Search in Publications

Found 5164 publications

Investigation of various matrix factorization methods for large recommender systems

Bottyán Németh István Pilászy Gábor Takács Domonkos Tikk

Publication Name: Proceedings IEEE International Conference on Data Mining Workshops Icdm Workshops 2008

Publication Date: 2008-12-01

Volume: Unknown

Issue: Unknown

Page Range: 553-562

Description:

Matrix Factorization (MF) based approaches have proven to be efficient for rating-based recommendation systems. In this work, we propose several matrix factorization approaches with improved prediction accuracy. We introduce a novel and fast (semi)-positive MF approach that approximates the features by using positive values for either users or items. We describe a momentum-based MF approach. A transductive version of MF is also introduced, which uses information from test instances (namely the ratings users have given for certain items) to improve prediction accuracy. We describe an incremental variant of MF that efficiently handles new users/ratings, which is crucial in a real-life recommender system. A hybrid MF-neighborbased method is also discussed that further improves the performance of MF. The proposed methods are evaluated on the Netflix Prize dataset, and we show that they can achieve very favorable Quiz RMSE (best single method: 0.8904, combination: 0.8841) and running time. © 2008 IEEE.

Open Access: Yes

DOI: 10.1109/ICDMW.2008.86

Resource allocation and cost reduction by means of alternative solutions

Zsolt Tibor Kosztyán Andrea Bencsik Z. Perjés

Publication Name: Innovations and Advanced Techniques in Systems Computing Sciences and Software Engineering

Publication Date: 2008-12-01

Volume: Unknown

Issue: Unknown

Page Range: 556-559

Description:

In the course of planning and implementation of projects it occurs frequently that after the preliminary calculation of the optimal resources allocation - in order to obtain minimal total cost - a project can not be realized at a price which is expected by the inviter of the tender. In the following study we present a method by which, in case of a delimited maximal budget, we are able to determine the schedule, cost demands and resource requirements of a project to be implemented with the highest possible quality. © Springer Science+Business Media B.V. 2008.

Open Access: Yes

DOI: 10.1007/978-1-4020-8735-6_103

Joint experiments on the Tokamaks CASTOR and Ta10

E. Dufkova J. Brotankova M. Berta R. Dejarnac A. Fonseca G. M. Vorobjev A. Talebitaher J. Sentkerestiova B. Tál R. Gonzales V. Soldatov E. Sukhov I. Ďuran I. El Chama Neto C. R.Gutierrez Tapia A. W. Kraemeraflecken J. Ferreira A. Singh H. Hegazy H. Hegazy P. Khorshid L. Ruchko G. Kirnev N. Kirneva E. Del Bosco V. Vershkov A. Malaquias M. Hron B. Kuteev G. Mank M. Gryaznevich L. I. Krupnik V. Budaev A. Ovsyannikov A. Melnikov G. Van Oost M. Sokolov Y. Kuznetsov P. Peleman J. Zajac S. Zoletnik V. Weinzettl J. Stockel

Publication Name: Aip Conference Proceedings

Publication Date: 2008-12-01

Volume: 996

Issue: Unknown

Page Range: 24-33

Description:

Small tokamaks may significantly contribute to the better understanding of phenomena in a wide range of fields such as plasma confiement and energy transport; plasma stability in different magnetic configurations; plasma turbulence and its impact on local and global plasma parameters; processes at the plasma edge and plasmaawall interaction; scenarios of additional heating and nonainductive current drive; new methods of plasma profile and parameter control; development of novel plasma diagnostics; benchmarking of new numerical codes and so on. Furthermore, due to the compactness, flexibility, low operation costs and high skill of their personnel small tokamaks are very convenient to develop and test new materials and technologies. Small tokamaks are suitable and important for broad international cooperation, providing the necessary environment and manpower to conduct dedicated joint research programmes. In addition, the experimental work on small tokamaks is very appropriate for the education of students, scientific activities of postagraduate students and for the training of personnel for large tokamaks. The first Joint (Host Laboratory) Experiment (JE1) has been carried out in 2005 on the CASTOR tokamak at the IPP Prague, Czech Republic. It was jointly organized by the IPPaASCR and KFKI HAC, Budapest, involved 20 scientists from 7 countries and was supported through the IAEA and the ICTP, Trieste. The objective of JE1 was to perform studies of plasma edge turbulence and plasma confinement. Following the success of JE1, JE2 has been performed on Ta10 at RRC "Kurchatov Institute" in Moscow; 30 scientists from 13 countries participated in this experiment. This experiment aimed to continue JE1 turbulence studies, now extending them to the plasma core. Results of JE1 and JE2 will be overviewed and compared. © 2008 American Institute of Physics.

Open Access: Yes

DOI: 10.1063/1.2917019

Comparison of fuzzy rule-based learning and inference systems

J. Botzheim Krisztián Balázs L. T. Koczy

Publication Name: 9th International Symposium of Hungarian Researchers on Computational Intelligence and Informatics Cinti 2008

Publication Date: 2008-12-01

Volume: Unknown

Issue: Unknown

Page Range: 61-75

Description:

In our work we have compared various fuzzy rule based learning and inference systems. The base of the investigations was a modular system that we have implemented in C language. It contains several alternative versions of the two key elements of rule based learning - namely, the optimization algorithm and the inference method - which can be found in the literature. We obtained very different properties when combining these alternatives (changing the modules and connecting them) in all possible ways. The investigations determined the values of the quality measures (complexity and accuracy) of the obtained alternatives both analitically and experimentally where it was possible. Based on these quality measures the combinations have been ordered according to different aspects.

Open Access: Yes

DOI: DOI not available

Matrix factorization and neighbor based algorithms for the netflix prize problem

Bottyán Németh István Pilászy Gábor Takács Domonkos Tikk

Publication Name: Recsys 08 Proceedings of the 2008 ACM Conference on Recommender Systems

Publication Date: 2008-12-01

Volume: Unknown

Issue: Unknown

Page Range: 267-274

Description:

Collaborative filtering (CF) approaches proved to be effective for recommender systems in predicting user preferences in item selection using known user ratings of items. This subfield of machine learning has gained a lot of popularity with the Netix Prize competition started in October 2006. Two major approaches for this problem are matrix factorization (MF) and the neighbor based approach (NB). In this work, we propose various variants of MF and NB that can boost the performance of the usual ensemble based scheme. First, we investigate various regularization scenarios for MF. Second, we introduce two NB methods: one is based on correlation coeficients and the other on linear least squares. At the experimentation part, we show that the proposed approaches compare favorably with existing ones in terms of prediction accuracy and/or required training time. We present results of blending the proposed methods. © 2008 ACM.

Open Access: Yes

DOI: 10.1145/1454008.1454049

The polarization method combined with the Newton-Raphson technique in magnetostatic field problems

Miklós Kuczmann

Publication Name: Przeglad Elektrotechniczny

Publication Date: 2008-12-01

Volume: 84

Issue: 12

Page Range: 198-201

Description:

Nonlinear magnetic field problems can be solved by using the polarization method. After applying the Finite Element Method (FEM), a system of nonlinear equations can be derived, which can only be solved by iterative techniques. The fixed point iteration scheme and the Newton-Raphson technique are the most widely used algorithms to solve nonlinear equations. The first is known as a convergent, but slow algorithm, while the other is a fast solver, which convergence can be critical. This paper presents and compares the two methods.

Open Access: Yes

DOI: DOI not available

Estimating "level of safety" in traffic modeling using human error

Zsuzsanna Tóth-Szabó K. Kovács

Publication Name: 15th World Congress on Intelligent Transport Systems and ITS America Annual Meeting 2008

Publication Date: 2008-12-01

Volume: 4

Issue: Unknown

Page Range: 2562-2573

Description:

The aim of the paper is to present the results of a study about involving the human error into traffic modeling. The analysis of driver's behavior in a selected traffic situation (drivers do or do not stop at red light) provided us to define the meaning human error in the context. Based on Bayesian theorem, the probability of bad decision is one of the results of the first part of this paper. Having built the junctions in VISSIM environment, and after a detailed investigation of output parameters (in simulated traffic there is no car passing at red), a software module was developed for having the bad decision probabilities using VISSIM. Grouping drivers by driving style (aggressive, normal, slow) gives opportunity (defining numbers of driver passing at red) to find the missing link. The trial version of this software now is able to define the probability of bad decision using VISSIM outputs.

Open Access: Yes

DOI: DOI not available

Implementation of a basic flow model in traffic controlling

No authors available

Publication Name: 9th International Symposium of Hungarian Researchers on Computational Intelligence and Informatics, CINTI 2008

Publication Date: 2008-12-01

Volume:

Issue:

Page Range: 523-530

Description:

The contribution introduces a basic model, a manner, how to control - using the specific tools - A crossing of a highway and a lower volume road, while we stop the highway's traffic only in a critical case otherwise just slow it down.

Open Access: No

DOI: DOI not available

Nodal and edge finite element analysis of eddy current field problems

Miklós Kuczmann

Publication Name: Przeglad Elektrotechniczny

Publication Date: 2008-12-01

Volume: 84

Issue: 12

Page Range: 194-197

Description:

Eddy current field problems can be solved by different potential formulations based on the "quasi-static" Maxwell's equations. The potential formulations are obtained using a vector potential and a scalar potential, the most widely used techniques are the A, V-A, the T,φ-φformulations and their combinations. Vector potentials can be approximated by nodal or edge finite elements, scalar potentials are approximated by scalar elements. The paper presents and compares these formulations through TEAM Problem No. 7 containing a multiply connected region.

Open Access: Yes

DOI: DOI not available