear aggregate scaling laws and accurately predict kilometre-level socioeconomic activity. And the theoretical analysis method is original which is based on growth instead of mean-field assumptions. The active population (AP) concept proposed by us is another contribution, which is a mixture of residential and working populations according to the duration of their activities in the region. AP is a more appropriate proxy than simply residential population for estimating socioeconomic activities. The density distribution of AP is ρ(r)∝r^(-β) (R_t^(1+β)-r^(1+β) )~r^(-β) which can also reconcile the conflict between area-size allometry and the exponential decay of population from city centre to urban fringe found in the literature. Our work opens a new avenue for uncovering the evolution of cities in terms of the interplay among urban elements, and it has a broad range of applications.

Abstract: Mention as a key feature on social networks can break through the effect of structural trapping and expand the visibility of a message. Although existing works usually use rank learning as implementation strategy before performing mention recommendation, these approaches may interfere with the influening factor exploration and cause some biases. In this paper, we propose a novel Context-aware Mention recommendation model based on Probabilistic Matrix Factorization (CMPMF). This model considers four important mention contextual factors including topic relevance, mention affinity, user profile similarity and message semantic similarity to measure the relevance score from users and messages dimensions. We fuse these mention contextual factors in latent spaces into the framework of probabilistic matrix factorization to improve the performance of mention recommendation. Through evaluation on a real-world dataset from Weibo, the empirically study demonstrates the effectiveness of discovered mention contextual factors. We also observe that topic relevance and mention affinity play a much significant role in the mention recommendation task. The results demonstrate our proposed method outperforms the state-of-the-art algorithms.

Abstract: After a sudden catastrophic event occurring in a population of individuals, panic can spread, persist and become more problematic than the catastrophe itself. In this paper, we explore through a computational approach the possibility to control the panic level in complex networks built with a recent behavioral model. After stating a rigorous theoretical framework, we propose a numerical investigation in order to establish the effect of the topology of the network on this control process, with randomly generated networks, and we compare the panic level for two distinct topology sets on a given network.

Abstract: Centrality has long been studied as a method of identifying node importance in networks. In this paper we study a variant of several walk-based centrality metrics based on the notion of a nonbacktracking walk, where the pattern iji is forbidden in the walk. Specifically, we focus our analysis on dynamic graphs, where the underlying data stream the network is drawn from is constantly changing. Efficient algorithms for calculating nonbactracking walk centrality scores in static and dynamic graphs are provided and experiments on graphs with several million vertices and edges are conducted. For the static algorithm, comparisons to a traditional linear algebraic method of calculating scores show that our algorithm produces scores of high accuracy within a theoretically guaranteed bound. Comparisons of our dynamic algorithm to the static show speedups of several orders of magnitude as well as a significant reduction in space required.

Abstract: Recognizing and diagnosing the state of traffic flow is an important research area, which is the basis of improving the level of traffic management and the quality of traffic information services. However, due to the increasing amount of traffic data collected, the traffic management system is facing the problem of "information surplus". After finishing several process, including data preprocessing, attribute reduction and rule acquisition, finally obtained the knowledge rules of the traffic flow’s state. Using the method of knowledge discovery can reveal some hidden, unknown and valuable information from the huge amount of traffic flow information, so as to provide rules and decisionmaking basis for traffic management department.

Abstract: Factor investing is one king of quantitative investing methodologies for portfolio construction based on factors. Factors with different style are extracted from multiple sources such as market data, fundamental information from financial statements, sentimental information from the Internet, etc. Numerous style factors are defined by Barra model proposed by Morgan Stanley Capital International(MSCI) to explain the return a portfolio. Multiple factors are usually integrated linearly when being put to use, which ensure stability of the process of integration and enhance the effectiveness of integrated factors. In this work, we integrate factors by machine learning and deep learning methodologies to explore deeper information among multiple style factors defined by MSCI Barra model. Multi-factors indexes are compiled using Smart Beta Index methodology proposed by MSCI. And the results shows non-linear integration by deep neural network can enhance the profitability and stability of the index compiled according to the integrated factor.

Abstract: As a core task and important part of Information Extraction，Entity Relation Extraction can realize the identification of the semantic relation between entity pairs. And it plays an important role in semantic understanding of sentences and the construction of entity knowledge base. It has the potential of employing distant supervision method, end-to-end model and other deep learning model with the creation of large datasets. In this review, we compare the contributions and defect of the various models that have been used for the task, to help guide the path ahead.

Abstract: Traffic congestion seriously affect citizens’ life quality. Many researchers have paid much attention to the task of short-term traffic congestion forecasting. However, the performance of the traditional traffic congestion forecasting approaches is not satisfactory. Moreover, most neural network models cannot capture the features at different moments effectively. In this paper, we propose an Attention-based long short-term memory (LSTM) recurrent neural network. We evaluate the prediction architecture on a real-time traffic data from Gray-Chicago-Milwaukee (GCM) Transportation Corridor in Chicagoland. The experimental results demonstrate that our method outperforms the baselines for the task of congestion prediction.

Abstract: Contemporary development in personalized medicine based both on extended clinical records and implementation of different high-throughput “omics” technologies has generated large amounts of data. To make use of these data, new approaches need to be developed for their search, storage, analysis, integration and processing. In this paper we suggest an approach for integration of data from diverse domains and various information sources enabling extraction of novel knowledge in cancer studies. Its application can contribute to the early detection and diagnosis of cancer as well as to its proper personalized treatment. The data used in our research consist of clinical records from two particular cancer studies with different factors and different origin, and also include gene expression datasets from different high-throughput technologies – microarray and next generation sequencing. An especially developed workflow, able to deal effectively with the heterogeneity of data and the enormous number of relations between patients and proteins, is used to automate the data integration process. During this process, our software tool performs advanced search for additional expressed protein relationships in a set of available knowledge sources and generates semantic links to them. As a result, a set of hidden common expressed protein mutations and their subsequent relations with patients is generated in the form of new knowledge about the studied cancer cases.

Abstract: Purkinje fibers are fundamental structures in the process of the electrical stimulation of the heart. To allow the contraction of the ventricle muscle, these fibers need to stimulate the myocardium in a syn- chronized manner. However, certain changes in the properties of these fibers may provide a desynchronization of the heart rate. This can oc- cur through failures in the propagation of the electrical stimulus due to conduction blocks occurring at the junctions that attach the Purkinje fibers to the ventricle muscle. This condition is considered a risk state for cardiac arrhythmias. The aim of this work is to investigate and analyze which properties may affect the propagation velocity and activation time of the Purkinje fibers, such as cell geometry, conductivity, coupling of the fibers with ventricular tissue and number of bifurcations in the network. In order to reach this goal, several Purkinje networks were generated by varying these parameters to perform a sensibility analysis. For the implementation of the computational model, the monodomain equation was used to describe mathematically the phenomenon and the numeri- cal solution was calculated using the Finite Volume Method. The results of the present work were in accordance with those obtained in the lit- erature: the model was able to reproduce certain behaviors that occur in the propagation velocity and activation time of the Purkinje fibers. In addition, the model was able to reproduce the characteristic delay in propagation that occurs at the Purkinje-muscle junctions.

Abstract: Tertiary protein structure prediction is one of the most challenging problems in Structural Bioinformatics, and it is an NP-Complete problem in computational complexity theory. The complexity is related to the significant number of possible conformations a single protein can assume. Metaheuristics became useful algorithms to find feasible solutions in viable computational time since exact algorithms are not capable. However, these stochastic methods are highly-dependent from parameter tuning for finding the balance between exploitation and exploration capabilities. Thus, self-adaptive techniques were created to handle the parameter definition task, since it is time-consuming. In this paper, we enhance the Self-Adaptive Differential Evolution with problem-domain knowledge provided by the angle probability list approach, comparing it with every single mutation we used to compose our set of mutation operators. Moreover, a population diversity metric is used to analyze the behavior of each one of them. The proposed method was tested with ten protein sequences with different folding patterns. Results obtained showed that the self-adaptive mechanism has a better balance between the search capabilities, providing better results regarding root mean square deviation and potential energy than the non-adaptive single-mutation methods.

Abstract: The prediction of protein structures is one of the most challenging problems in Structural Bioinformatics. In this paper, we present some variations of the Artificial Bee Colony algorithm to deal with the problem by the introducing of multi-objective optimization and knowledge from experimental proteins through the use of protein contact maps. Obtained results indicate that our approaches surpassed their previous version, demonstrating the importance of adapting the algorithm to deal with the particularities of the problem.

Abstract: Over the last decade, computational modeling has proved a useful tool to simulate and predict nosocomial transmission of pathogens and optimal control measures in healthcare settings. Nosocomial infections are a major public health issue espe-cially since the worldwide increase of antimicrobial resistance worldwide. Here, we present CTCmodeler, a framework that incorporate an agent-based model to simulate pathogen transmission through inter-individual contact in a hospital set-ting. CTCmodeler uses real admission, swab and contact data to deduce its own parameters, simulates individual-mediated transmission across hospital wards and produces weekly incidence estimates. Most earlier hospital models did not take into account the individual heterogeneity of contact patterns. By contrast, CTCmodeler explicitly captures temporal heterogeneous individual contact dy-namics by modelling close proximity interactions over time. Here, we illustrate how CTCmodeler may be used to simulate methicillin-resistant Staphylococcus aureus dissemination in a French long-term care hospital, using longitudinal data on sensor-recorded contacts and weekly swabs from the i-Bird study.

Abstract: Multi-criteria optimization problems belong to the hardest computational problems tackled, thus metaheuristic-based approach is necessary in order to deal with them. Evolutionary algorithms, swarm intelligence methods and other are very often used in such cases. Based on well-known ``no free lunch theorem'' there is always a need for creating new metaheuristics, though according to Sorensen, they should not be proposed without a firm background. In this paper a socio-cognitive ACO-type algorithm is proposed for multi-criteria TSP problem optimization. This algorithm is rooted in psychological inspirations and follows other socio-cognitive swarm intelligence methods proposed up to now. This paper presents the idea and shows the applicability of the proposed algorithm based on selected benchmark functions from the scope of well-known TSPLIB library.

Abstract: One of the key problems in parallel processing systems is the architecture of internodal connections, thus affecting the computational efficiency of the whole. In this work authors describe proposition of a new multi-channel hierarchical computational environment with distributed passive switching. According to authors, improvement of communication efficiency should be based on grouping of system components (processing nodes and channels). In the first group, processing nodes are combined into independent groups that communicate using a dedicated channel group. The second type of clustering groups channels available in the system. In particular, they are divided into smaller independent fragments that can be combined into clusters that support selected users. In this work, a model of proposed computational environment and basic reconfiguration protocol were described. The necessary components and management of reconfiguration, passive switching and hierarchization were discussed, highlighting related problems to be solved.

Abstract: This paper presents a model and implementation of a multi-agent system to support decisions to optimize a configuration of the production process in an company. Our goal is to choose the most desirable parameters of the technological process using computer simulation, which will help to avoid or reduce the number of much more expensive trial production processes, using physical production lines. These identified values of production process parameters will be applied later in the real mass production. Decision-making strategies are selected using different machine learning techniques that assist in obtaining products with the required parameters, taking into account sets of historical data. The focus was primarily on the analysis of the quality of prediction of the obtained product parameters for the different algorithms used and different sizes of historical data sets, and therefore different details of information, and secondly on the examining of the times necessary for building decision models for individual algorithms and data sets. The following algorithms were used: Multilayer Perceptron, Bagging, RandomForest, M5P and Voting. The experiments presented were carried out using data obtained for foundry processes. The JADE platform and the Weka environment were used to implement the multi--agent system.

Abstract: The increase of computing capabilities promises to address many scientific and engineering problems by enabling simulations to reach new levels of accuracy and scale. The field of uncertainty quantification (UQ) has recently been receiving an increasing amount of attention as it enables reliability study of modelled systems. However, performance of UQ analysis for high-fidelity simulations remains challenging due to exceedingly high complexity of computational workflows. In this paper, we present a UQ study on a complex workflow targeting a thermally stratified flow. We discuss different models that can be used to enable it. We then propose an abstraction at the level of the workflow specification that enables the modeller to quickly switch between UQ models and manage underlying compute infrastructure in a completely transparent way. We show that we can keep the workflow description almost unchanged while benefitting of all the insight the UQ study provides.

Abstract: We propose methods to eliminate redundant utility assessments in decision analysis applications. We abstract a set of utility assessments such that the set is represented as a matrix of ternary arrays. To achieve efficiency, arrays converted to decimal numbers for further processing. The resulting approach demonstrates excellent performance on random sets of utility assessments. The method eliminates the redundant questions for the decision maker and can serve for consistency check.

Abstract: In this paper, we propose an algorithm for designing a cryptosystem, in which the derivative disproportion functions are used. The symbols to be transmitted are encoded with the sum of at least two of these functions combined with random coefficients. A new algorithm is proposed for decoding the received messages by making use of important properties of the derivative disproportion functions. Numerical experiments are demonstrating the algorithm’s reliability and robustness.

Abstract: We study the effects of merging two separate markets each originally monopolized by a producer into a globalized duopoly market. We consider a linear inverse demand with cap price and quadratic cost functions. After globalization, we find the consistent conjectural variations equilibrium (CCVE) of the duopoly game. Unlike in the Cournot equilibrium, a complete symmetry (identical cost functions parameters of both firms) does not imply the strongest coincident profit degradation. For the situation where both agents are low-marginal cost firms, we find that the company with a technical advantage over her rival has a better ratio of the current and previous profits. Moreover, as the rival becomes ever weaker, that is, as the slope of the rival’s marginal cost function increases, the profit ratio improves.

Abstract: The element of the epistemic uncertainty quantification concerning the estimation of the approximation error is analyzed from the viewpoint of the ensemble of numerical solutions obtained via independent numerical algorithms. The analysis is based on the geometry considerations: the triangle inequality and measure concentration in spaces of great dimension. In result, the feasibility for nonintrusive postprocessing appears that provides the approximation error estimation on the ensemble of the solutions. The ensemble of numerical results obtained by five OpenFOAM solvers is analyzed. The numerical tests were made for the inviscid compressible flow around a cone at zero angle of attack and demonstrated the successful estimation of the approximation error.

Abstract: Yield estimation of antenna systems is important to check their robustness with respect to the uncertain sources. Since the Monte Carlo sampling-based real physics simulation model evaluations are computationally intensive, this work proposes the polynomial chaos-Kriging (PC-Kriging) metamodeling technique for fast yield estimation. PC-Kriging integrates the polynomial chaos expansion (PCE) as the trend function of Kriging metamodel since the PCE is good at cap-turing the function tendency and Kriging is good at matching the observations at training points. The PC-Kriging is demonstrated with an analytical case and a multi-band patch antenna case and compared with direct PCE and Kriging meta-models. In the analytical case, PC-Kriging reduces the computational cost by around 42% compared with PCE and over 94% compared with Kriging. In the antenna case, PC-Kriging reduces the computational cost by over 60% compared with Kriging and over 90% compared with PCE. In both cases, the savings are obtained without compromising the accuracy.

Abstract: Quasi-Newton methods are popular gradient-based optimization methods that can achieve rapid convergence using only first-order derivatives. However, the choice of the initial Hessian matrix upon which quasi-Newton updates are applied is an important factor that can significantly affect the performance of the method. This fact is especially true for limited-memory variants, which are widely used for large-scale problems where only a small number of updates are applied in order to minimize the memory footprint. In this paper, we introduce both a scalar and a sparse diagonal Hessian initialization framework, and we investigate its effect on the restricted Broyden-class of quasi-Newton methods. Our implementation in PETSc/TAO allows us to switch between different Broyden class methods and Hessian initializations at runtime, enabling us to quickly perform parameter studies and identify the best choices. The results indicate that a sparse Hessian initialization based on the diagonalization of the BFGS formula significantly improves the base BFGS methods and that other parameter combinations in the Broyden class may offer competitive performance.

Abstract: Electromagnetic (EM) analysis is the primary tool utilized in the design of high-frequency structures. In the vast majority of cases, simpler models (e.g., equivalent networks or analytical ones) are either not available or lack accuracy: they can only be used to yield initial designs that need to be further tuned. Consequently, EM-driven adjustment of geometry and/or material parameters of microwave and antenna components is a necessary design stage. This, however, is a computationally expensive process, not only because of the considerable computational cost of high-fidelity EM analysis but also due to a typically large number of parameters that need to be adjusted. In particular, conventional numerical optimization routines (both local and global) may be prohibitively expensive. In this paper, a reduced-cost trust-region-based gradient search algorithm is proposed for optimization of high-frequency components. Our methodology is based on a smart management of the system Jacobian enhancement which combines omission of (finite-differentiation-based) sensitivity updates for variables that exhibit small (relative) relocation in the directions of the corresponding coordinate system axes and selective utilization of a rank-one Broyden updating formula. Parameter selection for Broyden-based updating depends on the alignment between the direction of the latest design relocation and respective search space basis vectors. The proposed technique is demonstrated using an ultra-wideband antenna and a miniaturized coupler. In both cases, a significant reduction of the number of EM simulations involved in the optimization process is achieved as compared to the benchmark algorithm. At the same time, degradation of the design quality is minor.

Abstract: Reducing the measurement effort that is made for identification of parameters is an important task in some fields of technology. This work focuses on calibration of functions running on the electronic control unit (ECU), where measurements are the main expense factor. An algorithm for information content analysis of recorded measurement data is introduced that places the calibration engineer in the position to shorten future test runs. The analysis is based upon parameter sensitivities and utilizes the Fisher-information matrix to determine the value of certain measurement portions with respect to parameter identification. By means of a simple DC motor model the algorithm's working principle is illustrated. The first use on a real ECU function achieves a measurement time reduction of 67% while a second use case opens up new features for the calibration of connected cars.

Abstract: Large scale cell based blood flow simulations are expensive, both in time and resource requirements. HemoCell can perform such simulations on high performance computing resources by dividing the simulation domain into multiple blocks. This division has a performance impact caused by the necessary communication between the blocks. In this paper we implement an efficient algorithm for computing the mechanical model for HemoCell together with an improved communication structure. The result is an up to $4$ times performance increase for blood flow simulations performed with HemoCell.