Computer Science Journal (AGH University of Science and Technology, Krakow)
Not a member yet
476 research outputs found
Sort by
Application of linguistic cues in the analysis of language of hate groups
Get PDFHate speech and fringe ideologies are social phenomena that thrive on-line. Members of the political and religious fringe are able to propagate their ideas via the Internet with less effort than in traditional media. In this article, we attempt to use linguistic cues such as the occurrence of certain parts of speech in order to distinguish the language of fringe groups from strictly informative sources. The aim of this research is to provide a preliminary model for identifying deceptive materials online. Examples of these would include aggressive marketing and hate speech. For the sake of this paper, we aim to focus on the political aspect. Our research has shown that information about sentence length and the occurrence of adjectives and adverbs can provide information for the identification of differences between the language of fringe political groups and mainstream media
Pre-trained Deep Neural Network using Sparse Autoencoders and Scattering Wavelet Transform for Musical Genre Recognition
Get PDFResearch described in this paper tries to combine the approach of Deep Neural Networks (DNN) with the novel audio features extracted using the Scattering Wavelet Transform (SWT) for classifying musical genres. The SWT uses a sequence of Wavelet Transforms to compute the modulation spectrum coefficients of multiple orders, which has already shown to be promising for this task. The DNN in this work uses pre-trained layers using Sparse Autoencoders (SAE). Data obtained from the Creative Commons website jamendo.com is used to boost the well-known GTZAN database, which is a standard benchmark for this task. The final classifier is tested using a 10-fold cross validation to achieve results similar to other state-of-the-art approaches
Overview and Evaluation of Conceptual Strategies for Accessing CPU-Dependent Execution Resources in Grid Infrastructures
Get PDFThe emergence of many-core and massively-parallel computational accelerators (e.g., GPGPUs) has led to user demand for such resources in grid infrastructures. A widely adopted approach for discovering and accessing such resources has, however, yet to emerge. GPGPUs are an example of a larger class of computational resources, characterized in part by dependence on an allocated CPU. This paper terms such resources "CPU-Dependent Execution Resources" (CDERs). Five conceptual strategies for discovering and accessing CDERs are described and evaluated against key criteria, and all five strategies are compliant with GLUE 1.3, GLUE 2.0, or both. From this evaluation, two of the presented strategies clearly emerge as providing the greatest flexibility for publishing both static and dynamic CDER information and identifying CDERs that satisfy specific job requirements. Furthermore, a two-phase approach to job-submission is proposed for those jobs requiring access to CDERs. The approach is compatible with existing grid services. Examples are provided to illustrate job submission under each strategy
Document controversy classification based on the Wikipedia category structure
Get PDFDispute and controversy are parts of our culture and cannot be omitted on the Internet (where it becomes more anonymous). There have been many studies on controversy, especially on social networks such as Wikipedia. This free on-line encyclopedia has become a very popular data source among many researchers studying behavior or natural language processing. This paper presents using the category structure of Wikipedia to determine the controversy of a single article. This is the first part of the proposed system for classification of topic controversy score for any given text
Effects of Sparse Initialization in Deep Belief Networks
Get PDFDeep neural networks are often trained in two phases: first hidden layers are pretrained in an unsupervised manner and then network is fine-tuned with error backpropagation. Pretraining is often carried out using Deep Belief Networks (DBNs), with initial weights set to small random values. However, recent results established that well-designed initialization schemes, e.g. Sparse Initialization (SI), can greatly improve performance of networks that do not use pretraining. An interesting question arising from these results is whether such initialization techniques wouldn\u27t also improve pretrained networks? To shed light on this question, in this work we evaluate SI in DBNs that are used to pretrain discriminative networks. The motivation behind this research is our observation that SI has an impact on the features learned by a DBN during pretraining. Our results demonstrate that this improves network performance: when pretraining starts from sparsely initialized weight matrices networks achieve lower classification error after fine-tuning
Study of the Temporal-Statistics-Based Reputation Models for Q&A Systems
Get PDFQ&A systems are becoming a vital source of knowledge in many different domains. In some cases, they are also associated with services which provide employers with important information regarding the expertise of its potential employees. Therefore, the reputation earned in such communities can be associated with better job opportunities, and its significance is increasing. However, in a community where there is no direct financial motivation for participation, a reputation score is not solely an expertise metric. It is also a powerful motivator for remaining an active community member. Regardless of this complexity, algorithms for calculating reputation scores need to be as easy to understand (and implement) as possible. Therefore, the designers of the Q&A reputation system often implement a set of fixed rules, to some extent trading quality for quantity. Our goal is to study whether (and how) temporal statistics of a Q&A website can be incorporated into its reputation system. We want the proposed mechanism to dynamically adjust the impact of a single-answer evaluation on the reputation of its producer. We would like the proposed model to accurately reflect the expertise of content producers
Automated Credibility Assessment on Twitter
Get PDFn this paper, we make a practical approach to automated credibility assessment on Twitter. We describe the process behind the design of an automated classifier for information credibility assessment. As an addition, we propose practical implementation of TwitterBOT, a tool which is able to score submitted tweets while working in the native Twitter interface
Foundational Certification of Code Transformations Using Automatic Differentiation
Get PDFAutomatic Differentiation (AD) is concerned with the semantics augmentation of an input program representing a function to form a transformed program that computes the function\u27s derivatives. To ensure the correctness of the AD transformed code, particularly for safety critical applications, we aim at certifying the algebraic manipulations at the heart of the AD process. We have considered a WHILE-language and have shown how such proofs can be constructed by using an appropriate relational Hoare logic.In particular, we have shown how such inference rules can be constructed for both the forward and reverse mode AD by using an abductive logical reasoning
Agent-based hierarchical approach for executing bag-of-tasks in clouds
Get PDFNumerous unrelated, independent (no inter-task communication) tasks called “bag-oftasks”(BoTs) compared with message passing applications can be highly parallelised andexecuted in any acceptable order. A common practice when executing bag-of-tasks applications(BoT) is to exploit the master-slave topology. Cloud environments offer some featuresthat facilitate executing BoT applications. One of the approaches to control cloud resourcesis to use agents that can flexibly act in a dynamic environment. Given these assumptions wedesigned a combination of these approaches, which can be classified as: a distributed, hierarchicalsolution to the issue of scalable executing of bag-of-tasks. The concept of our systemrelates to a project that is focused on processing huge quantities of data incoming from anetwork of sensors by the Internet. Our aim is to create a mechanism for processing such dataas a system which executes jobs while exploiting load balancing for cloud resources using,e.g., Eucalyptus. The idea is to create a hybrid architecture which takes advantage of somecentralized parts of the system and full distributedness in other parts. On the other handwe balance dependencies between the system components using a hierarchic master-slavestructure
Comparison of incomplete data handling techniques for neuro-fuzzy system
Get PDFReal-life data sets sometimes miss some values. The incomplete data needs specialized algorithms or preprocessing that allows the use of the algorithms for complete data. The paper presents a comparison of various techniques for handling incomplete data in the neuro-fuzzy system ANNBFIS. The crucial procedure in the creation of a fuzzy model for the neuro-fuzzy system is the partition of the input domain. The most popular approach (also used in the ANNBFIS) is clustering. The analyzed approaches for clustering incomplete data are: preprocessing (marginalization and imputation) and specialized clustering algorithms (PDS, IFCM, OCS, NPS). The objective of our research is the comparison of the preprocessing techniques and specialized clustering algorithms to find the the most-advantageous technique for handling incomplete data with a neuro-fuzzy system. This approach is also the indirect validation of clustering