Computing and Informatics (E-Journal - Institute of Informatics, SAS, Bratislava)
Not a member yet
1506 research outputs found
Sort by
Time-Sensitive Adaptive Model for Adult Image Classification
Get PDFImages play an important role in modern internet communications, but not all of the images shared by the users are appropriate, and it is necessary to check and reject the inappropriate ones. Deep neural networks do this task perfectly, but it may not be necessary to use maximum power for all images. Many easier-to-identify images may be classified at a lower cost than running the full model. Also, the pressure on the system varies from time to time, so an algorithm that can produce the best possible results for different budgets is very useful. For this purpose, a deep convolutional neural network with the ability to generate several outputs from its various layers has been designed. Each output can be considered as a classifier with its own cost and accuracy. A selector is then used to select and combine the results of these outputs to produce the best possible result in the specified time budget. The selector uses a reinforcement learning model, which, despite the time-consuming learning phase, is fast at execution time. Our experiments on challenging social media images dataset show that the proposed model can reduce the processing time by 32 % by sacrificing only 1.4 % of accuracy compared to the VGG-f network. Also, using different metrics such as F1-score and AUC (the Area Under the Curve in the accuracy vs. time budget chart), the superiority of the proposed model at different time budgets over the base model is shown
Cookery: A Framework for Creating Data Processing Pipeline Using Online Services
Get PDFWith the increasing amount of data the importance of data analysis in various scientific domains has grown. A large amount of the scientific data has shifted to cloud based storage. The cloud offers storage and computation power. The Cookery framework is a tool developed to build scientific applications using cloud services. In this paper we present the Cookery systems and how they can be used to authenticate and use standard online third party services to easily create data analytic pipelines. Cookery framework is not limited to work with standard web services; it can also integrate and work with the emerging AWS Lambda which is part of a new computing paradigm, collectively, known as serverless computing. The combination of AWS Lambda and Cookery, which makes it possible for users in many scientific domains, who do not have any program experience, to create data processing pipelines using cloud services in a short time
Towards Exascale Computing Architecture and Its Prototype: Services and Infrastructure
Get PDFThis paper presents the design and implementation of a scalable compute platform for processing large data sets in the scope of the EU H2020 project PROCESS. We are presenting requirements of the platform, related works, infrastructure with focus on the compute components and finally results of our work
Improved Deep Forest Mode for Detection of Fraudulent Online Transaction
Get PDFAs the rapid development of online transactions, transaction frauds have also emerged seriously. The fraud strategies are characterized by specialization, industrialization, concealment and scenes. Anti-fraud technologies face many challenges under the trend of new situations. In this paper, aiming at sample imbalance and strong concealment of online transactions, we enhance the original deep forest framework to propose a deep forest-based online transaction fraud detection model. Based on the BaggingBalance method we propose, we establish a global sample imbalance processing mechanism to deal with the problem of sample imbalance. In addition, the autoencoder model is introduced into the detection model to enhance the representation learning ability. Via the three-month real online transactions data of a China's bank, the experimental results show that, evaluating by the metric of precision and recall rate, the proposed model has a beyond 10 % improvement compared to the random forest model, and a beyond 5 % improvement compared to the original deep forest model
New Simulation Software Technologies at the LHCb Experiment at CERN
Get PDFThe LHCb experiment at the Large Hadron Collider (LHC) at CERN has successfully performed a large number of physics measurements during Runs 1 and 2 of the LHC. Monte Carlo simulation is the key to the interpretation of these and future measurements. The LHCb experiment is currently undergoing a major detector upgrade for Run 3 of the LHC to process events with five times higher luminosity. New simulation software technologies have to be introduced to produce simulated data samples of sufficient size within the computing resources allocated for the next few years. Therefore, the LHCb collaboration is currently preparing an upgraded version of its Gauss simulation framework. The new version provides the LHCb specific functionality while its generic simulation infrastructure has been encapsulated in an experiment independent framework, Gaussino. The latter combines the Gaudi core software framework and the Geant4 simulation toolkit and fully exploits their multi-threading capabilities. A prototype of a fast simulation interface to the simulation toolkit is being developed as the latest addition to Gaussino to provide an extensive palette of fast simulation models, including new deep learning-based options
Lazy Repairing Backtracking for Dynamic Constraint Satisfaction Problems
Get PDFExtended Partial Dynamic Backtracking (EPDB) is a repair algorithm based on PDB. It deals with Dynamic CSPs based on ordering heuristics and retroactive data structures, safety conditions, and nogoods which are saved during the search process. In this paper, we show that the drawback of both EPDB and PDB is the exhaustive verification of orders, saved in safety conditions and nogoods, between variables. This verification affects remarkably search time, especially since orders are often indirectly deduced. Therefore, we propose a new approach for dynamically changing environments, the Lazy Repairing Backtracking (LRB), which is a fast version of EPDB insofar as it deduces orders directly through the used ordering heuristic. We evaluate LRB on various kinds of problems, and compare it, on the one hand, with EPDB to show its effectiveness compared to this approach, and, on the other hand, with MAC-2001 in order to conclude, from what perturbation rate resolving a DCSP with an efficient approach can be more advantageous than repair
Formal Verification of UML MARTE Specifications Based on a True Concurrency Real Time Model
Get PDFThe profile UML MARTE offers a general modeling framework for designing and analyzing real-time and embedded systems. Temporal aspects are critical criteria that should be taken into account during the design process. So, formal methods may be used to ensure the functional correctness of such systems. For this purpose, this paper defines an operational method for translating UML sequence diagrams annotated with MARTE stereotypes to time Petri nets with action duration specifications. The semantics of these specifications are defined in terms of duration action timed automata. This allows formal verification by means of several model checker tools like UPPAAL
Modeling Public Transport Network System by Using Statistics, Network Theory and Ant Colony Optimization
Get PDFIn some countries, bicycles are often used to access public transit stations, but the proportion of out-of-the-way travel is much smaller due to the limited availability of bicycles. Public bicycles are innovative rental or free bicycle schemes in urban areas that can be used for day-to-day mobility as one-way use is possible and can be considered as part of a public transport system. Different from traditional, mostly leisure bike rental services, they provide fast and easy access and have a variety of organizational layout, business models and useful technology for smart bikes (rented via smart cards or mobile phones). We find that bicycle-sharing systems that complement the traditional public transport system could potentially increase the competitiveness and attractiveness of sustainable modes of urban transport and thus help cities to promote sustainable daily mobility. Finally, we emphasize that the availability of open sources of urban transport information, such as public transport in our case, is crucial for analyzing urban mobility patterns. The aim of the research is to analyze and model PPP bicycle rentals using mathematical and computer methods. The article presents the application of the statistical and topological properties of bicycle rental and return network theory in city Novo mesto. The article uses swarm intelligence, a colony of ants to optimize the development of wheels across 14 stations. The wider city Novo mesto region with a population of almost 30 000 people, as a key industrial center, is heavily dependent on urban transport
Verifying Computation Tree Logic of Knowledge via Knowledge-Oriented Petri Nets and Ordered Binary Decision Diagrams
Get PDFComputation Tree Logic of Knowledge (CTLK) can specify many requirements of privacy and security of multi-agent systems (MAS). In our previous papers, we defined Knowledge-oriented Petri Net (KPN) to model MAS, proposed similar reachability graph to verify CTLK, gave their model checking algorithms and developed a related tool. In this paper, we use the technique of Ordered Binary Decision Diagrams (OBDD) to encode similar reachability graph in order to alleviate the state explosion problem, and verify more epistemic operators of CTLK. We design the corresponding symbolic model checking algorithms and improve our tool. We compare our model and method with MCMAS that is the state-of-the-art CTLK model checker, and experiments illustrate the advantages of our model and method. We also explain the reasons why our model and method can obtain better performances