Computer Science
Loading...
ISSN 1508-2806
e-ISSN: 2300-7036
Issue Date
2015
Volume
Vol. 16
Number
No. 4
Description
Reviewed by: Miguel Angel Guevara Lopez, Volodymyr Turchenko, Wacław Gudowski, Kamil Tucek, Bartosz Baliś, Depei Qian, Włodzimierz Funika, Renata Słota, Manuel Laso, Arnold Heemink, Grzegorz Jaśkiewicz, Aleksander Smywiński-Pohl, Curdin Derungs
Journal Volume
Computer Science
Vol. 16 (2015)
Projects
Pages
Articles
Effects of Sparse Initialization in Deep Belief Networks
(Wydawnictwa AGH, 2015) Grzegorczyk, Karol; Kurdziel, Marcin; Wójcik, Piotr Iwo
Deep neural networks are often trained in two phases: first, hidden layers are pretrained in an unsupervised manner, and then the 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 the performance of networks that do not use pretraining. An interesting question arising from these results is whether such initialization techniques wouldn’t 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 errors after fine-tuning.
The MCB code for numerical modeling of Fourth Generation nuclear reactors
(Wydawnictwa AGH, 2015) Oettingen, Mikołaj; Cetnar, Jerzy; Mirowski, Tomasz
R&D in the nuclear reactor physics demands state-of-the-art numerical tools that are able to characterize investigated nuclear systems with high accuracy. In this paper, we present the Monte Carlo Continuous Energy Burnup Code (MCB) developed at AGH University’s Department of Nuclear Energy. The code is a versatile numerical tool dedicated to simulations of radiation transport and radiation-induced changes in matter in advanced nuclear systems like Fourth Generation nuclear reactors.We present the general characteristics of the code and its application for modeling of Very-High-Temperature Reactors and Lead-Cooled Fast Rectors. Currently, the code is being implemented on the supercomputers of the Academic Computer Center (CYFRONET) of AGH University and will soon be available to the international scientific community.
Application of the Complex Event Processing system for anomaly detection and network monitoring
(Wydawnictwa AGH, 2015) Frankowski, Gerard; Jerzak, Marcin; Miłostan, Maciej; Nowak, Tomasz; Pawłowski, Marek
Protection of infrastructures for e-science, including grid environments and NREN facilities, requires the use of novel techniques for anomaly detection and network monitoring. The aim is to raise situational awareness and provide early warning capabilities. The main operational problem that most network operators face is integrating and processing data from multiple sensors and systems placed at critical points of the infrastructure. From a scientific point of view, there is a need for the efficient analysis of large data volumes and automatic reasoning while minimizing detection errors. In this article, we describe two approaches to Complex Event Processing used for network monitoring and anomaly detection and introduce the ongoing SECOR project (Sensor Data Correlation Engine for Attack Detection and Support of Decision Process), supported by examples and test results. The aim is to develop methodology that allows for the construction of next-generation IDS systems with artificial intelligence, capable of performing signature-less intrusion detection.
Overview and evaluation of conceptual strategies for accessing CPU-dependent execution resources in grid infrastructures
(Wydawnictwa AGH, 2015) Walsh, John; Dukes, Jonathan; Pierantoni, Gabriele; Coghlan, Brian
The 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.
Retrieval and interpretation of textual geolocalized information based on semantic geolocalized relations
(Wydawnictwa AGH, 2015) Korczyński, Wojciech
This paper describes a method for geolocalized information retrieval from natural language text and its interpretation by assigning it geographic coordinates. Proof-of-concept implementation is discussed, along with a geolocalized dictionary stored in a PostGIS/PostgreSQL spatial relational database. The discussed research focuses on the strongly inflectional Polish language, hence, additional complexity had to be taken into account. The presented method has been evaluated with the use of diverse metrics.