Computing and Informatics (E-Journal - Institute of Informatics, SAS, Bratislava)
Not a member yet
1506 research outputs found
Sort by
Experiment on Methods for Clustering and Categorization of Polish Text
Get PDFThe main goal of this work was to experimentally verify the methods for a challenging task of categorization and clustering Polish text. Supervised and unsupervised learning was employed respectively for the categorization and clustering. A profound examination of the employed methods was done for the custom-built corpus of Polish texts. The corpus was assembled by the authors from Internet resources. The corpus data was acquired from the news portal and, therefore, it was sorted by type by journalists according to their specialization. The presented algorithms employ Vector Space Model (VSM) and TF-IDF (Term Frequency-Inverse Document Frequency) weighing scheme. Series of experiments were conducted that revealed certain properties of algorithms and their accuracy. The accuracy of algorithms was elaborated regarding their ability to match human arrangement of the documents by the topic. For both the categorization and clustering, the authors used F-measure to assess the quality of allocation
Blog Style Classification: Refining Affective Blogs
Get PDFIn the constantly growing blogosphere with no restrictions on form or topic, a number of writing styles and genres have emerged. Recognition and classification of these styles has become significant for information processing with an aim to improve blog search or sentiment mining. One of the main issues in this field is detection of informative and affective articles. However, such differentiation does not suffice today. In this paper we extend the differentiation and suggest a fine-grained set of subcategories for affective articles. We propose and evaluate a classification method employing novel lexical, morphological, lightweight syntactic and structural features of written text. The results show that our method outperforms the existing approaches
Dynamic Monitoring in PANGEA Platform Using Event-Tracing Mechanisms
Get PDFThe use of distributed multi-agent systems (MAS) have increased in recent years, with the growing potential to handle large volumes of data and coordinate the operations of many organizations. In these systems, each agent independently handles a set of specialized tasks and cooperates to achieve the goals of the system and a high degree of flexibility. Multi-agent systems have become the most effective and widely used form of developing this type of application in which communication among various devices must be both reliable and efficient. One of the problems related to distribute computing is message passing, which is related to the interaction and coordination among intelligent agents. Consequently, a multi-agent architecture must necessarily provide a robust communication platform and control mechanisms. This paper presents the integration of an event-tracing model in an agent platform called PANGEA. Adding this new capability, the platform allows improving the monitoring and analysis of the information that agents can send/receive in order to fulfil their goals more efficiently
Using AVX2 Instruction Set to Increase Performance of High Performance Computing Code
Get PDFIn this paper we discuss new Intel instruction extensions - Intel Advance Vector Extensions 2 (AVX2) and what these bring to high performance computing (HPC). To illustrate this new systems utilizing AVX2 are evaluated to demonstrate how to effectively exploit AVX2 for HPC types of the code and expose the situation when AVX2 might not be the most effective way to increase performance
Parallel Computation of Smoke Movement During a Car Park Fire
Get PDFIn this paper the use of Fire Dynamics Simulator (FDS) for parallel computer simulation of the smoke movement during a fire of two passenger cars in an underground car park is investigated. The simulations were executed on a high-performance computer cluster. A specific problem of FDS parallel computation using Message-Passing Interface (MPI) is a separate solution of governing equations on computational subdomains causing a loss of accuracy. Therefore, the impact of parallelisation on simulation accuracy in the case of using a greater number of computational cores of the computer cluster is studied with the aim to increase the computational performance and enable practical application of such simulations for fire safety measures. The geometrical model and material properties of the cars used in the simulation have been verified by a full-scale fire experiment in open air. We describe the results of a series of simulations of several fire scenarios with different numbers of parked cars and ventilation configurations and determine times and locations at which conditions in the car park become untenable for human life. The simulation indicates that proper ventilation prolongs tenable conditions by several minutes
On-Line Load Balancing with Task Buffer
Get PDFOn-line load balancing is one of the most important problems for applications with resource allocation. It aims to assign tasks to suitable machines and balance the load among all of the machines, where the tasks need to be assigned to a machine upon arrival. In practice, tasks are not always required to be assigned to machines immediately. In this paper, we propose a novel on-line load balancing model with task buffer, where the buffer can temporarily store tasks as many as possible. Three algorithms, namely LPTCP1_α, LPTCP2_α, and LPTCP3_β, are proposed based on the Longest Processing Time (LPT) algorithm and a variety of planarization algorithms. The planarization algorithms are proposed for reducing the difference among each element in a set. Experimental results show that our proposed algorithms can effectively solve the on-line load balancing problem and have good performance in large scale experiments
Estimation-Based Queue Scheduling Model to Improve QoS for End Users in MANETs
Get PDFUsing MANETs for real time applications is always a challenge as the network is extremely dynamic with brisk topology changes. Despite this, several real time schedulers have been developed that aimed at providing QoS to ad hoc nodes. The quality of service (QoS) is standardized in terms of capacity, reliability, link quality, delays/jitters, and network cost. Thus, for QoS, the better transmission should be maintained at end user as well as at the transmitting unit. QoS of a network is affected by delays and bandwidth allocated for transmission. For an efficient network, it is required to predict these metrics during transmission. For this, in this paper, integration of quaternion-based Kalman filter is performed that predicts the required bandwidth and the network delays with higher accuracy. From the analysis, it is shown that bandwidth can be optimized but it is not possible to aloof delays in the network. Thus, while implementing such admission control procedures, estimation process allows control over delays and sustain them from going beyond a certain threshold value. The model proposed is a novel approach and has not been formulated in any of previous work related to QoS in MANETs. The effectiveness of model is demonstrated using both simulation and real time results
Efficient Two-Level Swarm Intelligence Approach for Multiple Sequence Alignment
Get PDFThis paper proposes two-level particle swarm optimization (TL-PSO), an efficient PSO variant that addresses two levels of optimization problem. Level one works on optimizing dimension for entire swarm, whereas level two works for optimizing each particle's position. The issue addressed here is one of the most challenging multiple sequence alignment (MSA) problem. TL-PSO deals with the arduous task of determination of exact sequence length with most suitable gap positions in MSA. The two levels considered here are: to obtain optimal sequence length in level one and to attain optimum gap positions for maximal alignment score in level two. The performance of TL-PSO has been assessed through a comparative study with two kinds of benchmark dataset of DNA and RNA. The efficiency of the proposed approach is evaluated with four popular scoring schemes at specific parameters. TL-PSO alignments are compared with four PSO variants, i.e. S-PSO, M-PSO, ED-MPSO and CPSO-Sk, and two leading alignment software, i.e. ClustalW and T-Coffee, at different alignment scores. Hence obtained results prove the competence of TL-PSO at accuracy aspects and conclude better score scheme
A Comprehensive Learning-Based Model for Power Load Forecasting in Smart Grid
Get PDFIn the big data era, learning-based techniques have attracted more and more attention in many industry areas such as smart grid, intelligent transportation. The power load forecasting is one of the most critical issues in data analysis of smart grid. However, learning-based methods have not been widely used due to the poor data quality and computational capacity. In this paper, we propose a comprehensive learning-based model to forecast heavy and over load (HOL) accidents according to the data from various information systems. At first, we present a combined random under- and over-sampling technique for imbalanced electric data, and choose an optimal sampling rate through several experiments. Then, we reduce the attributes that have significant impact on the power load by using learning-based methods. Finally, we provide an algorithm based on the random forest method to prevent the over-fitting problem. We evaluate the proposed model and algorithms with the real-world data provided by China Grid. The experimental results show that our model works efficiently and achieves low error rates
A New Approach of Dynamic Clustering Based on Particle Swarm Optimization and Application in Image Segmentation
Get PDFThis paper presents a new approach of dynamic clustering based on improved Particle Swarm Optimization (PSO) and which is applied to image segmentation (called DCPSONS). Firstly, the original PSO algorithm is improved by using diversity mechanism and neighborhood search strategy. The improved PSO is then combined with the well-known data clustering k-means algorithm for dynamic clustering problem where the number of clusters has not yet been known. Finally, DCPSONS is applied to image segmentation problem, in which the number of clusters is automatically determined. Experimental results in using sixteen benchmark data sets and several images of synthetic and natural benchmark data demonstrate that the proposed DCPSONS algorithm substantially outperforms other competitive algorithms in terms of accuracy and convergence rate