Computational Science – ICCS 2019

Aspect Based Sentiment Analysis is a type of fine-grained sentiment analysis. It is popular in both industry and academic communities, since it provides more detailed information on the user generated text in product reviews or social network. Therefore, we propose a novel framework based on neural network to determine the polarity of a review given a specific target. Not only the words close to the target but also the words far from the target determine the polarity of the review given a certain target, so we use self attention to solve the problem of long distance dependence. Briefly, we do multiple linear mapping on the review, do multiple attention and combine them to attend to the information from different representation sub-spaces. Besides, we use domain embedding to get close to the real word embedding in a certain domain, since the meaning of the same word may be different in different situation. Moreover, we use position embedding to underline the target and pay more attention to the words that are close to the target to get better performance on the task. We validate our model on four benchmarks, they are SemEval 2014 restaurant dataset, SemEval 2014 laptop dataset, SemEval 2015 restaurant dataset and SemEval 2016 restaurant dataset. The final results show that our model is effective and strong, which brings a 0.74% boost averagely based on the previous state-of-the-art work.

Law enforcement agencies and private security companies work to prevent, detect and counteract any threat with the resources they have, including alarms and video surveillance. Even so, there are still terrorist attacks or shootings in schools in which armed people move around a venue exercising violence and generating victims, showing the limitations of current systems. For example, they force security agents to monitor continuously all the images coming from the installed cameras, and potential victims nearby are not aware of the danger until someone triggers a general alarm, which also does not give them information on what to do to protect themselves. In this article we present a project that is being developed to apply the latest technologies in early threat detection and optimal response. The system is based on the automatic processing of video surveillance images to detect weapons and a mobile app that serves both for detection through the analysis of mobile device sensors, and to send users personalised and dynamic indications. The objective is to react in the shortest possible time and minimise the damage suffered.

This paper is devoted to new aspects of Natural Language Interface to Relational Database (NLIDB) integration into third-party corporate environments related to control data access. Because there is no schema information in the input NL query and the different relational database management system (RDBMS) requires different meta-data types and rules to control data access, developers meet a problem addressed to automatic data access control in the case of NL interface implementation to relational databases. In the paper, we suggest a comprehensive approach which takes into account permissions throughout the pipeline of transforming NL query into SQL query with an intermediate SPARQL representation. Our integration solutions based on well-known Ontology Based Data Access (OBDA) approach, which gives us the opportunity to adapt the proposed solutions to the specifics of access control facilities of various RDBMS. Suggested approach has been implemented within intelligent service, named Reply.

Cross-modal retrieval has attracted increasing attention in recent years. Deep supervised hashing methods have been widely used for cross-modal similarity retrieval on large-scale datasets, because the deep architectures can generate more discriminative feature representations. Traditional hash methods adopt a symmetric way to learn the hash function for both query points and database points. However, those methods take an immense amount of work and time for model training, which is inefficient with the explosive growth of data volume. To solve this issue, an Asymmetric Deep Cross-modal Hashing (ADCH) method is proposed to perform more effective hash learning by simultaneously preserving the semantic similarity and the underlying data structures. More specifically, ADCH treats the query points and database points in an asymmetric way. Furthermore, to provide more similarity information, a detailed definition for cross-modal similarity matrix is also proposed. The training of ADCH takes less time than traditional symmetric deep supervised hashing methods. Extensive experiments on two widely used datasets show that the proposed approach achieves the state-of-the-art performance in cross-modal retrieval.

In Dial-a-Ride Problem (DARP) customers request from an operator a transportation service from a pick-up to a drop-off place. Depending on the formulation, the problem can address several constraints, being associated with problems such as door-to-door transportation for elderly/disabled people or occasional private drivers. This paper addresses the latter case where a private drivers company transports passengers in a heterogeneous fleet of saloons, estates, people carriers and minibuses. The problem is formulated as a multiple objective DARP which tries to minimize the total distances, the number of empty seats, and the wage differential between the drivers. To solve the problem a Non-dominated Sorting Genetic Algorithm-II is hybridized with a local search. Results for daily scheduling are shown.

The number of users of the parking lots from the campus of the Polytechnic of Leiria, a higher education institution in Portugal, has been increasing each year. It is becoming a major concern to the organization to address the high demand for a free parking spot on campus. In order to ease this problem, this paper proposes the design of a smart parking system that can help users to easily find a free parking spot, using an integrated system that includes sensors and a mobile application.The system is based on the information about the occupation status of parking lots generated by parking sensors. This information is available in the mobile application that consumes a REST webservice and is presented to end-users, thus contributing to the decrease of time wasted on the quest of finding a free spot. The software architecture consists on a set of decoupled modules that compute and share the information generated by sensors. This architectural approach is noteworthy because it maximizes the system scalability and responsiveness to change. It allows the system to expand with the integration of new applications and perform updates on the existing ones, without an overall impact on the operations of the other system modules.

The Internet of Things (IoT) is a novel paradigm based on the connectivity among different entities or “things”. IoT environment in the form of interconnected smart “things” represents a great potential in terms of effective and efficient solutions related to urban context (e.g., system architecture, design and development, human involvement, data management and applications). On the other hand, with the introduction of the IoT environment, devices and network security have become a fundamental and challenging issue. Moreover, growing number of users connected via IoT system necessitates focusing on the vulnerability of complex networks and defence challenges at the topological level. This paper addresses these challenges from the perspective of graph theory. In this work, the authors introduce a novel AV11 algorithm to identify the most critical and influential IoT nodes in a Social IoT (SIoT) network in a smart city context using ENEA Portici CRESCO infrastructure.

Interactive visualization of spatio-temporal data is a very active area that has experienced remarkable advances in the last decade. This is due to the emergence of fields of research such as big data and advances in hardware that allow better analysis of information. This article describes the methodology followed and the design of an open source tool, which in addition to interactively visualizing spatio-temporal data that are represented in an ontology, allows the definition of what to visualize and how to do it. The tool allows selecting, filtering and visualizing in a graphical way the entities of the ontology with spatiotemporal data, as well as the instances related to them. The graphical elements used to display the information are specified on the same ontology, extending the VISO graphic ontology, used for mapping concepts to graphic objects with RDFS/OWL Visualization Language (RVL). This extension contemplates the data visualization on rich real-time 3D environments, allowing different modes of visualization according to the level of detail of the scene, while also emphasizing the treatment of spatio-temporal data, very often used in cultural heritage models. This visualization tool involves simple visualization scenarios and high interaction environments that allow complex comparative analysis. It combines traditional solutions, like hypercube or time-animations with innovative data selection methods.

The paper is devoted to the development of high-level tools to automate tangible human-machine interfaces creation bringing together IoT technologies and ontology engineering methods. We propose using ontology-driven approach to enable automatic generation of firmware for the devices and middleware for the applications to design from scratch or transform the existing M2M ecosystem with respect to new human needs and, if necessary, to transform M2M systems into human-centric ones. Thanks to our previous research, we developed the firmware and middleware generator on top of SciVi scientific visualization system that was proven to be a handy tool to integrate different data sources, including software solvers and hardware data providers, for monitoring and steering purposes. The high-level graphical user SciVi interface enables to design human-machine communication in terms of data flow and ontological specifications. Thereby the SciVi platform capabilities are sufficient to automatically generate all the necessary components for IoT ecosystem software. We tested our approach tackling the real-world problems of creating hardware device turning human gestures into semantics of spatiotemporal deixis, which relates to the verbal behavior of people having different psychological types. The device firmware generated by means of SciVi tools enables researchers to understand complex matters and helps them analyze the linguistic behavior of users of social networks with different psychological characteristics, and identify patterns inherent in their communication in social networks.

Reliable vessel re-identification would enable maritime surveillance systems to analyze the behavior of vessels by drawing their accurate trajectories, when they pass along different camera locations. However, challenging outdoor conditions and varying viewpoint appearances combined with the large size of vessels limit conventional methods to obtain robust re-identification performance. This paper employs CNNs to address these challenges. In this paper, we propose an Identity Oriented Re-identification network (IORnet), which improves the triplet method with a new identity-oriented loss function. The resulting method increases the feature vector similarities between vessel samples belonging to the same vessel identity. Our experimental results reveal that the proposed method achieves 81.5% and 91.2% on mAP and Rank1 scores, respectively. Additionally, we report experimental results with data augmentation and hyper-parameters optimization to facilitate reliable ship re-identification. Finally, we provide our real-world vessel re-identification dataset with various annotated multi-class features to public access.

Indoor localisation is a fundamental problem in robotics, which has been the subject of several research works over the last few years. Indeed, while solutions based on fusion of inertial and global navigation satellite system (GNSS) measurements have proved their efficiency in outdoor environments, indoor localisation remains an open research problem. Although commercial motion tracking systems offer very accurate position estimation, their high cost cannot be afforded by all research laboratories. This paper presents an indoor localisation solution based on a multi-camera setup. The proposed system relies on low-cost sensors, which makes it very affordable compared to commercial motion-tracking systems. We show through the experiments conducted that the proposed approach, although being cheap, can provide real-time position measurements with an error of less than 2 cm up to a distance of 2 m.

In our previous work on automated microalgae classification system we proposed the multi-resolution image segmentation that can handle well with unclear boundary of algae bodies and noisy background, since an image segmentation is the most important preprocessing step in object classification and recognition. The previously proposed approach was able to classify twelve genera of microalgae successfully; however, when we extended it to work with new genera of filamentous algae, new challenging problems were encountered. These difficulties arise due to a variety of the forms of filamentous algae, which complicates both image segmentation and classification processes, resulting in substantial degradation of classification accuracy. Thus, in this work we propose a modified version of our multi-resolution segmentation algorithm by combining them in such a way that the strengths of both algorithms complement each other’s weaknesses. We also propose a new skeleton-based shape descriptor to alleviate an ambiguity caused by multiple morphologies of filamentous forms of algae in classification process. Effectiveness of the two proposed approaches are evaluated on five genera of filamentous microalgae. SMO is used as a classifier. Experimental result of 91.30% classification accuracy demonstrates a significant improvement of our proposed approaches.

Monitoring public space with imaging sensors to perform an object- or person-tracking is often associated with privacy concerns. We present a Dynamic Vision Sensor (DVS) based approach to achieve this tracking that does not require the creation of conventional grey- or color images. These Dynamic Vision Sensors produce an event-stream of information, which only includes the changes in the scene.The presented approach for tracking considers the scenario of fixed mounted sensors. The method is based on clustering events and tracing the resulting cluster centers to accomplish the object tracking. We show the usability of this approach with a first proof-of-concept test.

One of the most crucial steps of preprocessing of document images subjected to further text recognition is their binarization, which influences significantly obtained OCR results. Since for degrades images, particularly historical documents, classical global and local thresholding methods may be inappropriate, a challenging task of their binarization is still up-to-date. In the paper a novel approach to the use of Generalized Gaussian Distribution for this purpose is presented. Assuming the presence of distortions, which may be modelled using the Gaussian noise distribution, in historical document images, a significant similarity of their histograms to those obtained for binary images corrupted by Gaussian noise may be observed. Therefore, extracting the parameters of Generalized Gaussian Distribution, distortions may be modelled and removed, enhancing the quality of input data for further thresholding and text recognition. Due to relatively long processing time, its shortening using the Monte Carlo method is proposed as well. The presented algorithm has been verified using well-known DIBCO datasets leading to very promising binarization results.

This paper presents a comparative analysis of algorithms belonging to manifold learning and linear dimensionality reduction. Firstly, classical texture image descriptors, namely Gray-Level Co-occurrence Matrix features, Haralick features, Histogram of Oriented Gradients features and Local Binary Patterns are combined to characterize and discriminate textures. For patches extracted from several texture images, a concatenation of the image descriptors is performed. Using four algorithms to wit Principal Component Analysis (PCA), Locally Linear Embedding (LLE), Isometric Feature Mapping (ISOMAP) and Laplacian Eigenmaps (Lap. Eig.), dimensionality reduction is achieved. The resulting learned features are then used to train four different classifiers: k-nearest neighbors, naive Bayes, decision tree and multilayer perceptron. Finally, the non-parametric statistical hypothesis test, Wilcoxon signed-rank test, is used to figure out whether or not manifold learning algorithms perform better than PCA. Computational experiments were conducted using the Outex and Salzburg datasets and the obtained results show that among twelve comparisons that were carried out, PCA presented better results than ISOMAP, LLE and Lap. Eig. in three comparisons. The remainder nine comparisons did not presented significant differences, indicating that in the presence of huge collections of texture images (bigger databases) the combination of image feature descriptors or patches extracted directly from raw image data and manifold learning techniques is potentially able to improve texture classification.

Keyphrase extraction, as a basis for many natural language processing and information retrieval tasks, can help people efficiently discover their interested information from vast streams of online documents. Previous methods are mostly proposed in general purpose, where keyphrases that represent the main topics are extracted. However, such keyphrases can hardly distinguish events from massive streams of long text documents that share similar topics and contain highly redundant information. In this paper, we address the task of keyphrase extraction for event-oriented retrieval. We propose a novel bi-clustering model for clustering the documents and keyphrases simultaneously. The model consequently makes the extracted keyphrases more specific and related to the event. We conduct a series of experiments on a real-world dataset. The experimental results demonstrate the better performance of our approach than other unsupervised approaches.

This article introduces a path-finding method based on evolutionary algorithms and a fully vectorized GPU implementation of it. The algorithm runs on real-time and it can handle dynamic obstacles in maps of arbitrary size. The experiments show the proposed approach outperforms other traditional path-finding algorithms (e.g. A*). The conclusions present further improvement possibilities to the proposed approach like the application of multi-objective algorithms to represent full crowd models.

As the result of the intense research activity of the past decade, Semantic Web technology has achieved a notable popularity and maturity. This technology is leading the evolution of the Web via interoperability by providing structured metadata. Because of the adoption of rich data models on a large scale to support the representation of complex relationships among concepts and automatic reasoning, the computational performance of ontology-based systems can significantly vary. In the evaluation of such a performance, a number of critical factors should be considered. Within this paper, we provide an empirical framework that yields an extensive analysis of the computational performance of ontology-based systems. The analysis can be seen as a decision tool in managing the constraints of representational requirements versus reasoning performance. Our approach adopts synthetic ontologies characterised by an increasing level of complexity up to OWL 2 DL. The benefits and the limitations of this approach are discussed in the paper.

We present a framework for the distributed approximation of moments, enabling the evaluation of the uncertainty in a dynamical system. The first and second moment, mean, and variance are computed with up to third-order Taylor series expansion. The required derivatives for the expansion are generated automatically by automatic differentiation and propagated through an implicit time stepper. The computational kernels are the accumulation of the derivatives (Jacobian, Hessian, tensor) and the covariance matrix. We apply distributed parallelism to the Hessian or third-order tensor, and the user merely has to provide a function for the differential equation, thus achieving similar ease of use as Monte Carlo-based methods. We demonstrate our approach using with benchmarks on Theta, a KNL-based system at the Argonne Leadership Computing Facility.

The main purpose of this paper is modelling and solving boundary value problems simultaneously considering uncertainty of all of input data such as: shape of boundary, boundary conditions and other parameters. The strategy is presented on the basis of problems described by Navier-Lamé equations. Therefore, the uncertainty of parameters here, means the uncertainty of the Poisson’s ratio and Young’s modulus. For solving uncertainly defined problems we use implementation of interval parametric integral equations system method (IPIES). In this method we propose modification of directed interval arithmetic for modeling and solving uncertainly defined problems. We consider an examples of uncertainly defined, 2D elasticity problems. We present boundary value problems with linear as well as curvelinear (modelled using NURBS curves) shape of boundary. We verify obtained interval solutions by comparison with precisely defined (without uncertainty) analytical solutions. Additionally, to obtain errors of such solutions, we decided to use the total differential method. We also analyze influence of input data uncertainty on interval solutions.

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.

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 numbers. To achieve efficiency, the matrix is converted to a decimal vector 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.

The paper develops an algorithm based on derivative disproportion functions (DDF) for modeling a cryptosystem for transmitting and receiving devices. The transmitted symbols are encoded with the aid of sums of at least two of those functions weighted with random coefficients. Some important properties of the derivative disproportion functions are also discussed. Numerical experiments demonstrate that the algorithm is quite reliable and robust.

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.

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.

The task of assessing accuracy in mathematical modeling of gas-dynamic processes is of utmost importance and relevance. Modern software packages include a large number of models, numerical methods and algorithms that allow solving most of the current CFD problems. However, the issue of obtaining a reliable solution in the absence of experimental data or any reference solution remains relevant. The paper provides a brief overview of some useful approaches to solving the problem, including such approaches as a multi-model approach, the study of an ensemble of solutions, the construction of a generalized numerical experiment.

By their nature, the composition of black box models is opaque. This makes the ability to generate explanations for the response to stimuli challenging. The importance of explaining black box models has become increasingly important given the prevalence of AI and ML systems and the need to build legal and regulatory frameworks around them. Such explanations can also increase trust in these uncertain systems. In our paper we present RICE, a method for generating explanations of the behaviour of black box models by (1) probing a model to extract model output examples using sensitivity analysis; (2) applying CNPInduce, a method for inductive logic program synthesis, to generate logic programs based on critical input-output pairs; and (3) interpreting the target program as a human-readable explanation. We demonstrate the application of our method by generating explanations of an artificial neural network trained to follow simple traffic rules in a hypothetical self-driving car simulation. We conclude with a discussion on the scalability and usability of our approach and its potential applications to explanation-critical scenarios.

Organizations are turning to predictive analytics to help solve difficult problems and uncover new opportunities. Nowadays, the processes are saturated in data, which must be used properly to generate the necessary key information in the decision making process. Although there are several useful techniques to process and analyze data, the main value starts with the treatment of key factors. In this way, a Predictive Factor Variance Association (PFVA) is proposed to solve a multi-class classification problem. The methodology combines well-known machine learning techniques along with linear algebra and statistical models to provide the probability that a particular sample belongs to a class or not. It can also give predictions based on regression for quantitative dependent variables and carry-out clustering of samples. The main contribution of this research is its robustness to execute different processes simultaneously without fail as well as the accuracy of the results.

Combinatorial scientific computing refers to the field of using combinatorial algorithms to solve problems in computational science and data science. Teaching even elementary topics from this area is difficult because it involves bridging the gap between scientific computing and graph theory. Furthermore, it is often necessary to understand not only the methodologies from mathematics and computer science, but also from different scientific domains from which the underlying problems arise. To enrich the learning process in combinatorial scientific computing, we designed and implemented a set of interactive educational modules called EXPLAIN. The central idea behind EXPLAIN is its focus on describing the equivalence of a problem in terms of scientific computing and graph theory. That is, in EXPLAIN, the scientific computing problem and its graph theoretical representation are treated as two sides of the same coin. The process of solving a problem is interactively explored by visualizing transformations on an object from scientific computing, simultaneously, with the corresponding transformations on a suitably defined graph. We describe the redesign of the EXPLAIN software with an emphasis on integrating a domain-specific scripting language and a hierarchical visualization for recursively defined problems.

One of the core missions of Michigan State University’s new Department of Computational Mathematics, Science, and Engineering is to provide education in computational modeling and data science to MSU’s undergraduate and graduate students. In this paper, we describe our creation of CMSE 201, “Introduction to Computational Modeling and Data Analysis,” which is intended to be a standalone course teaching students core concepts in data analysis, data visualization, and computational modeling. More broadly, we discuss the education-research-based rationale behind the “flipped classroom” instructional model that we have chosen to use in CMSE 201, which has also informed the design of other courses taught in the department. We also explain the course’s design principles and implementation.

Infiltration of data-driven computational methods of humanities research has generated mutual interests between the two communities of computer science and humanities. Larger institutions have adopted drastic structural reforms to meet the challenges to bridge the two fields. Successful examples include the integrated major programs launched at Stanford University and the collaborative workshop at Carnegie Mellon University. These types of exploratory experiments require (1) intensive resources as well as (2) strong support of faculty and administration. At a small college, both can be luxuries. We present an innovative model to carry out effective synchronized courses of computational humanities and digital humanities that pulls together efforts between two small programs and needs little additional support. This paper reviews the proposal, design, and delivery of a pair of interdisciplinary graduate courses in the small college setting. We discuss the details of our implementation and provided our observations and recommendations.

This paper describes the motivation and design of a 10-week graduate course that teaches practices for developing research software; although offered by an engineering program, the content applies broadly to any field of scientific research where software may be developed. Topics taught in the course include local and remote version control, licensing and copyright, structuring Python modules, testing and test coverage, continuous integration, packaging and distribution, open science, software citation, and reproducibility basics, among others. Lectures are supplemented by in-class activities and discussions, and all course material is shared openly via GitHub. Coursework is heavily based on a single, term-long project where students individually develop a software package targeted at their own research topic; all contributions must be submitted as pull requests and reviewed/merged by other students. The course was initially offered in Spring 2018 with 17 students enrolled, and will be taught again in Spring 2019.

The widespread of data science programming languages and libraries have raised new interest in teaching computational science coding in ways that leverage the capabilities of both single-computer and cluster-based computation infrastructures. Some of the programming patterns and idioms are converging, yet there are specialized uses and cases that require learners to switch from one to another. In this paper, we report on the experience in action research with more than ten cohorts of mixed background students in postgraduate level data science classes. We first discuss the key mental models found to be essential to understanding solution design, and then review the three fundamental paradigms that students must face when coding data manipulation and their interrelation. Finally, we discuss some insights on additional elements found important in understanding the specificities of current practice in data analysis tasks.

Designing good projects involving programming in numerical analysis for large groups of students with different backgrounds is a challenging task. The assignment has to be manageable for the average student, but to additionally inspire the better students it is preferable that it has some depth and leads to them to think about the subject. We describe a project that was assigned to the students of an introductory Numerical Analysis course at the University of Iceland. The assignment is to numerically compute the length of solution trajectories of a system of ordinary differential equations with a stable equilibrium point. While not difficult to do, the results are somewhat surprising and got the better students to get interested in what was happening. We describe the project, its solution using Matlab, and the underlying mathematics in some detail. Further, we discuss the pedagogical aspects of the project and the results in terms of its success and shortcomings.

The present work is dedicated to studying the transfer of natural gas in shale formations. The governing model was developed on the basis of the model of dual-porosity dual-permeability (DPDP). The mixed finite element method (MFEM) is employed to solve the governing equations numerically. Numerical example is presented and results discussed such as production cumulative rate, pressure and apparent permeability.

Process mining is gaining significant importance in the healthcare domain, where the quality of services depends on the suitable and efficient execution of processes. A pivotal challenge for the application of process mining in the healthcare domain comes from the growing importance of multi-centric studies, where privacy-preserving techniques are strongly needed.In this paper, building on top of the well-known Alpha algorithm, we introduce a distributed process mining approach, that allows to overcome problems related to privacy and data being spread around. The introduced technique allows to perform process mining without sharing any patients-related information, thus ensuring privacy and maximizing the possibility of cooperation among hospitals.

Roadworks are required to keep roads in acceptable condition, and to perform maintenance of essential infrastructure. Road agencies are facing the problem of how to effectively plan frequent roadworks. In this paper, we exploit Automated Planning for roadworks planning. We introduce a planning domain model that allows us to plan a set of required roadworks, over a period of time, in a large urban region, by specifying constraints to be satisfied and suitable quality metrics. Our empirical analysis shows the suitability of the proposed approach.

Present work is dedicated to development of the software for interactive visualization of results of simulation of gas dynamics problems on meshes of extra large sizes. Kitware ParaView visualization tool, which is popular among engineers and scientists is used as a frontend. The coupling of client and server instances of ParaView is used in the project. The crucial feature of the work is an application of Apache Hadoop and Apache Spark for distributed retrieving of simulation data from files on hard disk. The data is stored on the cluster in Hadoop Distributed File System (HDFS) managed by Apache Hadoop and is provided to ParaView server by Apache Spark data processing tool.

The dolphin kick assumes a vital role in swimming competitions, as it is used after dives and turns in several swimming styles. To improve the swimmer’s dolphin kick performance, flows around him were simulated. Using video footage of a male swimmer’s joint angles, a 3D model simulation was created. The flows were computed using the unstructured moving grid finite volume method to express the complicated motion of swimmers. The mesh around the swimmer is moved according to his motion. In this method, a geometric conservation law is satisfied as well as a physical one. Furthermore, the moving computational domain method is also adopted for calculation efficiency. The numerical swimmer is finally completed by a coupled computation between motion of human and fluid. In this paper, the simulation results revealed that the influence of the maximum knee oscillation angles affect the speed of the swimmer.

The demands for the optimization of particle-based methods with short-range interaction forces such as those in smoothed particle hydrodynamics (SPH) is increasing, especially for many-core architectures. However, because particle-based methods require large amount of memory access, it is challenging to obtain high efficiency for low-byte/FLOP many-core architectures. Hence, an efficient technique, the so-called “multiwalk” method, was developed in an N-body gravitational field. The key of the multiwalk method is in sharing of the interaction lists with multiple particles to offer an efficient use of the cache memory in the double-loops operation for calculating the interactions and reducing the main memory access. However, such performance improvement is not clear for the problems with short-range interaction forces such as those in SPH. In this paper, we proposed a theoretical performance model to examine the tradeoff relations between the memory and the cost of floating point operations to optimise the SPH code. We also validated the model with the wall-clock time spent on the PEZY-SCs (SC1 and SC2).

The Sub-NUMA Clustering 4 (SNC-4) affinity mode of the Intel Xeon Phi Knights Landing introduces a new environment for parallel applications, that provides a NUMA system in a single chip.The main target of this work is to characterize the behaviour of this system, focusing in nested parallelization for a well known algorithm, with regular and predictable memory access patterns as the matrix multiplication. It has been studied the effects of thread distribution in the processor on the performance when using SNC-4 affinity mode, the differences between cache and flat modes of the MCDRAM and the improvements due to vectorization in different scenarios in terms of data locality.Results show that the best thread location is the scatter distribution, using 64 or 128 threads. Differences between cache and flat modes of the MCDRAM are, generally, not significant. The use of optimization techniques as padding to improve locality has a great impact on execution times. Vectorization resulted to be efficient only when the data locality is good, specially when the MCDRAM is used as a cache.

In the last decade, planning with domains modelled in the hybrid PDDL+ formalism has been gaining significant research interest. A number of approaches have been proposed that can handle PDDL+, and their exploitation fostered the use of planning in complex scenarios. In this paper we introduce a PDDL+ reformulation method that reduces the size of the grounded problem, by reducing the arity of sparse predicates, i.e. predicates with a very large number of possible groundings, out of which very few are actually exploited in the planning problems. We include an empirical evaluation which demonstrates that these methods can substantially improve performance of domain-independent planners on PDDL+ domains.

Platforms are multi-sided marketplaces that bring together groups of users that would otherwise not have been able to connect or transact. The app markets for Apple iOS and Google Android are examples of such markets. System dynamics is a powerful method to gain useful insight into environments of dynamic complexity and policy resistance. In this paper, we argue that adapted to the context of digital business platforms, the practice of system dynamics facilitates understanding of the role of incentives in such marketplaces for increasing participation, value generation, and market growth. In particular, we describe our efforts to simulate the market competition between iOS and Android in terms of the interacting markets for devices and their apps.

Sockpuppet detection is a valuable and challenging issue in social network. Current works are continually making efforts to detect sockpuppet based on verbal, non-verbal or network-structure features. However, they do not consider the propagation characteristic and propagation structure of sockpuppet. With our observation, the propagation trees of sockpuppet and ordinary account are different. Sockpuppet’ propagation tree is evidently wider and deeper than that of the ordinary one. Based on these observations, we propose a propagation-structure based method to tackle sockpuppet detection problem. The experiment on two real-world datasets of Sina Weibo demonstrates that our method is more robust and outperforms previous methods.

Whether to use multiple threads in one process (MPI+X) or multiple processes (pure MPI) has long been an important question in HPC. Techniques like in situ analysis and visualization further complicate matters, as it may be very difficult to couple the different components in a way that would allow them to run in the same process. Combined with the growing interest in task-based programming models, which often rely on fine-grained tasks and synchronization, a question arises: Is it possible to run two tightly coupled task-based applications in two separate processes efficiently or do they have to be combined into one application? Through a range of experiments on the latest Intel Xeon Scalable (Skylake) and AMD EPYC (Zen) many-core architectures, we have compared performance of fine-grained synchronization and data exchange between threads in the same process and threads in two different processes. Our experiments show that although there may be a small price to pay for having two processes, it is still possible to achieve very good performance. The key factors are utilizing shared memory, selecting the right thread affinity, and carefully selecting the way the processes are synchronized.

In the last years, forest fire spread simulators have proven to be very promising tools in the fight against these disasters. Due to the necessity to achieve realistic predictions of the fire behavior in a relatively short time, execution time may be reduced. Moreover, several studies have tried to apply the computational power of GPUs (Graphic Processors Units) to accelerate the simulation of the propagation of fires. Most of these studies use forest fires simulators based on Cellular Automata (CA). CA approaches are fast and its parallelization is relatively easy; conversely, they suffer from precision lack. Elliptical wave propagation is an alternative approach for performing more reliable simulations. Unfortunately, its higher complexity makes their parallelization challenging. Here we explore two different parallel strategies based on Elliptical wave propagation forest fire simulators; the multicore architecture of CPU (Central Processor Unit) and the computational power of GPUs to improve execution times. The aim of this work is to assess the performance of the simulation of the propagation of forest fires on a CPU and a GPU, and finding out when the execution on GPU is more efficient than on CPU. In this study, a fire simulator has been designed based on the basic model for one point evolution in the FARSITE simulator. As study case, a synthetic fire with an initial circular perimeter has been used; the wind, terrain and vegetation conditions have been maintained constant throughout the simulation. Results highlighted that GPUs allow obtaining more accurate results while reducing the execution time of the simulations.

Despite a rapid technological evolution in the field of technical assistance for people with motor disabilities, their ability to move independently in a wheelchair is still limited. New information and communication technologies (NICT) such as augmented reality (AR) are a real opportunity to integrate people with disabilities into their everyday life and work. AR can afford real-time information about buildings and locations’ accessibility through mobile applications that allow the user to have a clear view of the building details. By interacting with augmented environments that appear in the real world using a smart device, users with disabilities have more control of their environment. In this paper, we propose a decision support system using AR for motor disabled people navigation assistance. We describe a real-time wheelchair navigation system equipped with geological mapping that indicates access path to a desired location, the shortest route towards it and identifies obstacles to avoid. The prototyped wheelchair navigation system was developed for use within the University of Lille campus.

Due to the advent of quantum computers, quantum-safe cryptographic alternatives are required. Promising candidates are based on lattices. The hardness of the underlying problems must also be assessed on classical hardware. In this paper, we present the open source framework p3Enum for solving the important lattice problem of finding the shortest non-zero vector in a lattice, based on enumeration with extreme pruning. Our parallelized enumeration routine scales very well on SMP systems with an extremely high parallel efficiency up to 0.91 with 60 threads on a single node. A novel parameter $$\nu $$ within the pruning function increases the probability of success and the workload of the enumeration. This enables p3Enum to achieve runtimes for parallel enumerations which are comparable to single-threaded cases but with higher success rate. We compare the performance of p3Enum to publicly available libraries and results in the literature. For lattice dimensions 66 to 88, p3Enum performs the best which makes it a good building block in lattice reduction frameworks.

We introduce a method of calculating and rendering shapes in a non-Euclidean 2D space in real-time using hyperbolic and spherical trigonometry. We record the objects’ parameters in a polar coordinate system and use azimuthal equidistant projection to render the space onto the screen. We discuss the complexity of this method, renderings produced, limitations and possible applications of the created software as well as potential future developments.

Identifying academic homepages is a fundamental work of many tasks, such as expert finding, researcher profile extraction and homonym researcher disambiguation. Many works have been proposed to obtain researcher homepages using search engines. These methods only extract features at the lexical-level from each single retrieval result, which is not enough to identify homepage from retrieval results with high similarity. To address this problem, we first make deep-insight improvements on three aspects. (1) Fine-gained features are designed to efficiently detect whether the researcher’s name appears in retrieval results; (2) Establishing correlation of multiple retrieval results for the same researcher; (3) Obtaining semantic information involved in URL, title and snippet of each retrieval result by recurrent neural networks. Afterwards, we employ a joint neural network framework which is able to make comprehensive use of these informative information. In comparison with previous work, our approach gives a substantial increase of 10%–11% accuracy on a real-world dataset provided by AMiner. Experimental results demonstrate the effectiveness of our method.

Road traffic has become a main source of air pollution in urban areas. For this reason, governments are applying traffic regulations trying to fulfill the recommendations of Air Quality (AQ) standards in order to reduce the pollution level. In this paper, we present a novel proposal to improve AQ in cities by combining fuzzy logic and Complex Event Processing (CEP) technology. In particular, we propose a flexible fuzzy inference system to improve the decision-making process by recommending the actions to be carried out on each pollution scenario. This fuzzy inference system is fed with pollution data obtained by a CEP engine and weather forecast from domain experts.

Parametric linear programming is central in polyhedral computations and in certain control applications. We propose a task-based scheme for parallelizing it, with quasi-linear speedup over large problems.

The Analytic Hierarchy Process (AHP) method is widely used to deal with multi-criteria decision-making problems thanks to its simplicity and flexibility. However, it is often criticized for subjectivity and inconsistency in assigning the comparison weights that are based on expert judgments. In order to remedy these shortcomings, we propose in this paper an algorithm that automatically generates the pairwise comparison weights of alternatives according to each considered criterion. In addition, we demonstrate through an example that the judgment matrices constructed by the algorithm are very consistent.

Neural Networks (NNs) are frequently applied to Multi Input Multi Output (MIMO) problems, where the amount of data to manage is extremely high and, hence, the computational time required for the training process is too large. Therefore, MIMO problems are often split into Multi Input Single Output (MISO) problems; MISOs are further decomposed into several Single Input Single Output (SISO) problems. The aim of this paper is to present an optimized approach for NNs training based on properties of Singular Value Decomposition (SVD), allowing to decompose the MISO NN into a collection of SISO NNs. The decomposition provides a two-fold advantage: firstly, each SISO NN can be trained by using a one-dimensional function, namely a limited dataset, and then a parallel architecture can be implemented on a PC-cluster, decreasing the computational cost. The parallel algorithm performance are validated by using magnetic hysteresis dataset with the aim to prove the computational speed up by preserving the accuracy.

We propose a movie-based visualization for High Performance Computing (HPC) visualization. In this method, a viewer interactively explores a movie database with a specially designed application program called a movie data browser. The database is a collection of movie files that are tied with the spatial coordinates of their viewpoints. One can walk through the simulation’s data space by extracting a sequence of image files from the database with the browser. In this method, it is important to scatter as many viewpoints as possible for smooth display. Since proposing the movie-based visualization method, we have been developing some critical tools for it. In this paper, we report the latest development for supercomputers to apply many in-situ visualizations with different viewpoints in a Multiple Program Multiple Data (MPMD) framework. A key feature of this framework is to place a membrane-like layer between the simulation program and the visualization program. Hidden behind the membrane layer, the simulation program is not affected by the visualization program even if the number of scattered viewpoints is large.

The popularity of real-time on-demand transit as a fast evolving mobility service has paved the way to explore novel solutions for point-to-point transit requests. In addition, strict government regulations on greenhouse gas emission calls for energy efficient transit solutions. To this end, we propose an on-demand public transit system using a fleet of heterogeneous electric vehicles, which provides real-time service to passengers by linking a zone to a predetermined rapid transit node. Subsequently, we model the problem using a Genetic Algorithm, which generates routes and schedules in real-time while minimizing passenger travel time. Experiments performed using a real map show that the proposed algorithm not only generates near-optimal results but also advances the state-of-the-art at a marginal cost of computation time.

The output power of photovoltaic (PV) systems is heavily influenced by mismatching conditions that can drastically reduce the power produced by PV arrays. The mismatching power losses in PV systems are mainly related to partial or full shading conditions, i.e. non-uniform irradiation of the array. An essential point is the detection of the irradiance level in the whole PV plant. The use of irradiance sensors is generally avoided because of their cost and necessity for periodic calibration. In this work, an Artificial Neural Network (ANN) based method is proposed to forecast the irradiance value of each panel constituting the PV module, starting from a number of spatially distributed analytical irradiance computations on the array. A 2D random and cloudy 12 h irradiance profile is generated considering wind action; the results show that the implemented system is able to provide an accurate temporal prevision of the PV plant irradiance distribution during the day.

iPIC3D is a widely used massively parallel Particle-in-Cell code for the simulation of space plasmas. However, its current implementation does not support execution on multiple GPUs. In this paper, we describe the porting of iPIC3D particle mover to GPUs and the optimization steps to increase the performance and parallel scaling on multiple GPUs. We analyze the strong scaling of the mover on two GPU clusters and evaluate its performance and acceleration. The optimized GPU version which uses pinned memory and asynchronous data prefetching outperform their corresponding CPU versions by $$5-10\times $$ on two different systems equipped with NVIDIA K80 and V100 GPUs.

Lossless data compression is emerged to utilize in the BigData applications in the recent days. The conventional algorithms mainly generate a symbol lookup table to replace a frequent data pattern in the inputted data to a symbol, and then compresses the information. This kind of the dictionary-based compression mechanism potentially has an overhead problem regarding the number of symbol matchings in the table. This paper focuses on a novel method to reduce the number of searches in the table using a bank separation technique. This paper reports design and implementation of the bank select method on the LCT-DLT, and shows the performance evaluations to validate the effects of the method.

In this work a stabilized variational formulation is proposed to solve the interface problem describing the electric response of cells to an applied electric field. The proposed stabilized formulation is attractive since the discrete operator resulting from finite element discretization generates a definite linear system for which efficient iterative solvers can be applied. The interface problem describing the cell response is solved with a primal variational formulation and the proposed stabilized formulation. Both methods are compared in terms of the approximation properties of the primal and the Lagrange multiplier variable. The computational performance of the methods are also compared in terms of the mean number of iterations needed to solve one time step during the polarization process of an isolated square cell. Moreover, numerical experiments are performed to validate the convergence properties of the methods.

The data-driven models are able to study the model structure in cases when a priori information is not sufficient to build other types of models. The possible way to obtain physical interpretation is the data-driven differential equation discovery techniques. The existing methods of PDE (partial derivative equations) discovery are bound with the sparse regression. However, sparse regression is restricting the resulting model form, since the terms for PDE are defined before regression. The evolutionary approach, described in the article, has a symbolic regression as the background instead and thus has fewer restrictions on the PDE form. The evolutionary method of PDE discovery (EPDE) is tested on several canonical PDEs. The question of robustness is examined on a noised data example.

Disease prediction can be extremely helpful in saving people, especially when we are diagnosed with cancer. Cervical cancer, also known as uterine cancer, is the fourth most frequent cancer in women with an estimated 570,000 new cases in 2018 representing 6.6% of all female cancers. In accordance with World Health Organization (WHO), the mortality rate for cervical cancer reaches 90% in the underdeveloped nations and that the high mortality rate found in it could suffer a substantial reduction if there were: prevention, effective screening, treatment programs and early diagnosis. Artificial Neural Networks (ANN) has been helping to provide predictions in healthcare for several decades. Most research works utilize neural classifiers trained with backpropagation (BP) learning algorithm to achieve cancer diagnosis. the traditional BP algorithm has some significant disadvantages, such as training too slowly, easiness to fall into local minima, and sensitivity of the initial weights and bias. In this work, we use a type of Recurrent Neural Network (RNN), known as Long Short-Term Memory (LSTM), whose main characteristic is the ability to store information in a series of temporal data. Instead of training the network with the backpropagation, the LSTM network was trained using five different metaheuristic algorithms: Cuckoo Search (CS), Genetic Algorithm (GA), Gravitational Search (GS), Gray Wolf Optimizer (GWO) and Particle Swarm Optimization (PSO). From results obtained can be observed that metaheuristic algorithms had performances above 96%.

The identification of cohesive communities (dense subgraphs) is a typical task applied to the analysis of social and biological networks. Different definitions of communities have been adopted for particular occurrences. One of these, the 2-club (dense subgraphs with diameter value at most of length 2) has been revealed of interest for applications and theoretical studies. Unfortunately, the identification of 2-clubs is a computationally intractable problem, and the search of approximate solutions (at a reasonable time) is therefore fundamental in many practical areas. In this article, we present a genetic algorithm based heuristic to compute a collection of Top k 2-clubs, i.e., a set composed by the largest k 2-clubs which cover an input graph. In particular, we discuss some preliminary results for synthetic data obtained by sampling Erdös-Rényi random graphs.

Road passage time is an important measure of urban traffic. Accurate estimation of road passage time contributes to the route programming and the urban traffic planning. Currently, the estimation of road passage time for a particular road is usually based on its historical data which is simple to express the general law of road traffic. However, with the increase of the number of roads in the urban area, the connection between roads becomes more complex. The existing methods fail to make use of the connection between different roads and the road passage time, merely based on its own historical data. In this paper, we propose a road passage time estimating model, called “CA-RPT”, which utilizes the contextual information between road connections as well as the date and time period. We evaluate our method based on a real geolocation information data set collected by mobile APP anonymously. The results demonstrate that our method is more accurate than the state-of-the-art methods.

This article describes the results of multidisciplinary research in the areas of analysis and modeling of complex processes of treatment on the example of patients with cardiovascular diseases. The aim of this study is to develop tools and methods for the analysis of highly variable processes. In the course of the study, methods and algorithms for processing large volumes of various and semi-structured series data of medical information systems were developed. Moreover, the method for predicting treatment events has been developed. Treatment graph and algorithms of community detection and machine learning method are applied. The use of graphs and machine learning methods has expanded the capabilities of process mining for a better understanding of the complex process of medical care. Moreover, the algorithms for parallel computing using CUDA for graph calculation is developed. The improved methods and algorithms are considered in the corresponding developed visualization tool for complex treatment processes analysis.

Calcium signaling is vital for the contraction of the heart. Physiologically realistic simulation of this subcellular process requires nanometer resolutions and a complicated mathematical model of differential equations. Since the subcellular space is composed of several irregularly-shaped and intricately-connected physiological domains with distinct properties, one particular challenge is to correctly compute the diffusion-induced calcium fluxes between the physiological domains. The common approach is to pre-calculate the effective diffusion coefficients between all pairs of neighboring computational voxels, and store them in large arrays. Such a strategy avoids complicated if-tests when looping through the computational mesh, but suffers from substantial memory overhead. In this paper, we adopt a memory-efficient strategy that uses a small lookup table of diffusion coefficients. The memory footprint and traffic are both drastically reduced, while also avoiding the if-tests. However, the new strategy induces more instructions on the processor level. To offset this potential performance pitfall, we use AVX-512 intrinsics to effectively vectorize the code. Performance measurements on a Knights Landing processor and a quad-socket Skylake server show a clear performance advantage of the manually vectorized implementation that uses lookup tables, over the counterpart using coefficient arrays.

The growth of the activities of the Petroleum Industry in the Brazilian Equatorial Margin, reinforces the need for the environmental knowledge of the region, which will be potentially exposed to risks related to such activities. The environmental importance of this region evidences the need to deepen and systematize not only the knowledge about the environmental sensitivity of the region, but also about the characteristics that will exert influence over it. The Costa Norte Project can be identified with one of these initiatives. The project has as one of the main objectives to evaluate the efficiency of the use of marine hydrodynamic environmental computational modeling methods to represent the marine dynamics over that region. In this paper a regional ocean computational model was used to produce an inedited ten year hindcast simulation in order to represent the main aspects associated with mesoscale climatological ocean circulation at the Brazilian equatorial margin. This article aims to present the methodology and the results analysis and evaluation associated to the cited hydrodynamic computational simulation. The obtained results clearly demonstrated the ocean model potential to represent the most important ocean variables space and time distribution over the studied region. Comparative analysis with observed data demonstrated good agreement with temperature, salinity and sea surface height fields generated by the implemented model. The Costa Norte Project is carrying out under the Brazilian National Petroleum Agency (ANP) R&D levy as “Investment Commitment to Research and Development” and is financially supported by Enauta O&G company.

The stationary neutron transport equation describes the neutron population and thus, the generated heat, inside a nuclear reactor core. Obtaining the solution of this equation requires to solve a generalized eigenvalue problem efficiently. The majority of the eigenvalue solvers use the factorization of the system matrices to construct preconditioners, such as the ILU decomposition or the ICC decomposition, to speed up the convergence of the methods. The storage of the involved matrices and incomplete factorization demands high quantities of computational memory although a the sparse format is used. This makes the computational memory the limiting factor for this kind of calculations in some personal computers. In this work, we propose a matrix-free preconditioned eigenvalue solver that does not need to have the matrices allocated in memory explicitly. This method is based on the block inverse-free preconditioned Arnoldi method (BIFPAM) with the innovation that uses a preconditioner that is applied from matrix-vector operations. As well as reducing enormously the computational memory, this methodology removes the time to assembly the sparse matrices involved in the system. A two-dimensional and three-dimensional benchmarks are used to study the performance of the methodology proposed.

We derive numerical series representations for option prices on interest rate index for affine jump-diffusion models in a stochastic jump intensity framework with an adaptation of the Fourier-cosine series expansions method, focusing on the European vanilla derivatives. We give the price for nine different Ornstein-Uhlenbeck models enhanced with different jump size distributions. The option prices are accurately and efficiently approximated by solving the corresponding set ordinary differential equations and parsimoniously truncating the Fourier series.

This paper presents new optimization approaches aiming at reducing the impact of memory accesses on the performance of dataflow programs. The approach is based on introducing a high level management of composite data types in dynamic dataflow programming language for the memory processing of data tokens. It does not require essential changes to the model of computation (MOC) or to the dataflow program itself. The objective of the approach is to remove the unnecessary constraints of memory isolations without introducing limitations to the scalability and composability properties of the dataflow paradigm. Thus the identified optimizations allow to keep the same design and programming philosophy of dataflow, whereas aiming at improving the performance of the specific configuration implementation. The different optimizations can be integrated into the current RVC-CAL design flows and synthesis tools and can be applied to different sub-networks partitions of the dataflow program. The paper introduces the context, the definition of the optimization problem and describes how it can be applied to dataflow designs. Some examples of the optimizations are provided.

We investigate, through data sets correlation analysis, how relevant to the simulation of refugee dynamics the food situation is. Armed conflicts often imply difficult food access conditions for the population, which can have a great impact on the behaviour of the refugees, as is the case in South Sudan. To test our approach, we adopt the Flee agent-based simulation code, combining it with a data-driven food security model to enhance the rule set for determining refugee movements. We test two different approaches for South Sudan and find promising yet negative results. While our first approach to modelling refugees response to food insecurity do not improve the error of the simulation development approach, we show that this behaviour is highly non-trivial and properly understanding it could determine the development of reliable models of refugee dynamics.

Three dimensional (3D) virtual representations of the internal structure of textiles are of interest for a variety of purposes related to fashion, industry, education or other areas. The modeling of ancient weaving techniques is relevant to understand and preserve our heritage, both tangible and intangible. However, ancient techniques cannot be reproduced with standard approaches, which usually are aligned with the characteristics of modern, mechanical looms. The aim of this paper is to propose a mathematical modelling of ancient weaving techniques by means of matrices in order to be easily mapped to a virtual 3D representation. The work focuses on ancient silk textiles, ranging from the 15th to the 19th centuries. We also propose a computer vision-based strategy to extract relevant information from digital imagery, by considering different types of images (textiles, technical drawings and macro images). The work here presented has been carried out in the scope of the SILKNOW project, which has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No. 769504.

In the paper, we study one of the fundamental problems of grammatical inference, namely the induction of nondeterministic finite automata (NFA). We consider the induction of NFA consistent with the given sets of examples and counterexamples. We transform the induction problem into a constraint satisfaction problem and propose two variable ordering methods to solve it. We evaluate experimentally the proposed variable ordering methods and compare them with a state-of-the-art method. Additionally, through the experiments, we assess the impact of sample sets sizes on the performance of the induction algorithm using the respective variable ordering methods.

The last few years marked a substantial development in the domain of Deep Reinforcement Learning. However, a crucial and not yet fully achieved objective is to devise intelligent agents which can be successfully taken out of the laboratory and employed in the real world. Intelligent agents that are successfully deployable in true physical settings, require substantial prior exposure to their intended environments. When this is not practical or possible, the agents benefit from being trained and tested on powerful test-beds, effectively replicating the real world. To achieve traffic management at an unprecedented level of efficiency, in this paper, we introduce a significantly richer new traffic simulation environment; Traffic3D. Traffic3D is a unique platform built to effectively simulate and evaluate a variety of 3D-road traffic scenarios, closely mimicking real-world traffic characteristics including faithful simulation of individual vehicle behavior, precise physics of movement and photo-realism. We discuss the merits of Traffic3D in comparison to state-of-the-art traffic-based simulators. Along with deep reinforcement learning, Traffic3D facilitates research across various domains such as object detection and segmentation, unsupervised representation learning, visual question answering, procedural generation, imitation learning and learning by interaction.

Rapid growth in Big Data and Cloud technologies has fueled rising energy demands in large server systems such as data centers, leading to a need for effective power management. In this paper, we investigate the energy consumption characteristics of data-intensive distributed applications in terms of the CPU and memory subsystem. To this end, we develop PowerSave as a lightweight software framework that enables dynamic reconfiguration of power limits. PowerSave uses Running Average Power Limit (RAPL) to impose power limits. Our evaluation study, conducted on three different real systems, demonstrates that for workloads typical of servers used in data centers, higher power caps correlate with higher overall CPU energy use.

In this paper, we present AcausalNets.jl - a library supporting inference in a quantum generalization of Bayesian networks and their application to quantum games. The proposed solution is based on modern approach to numerical computing provided by Julia language. The library provides a high-level functions for Bayesian inference that can be applied to both classical and quantum Bayesian networks.

Motif discovery and analysis in time series data-sets have a wide-range of applications from genomics to finance. In consequence, development and critical evaluation of these algorithms is required with the focus not just detection but rather evaluation and interpretation of overall significance. Our focus here is the specific algorithm, VALMOD, but algorithms in wide use for motif discovery are summarised and briefly compared, as well as typical evaluation methods with strengths. Additionally, Taxonomy diagrams for motif discovery and evaluation techniques are constructed to illustrate the relationship between different approaches as well as inter-dependencies. Finally evaluation measures based upon results obtained from VALMOD analysis of a GBP-USD foreign exchange (F/X) rate data-set are presented, in illustration.

In this study we apply clustering techniques for analyzing and understanding households’ electricity consumption data. The knowledge extracted by this analysis is used to create a model of normal electricity consumption behavior for each particular household. Initially, the household’s electricity consumption data are partitioned into a number of clusters with similar daily electricity consumption profiles. The centroids of the generated clusters can be considered as representative signatures of a household’s electricity consumption behavior. The proposed approach is evaluated by conducting a number of experiments on electricity consumption data of ten selected households. The obtained results show that the proposed approach is suitable for data organizing and understanding, and can be applied for modeling electricity consumption behavior on a household level.

Numerous studies have been conducted over the past few decades on human behavior modeling and simulation by incorporating the dynamic behaviors of people for different Facility Management (FM) applications. For example; the Drivers, Needs, Actions and Systems (DNAS) framework which provides a standardized way to conceptually represent energy-related occupant behaviors in buildings and allows the exchange of occupant behavior information and integration with building simulation tools. Despite numerous studies dealing with dynamic interactions of the building occupants, there is still a gap exists in the knowledge modeling of occupant behaviors for dynamic building environments. Such environments are best observed on construction sites where the contextual information linked to the building spaces evolve often over time in terms of their location, size, properties and relationships with the site environment. The evolving contextual information of a building is required to be mapped with the occupant interactions for an improved understanding of their changing behaviors. To fill this research gap, a framework is designed for providing a ‘blueprint map’ to integrate DNAS framework with our Semantic Trajectories in Dynamic Environments (STriDE) data model to incorporate the dynamicity of building environments. The proposed framework extends the usability of a DNAS framework by providing a centralized knowledge base that holds the mobility data of occupants with relevant historicized contextual information of the building environment to study occupant behaviors for different FM applications.

Coastal forecast systems provide coastal managers with accurate and timely hydrodynamic predictions, supporting multiple uses such as navigation, water monitoring, port operations and dredging activities. They are also useful to support recreational activities. Still, the widespread use of coastal forecasts is limited by the unavailability of open forecasts for consultation, the expertise needed to build operational forecast systems and the human and computational resources required to maintain them in operation every day. A new service for the generic deployment of forecast systems at user-specified locations was developed to address these limitations. Denoted OPENCoastS, this service builds circulation forecast systems for user-selected coastal areas and maintains them in operation using the European Open Science Cloud (EOSC) computational resources. OPENCoastS can be applied to any coastal region and has been in operation since 2018, forced by several regional and global forecasts of the atmospheric and ocean dynamics. It has attracted over 150 users from around 45 institutions across the globe. However, most users come from research institutions. The only requirement needed to use this service – a computational grid of the domain of interest – has proven difficult to obtain by most coastal managers. Herein, a new way to bring coastal managers and the general public to the OPENCoastS community is proposed. By creating an open, scalable and organized repository of computational grids, shared by expert coastal modelers across the globe, the benefits from the use of OPENCoastS can now be extended to all coastal actors.