Computing and Informatics (E-Journal - Institute of Informatics, SAS, Bratislava)
Not a member yet
1506 research outputs found
Sort by
Efficient Keyword Extraction and Text Summarization for Reading Articles on Smart Phone
Get PDFThese days, we can connect to the internet from almost anywhere, allowing us to access web content, including newspapers, magazines, blogs and websites, using mobile devices such as a smart phone. However, people sometimes struggle to read and use the contents due to the nature of these devices such as a small display, low display resolution and limited computing resources (low CPU speed and little memory). This paper aims to develop a convenient interface that provides keyword extraction, summary generation and search engine to users. We apply the proposed summarization method to Korean and English news articles and evaluate it using several experiments on single and multiple news article test collections and user-receptiveness tests. Since the proposed method shows a good performance on these experiments and tests, we think that this interface can help users more efficiently to read the news articles on various mobile devices
Oversampling Method for Imbalanced Classification
Get PDFClassification problem for imbalanced datasets is pervasive in a lot of data mining domains. Imbalanced classification has been a hot topic in the academic community. From data level to algorithm level, a lot of solutions have been proposed to tackle the problems resulted from imbalanced datasets. SMOTE is the most popular data-level method and a lot of derivations based on it are developed to alleviate the problem of class imbalance. Our investigation indicates that there are severe flaws in SMOTE. We propose a new oversampling method SNOCC that can compensate the defects of SMOTE. In SNOCC, we increase the number of seed samples and that renders the new samples not confine in the line segment between two seed samples in SMOTE. We employ a novel algorithm to find the nearest neighbors of samples, which is different to the previous ones. These two improvements make the new samples created by SNOCC naturally reproduce the distribution of original seed samples. Our experiment results show that SNOCC outperform SMOTE and CBSO (a SMOTE-based method)
Statistical Based Slicing Method for Prioritizing Program Fault Relevant Statements
Get PDFThe aim of this paper is to integrate the strong points of statistical debugging and program slicing techniques for efficient program fault localization. The dynamic slices could be inappropriately large including redundant information considering that the statements are not ranked according to their fault relevance in the computed slice. On the other hand, the conventional statistical debugging techniques do not consider the dependence relationships between the faulty code (i.e., the cause) and the erroneous output (i.e., the effect). This information is very useful during the debugging. In this paper, a new method Stat-Slice for locating the latent bugs in programs is presented which could find a wider range of bugs, e.g. the omitted code faults, the header file faults, and etc., comparing with other techniques. Unlike the traditional program slicing techniques, the proposed method computes the backward dynamic slices of several failing and passing runs. Using K-means clustering in addition to a new ranking and pruning technique, we prioritize statements according to their likelihood to be the cause for failure. Our experiments on Siemens, grep, gzip, and flex test suites manifest that ranking statements according to their suspiciousness has considerably reduced the effort for fault localization
Timed and Hybrid Petri Nets at Solving Problems of Computational Intelligence
Get PDFTimed Petri nets (TPN) and hybrid Petri nets (HPN), more precisely first-order HPN (FOHPN), are used here in order to model, analyse and control discrete-event systems (DES) and hybrid systems (HS) of different kinds consisting of cooperating subsystems (modules, agents). Foremost, principles of the particular kinds of Petri nets (PN), including the primary place/transition PN (P/T PN), are explained. Then, two methods of the supervision are introduced. Finally, the corresponding kinds of PN are applied in four case studies in order to solve the problems of modelling, analysing and control. Namely, the throughput of transport systems (rail and road), flexible manufacturing systems (FMS) and the workflow (WF) of the evacuation process from an endangered area (EA) are handled
Quantifying Productivity of Individual Software Programmers: Practical Approach
Get PDFSoftware measurement is a crucial part of a good software engineering. Software developers quantify the software to see if the use cases are complete, if the analysis model is consistent with requirements and if the code is ready to be tested. Software project managers assess the software process and the software product to determine if it is going to be finished on time and within budget. Customers evaluate the final product if it meets their needs. Overall, the main purpose of software engineering is to make software systems controllable and foreseeable, activities with a solid method rather than intuitional, complicated or unprincipled. Software measurement studies are about quantifying the software engineering entities and attributes, both of which aim to support software development efforts and quality improvement. In this paper, we quantify a set of relationships using the current size, defect and object-oriented software metrics practically and pragmatically. Our paper proposes a method to measure the productivity of individual software programmers. Furthermore, this method provides a common opinion for understanding, controlling and improving the software engineering practices
Effect of Data Layout in the Evaluation Time of Non-Separable Functions on GPU
Get PDFGPUs are able to provide a tremendous computational power, but their optimal usage requires the optimization of memory access. The many threads available can mitigate the long memory access latencies, but this usually demands a reorganization of the data and algorithm to reach the performance peak. The addressed problem is to know which data layout produces a faster evaluation when dealing with population-based evolutionary algorithms optimizing non-separable functions. This knowledge will allow a more efficient design of evolutionary algorithms. Depending on the fitness function and the problem size, the most suitable layout can be implemented at the design phase of the algorithm, avoiding later costly code or data layout redesigns. In this paper, diverse non-separable functions, such as Rosenbrock and Rana functions, and data layouts are evaluated. The implemented layouts cover main techniques to maximize the performance: coalesced access to global memory, intensive use of on-chip memory: shared memory and registers, and variable reuse to minimize the global memory transactions. Conclusions about the optimum data layout related to the characteristics of the fitness function and the problem size are stated. Besides, the conclusions ease the decision-making process for future implementations of other non-separable functions
Network Proactive Defense Model Based on Immune Danger Theory
Get PDFRecent investigations into proactive network defense have not produced a systematic methodology and structure; in addition, issues including multi-source information fusion and attacking behavior analysis have not been resolved. Borrowing ideas of danger sensing and immune response from danger theory, a proactive network defense model based on danger theory is proposed. This paper defines the signals and antigens in the network environment as well as attacking behavior analysis algorithm, providing evidence for future proactive defense strategy selection. The results of preliminary simulations demonstrate that this model can sense the onset of varied network attacks and corresponding endangered intensities, which help to understand the attack methods of hackers and assess the security situation of the current network, thus a better proactive defense strategy can be deployed. Moreover, this model possesses good robustness and accuracy
Evaluation of Database Replication Techniques for Cloud Systems
Get PDFCloud computing is becoming one of the preferred paradigms to deploy highly available and scalable systems. These systems usually demand the management of huge amounts of data, which cannot be solved with traditional database systems. Traditional replication protocols are not scalable enough for a cloud environment. This paper evolves different static replication techniques to achieve transactional support providing high availability and scalability as needed in cloud systems. This proposal offers different consistency levels according to the demands of client applications using a replication strategy based on a combination of traditional replication techniques with asynchronous epidemic updates. We have run several simulations that show this is an interesting approach to provide transactional support to clients with different consistency guaranties while leveraging the resources used
Improving Accuracy of Virtual Machine Power Model by Relative-PMC Based Heuristic Scheduling
Get PDFConventional utilization-based power model is effective for measuring the power consumption of physical machines. However, in virtualized environments its accuracy cannot be guaranteed because of the recursive resource accessing among multiple virtual machines. In this paper, we present a novel virtual machine scheduling algorithm, which uses Performance-Monitor-Counter as heuristic information to compensate the recursive power consumption. Theoretical analysis indicates that the error of virtual machine power model can be quantitative bounded when using the proposed scheduling algorithm. Extensive experiments based on standard benchmarks show that the error of virtual machine power measurements can be significantly reduced comparing with the classic credit-based scheduling algorithm
Automated Implementation Process of Machine Translation System for Related Languages
Get PDFThe paper presents an attempt to automate all data creation processes of a rule-based shallow-transfer machine translation system. The presented methods were tested on four fully functional translation systems covering language pairs: Slovenian paired with Serbian, Czech, English and Estonian language. An extensive range of evaluation tests was performed to assess the applicability of the methods