Otakar Fojt - Academia.edu (original) (raw)

Papers by Otakar Fojt

On selecting a constituent part of MU the" Overview of publishing activities" page will... more On selecting a constituent part of MU the" Overview of publishing activities" page will be displayed with information relevant to the selected constituent part. The" Overview of publishing activities" page is not available for non-activated items.

Hungarian Educational Research Journal

The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative eva... more The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative evaluation of research policies – based either on inputs (public investment, number of researchers), or outputs (publications, EU grants, number of patents). To give an example, we compare selected indicators across Visegrad countries in the period between 2006 and 2015. We conclude that both quantitative and qualitative perspectives as well as societal and political contexts should be taken into account when the performance of any R&D system and the impact of public investments into a public R&D sector are scrutinized.

ERGO

The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative eva... more The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative evaluation of research policies – based either on inputs (public investment, number of researchers), or outputs (publications, number of patents). To give an example we compare selected indicators across Visegrad countries in the period between 2006 and 2015. We conclude that both quantitative and qualitative perspectives as well as societal and political context should be taken into account when the performance of any R&D system and the impact of public investments into a public R&D sector are scrutinized.

This paper discusses the application of the technique of principal component analysis (PCA) to no... more This paper discusses the application of the technique of principal component analysis (PCA) to nonlinear time series analysis. First, we briefly describe state space reconstruction, then we continue with an overview of PCA with the emphasis on state space reconstruction applications. Next, we summarise our results from the analysis of industrial data and present the periodicity problem observed in the data. The report is concluded with a suggestion of alternative approaches to building Takens' matrix for reconstruction techniques. N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p, N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p,

In this paper we outline an approach to the analysis of sequential manufacturing data from indust... more In this paper we outline an approach to the analysis of sequential manufacturing data from industry using techniques from nonlinear dynamics. The basic idea is to consider a factory as a dynamical system. A process in the factory generates data, which contains information about the state of the system. If it is possible to analyse this data in such a way that knowledge of the system is increased, control and decision making processes can be improved. This will result, if applied, in a basis of competitive advantage to the factory. First, we give details of the type of recorded data and the necessary preprocessing techniques. We follow this with a description of our analysis. Our approach consists of state space reconstruction, applications of principal component analysis, a technique used to reveal structures in data and nonlinear deterministic prediction algorithms. The paper concludes with suggestions for future work and recommendations for improved data management. Appendices at the end of the paper contain more detailed information.

scientific articles have been written on the subject of data analysis and data knowledge discover... more scientific articles have been written on the subject of data analysis and data knowledge discovery. Nevertheless, the neophyte who wants to enter this field for the first time can have difficulties because of the large variety of many different methods, approaches and software tools. This overview gives a brief summary of modern methods of data analysis together with a description of some problems that may emerge. References at the end of the paper will guide the reader through essential books and articles written on the particular subject. The text focuses on real practical needswe have a problem, which methods are suitable for its solution? This is not a typical academic article. N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p, Department of Mathematics, University of York, England N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p, Department of Mathematics, University of York, England N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p,

In this paper we present rst a modi ed form of the fuzzy time series predictor used by Pang et al... more In this paper we present rst a modi ed form of the fuzzy time series predictor used by Pang et al. in 10] to predict telecom tra c and then the nonlinear deterministic predictor used by Ott et al. in 9] to predict chaotic time series. After describing the close connection between the two approaches, we produce a new nonlinear adaptive time series predictor, combining the best features of both methods. We test the new method on time series originating from nature and industry. Finally, we discuss the potential applications of the predictor and its limitations.

In our previous paper , we introduced the general algorithm of a nonlinear adaptive time series p... more In our previous paper , we introduced the general algorithm of a nonlinear adaptive time series predictor with fuzzy weighting and presented some experimental prediction results. This non-parametric (model-free) predictor is based on the pattern matching principle and uses a fuzzy weighting scheme for evaluation of relevant patterns. In this paper, we describe the practical implementation of this algorithm to a highspeed telecommunication networks, particularly for the prediction of the volume of incoming traffic at an aggregation point. We describe the optimisation system for adaptive selection of an appropriate embedding dimension and number of nearest neighbours. Furthermore, we test the optimiser's performance on both real and simulated telecom traffic data. We compare the obtained results with a standard 'moving average' prediction and finally discuss the results of the optimiser.

In this paper, we present a brief review of network control mechanisms and quality of service par... more In this paper, we present a brief review of network control mechanisms and quality of service parameters with the aim to narrow the focus of the research down to bandwidth allocation (BA) by using prediction techniques. A simple flow model of a telecom server is introduced and one possible switch variation is described for a queuing trial. Three methods of bandwidth allocation were applied: static BA (SBA), proportional BA (PBA) using the current queue lengths in each buffer and dynamic BA (DBA) using the current queue lengths of each buffer together with a prediction of incoming traffic for each queue. The Fuzzy-Non-Linear Traffic predictor introduced in [4] and optimised in [5] was used. The simulations were carried out using real traffic data of LAN traffic observed on an Ethernet at the Bellcore Morristown Research and Engineering facility [7,8] and data recorded at the University of York [6]. The simulation results of DBA have shown decrease in Cell Loss Ratio (CLR) and increase in Link Utilisation (LU) compared to PBA and SBA results. Further tests in a telecom lab are necessary to confirm the stability and long-term improvement of LU and CLR for DBA.

Science and Innovation Network, Feb 2012

Scan article, Aug 1, 2012

NNDG, Department of Mathematics, …, 1999

A deterministic model based on state space reconstruction and nonlinear dynamical methods was use... more A deterministic model based on state space reconstruction and nonlinear dynamical methods was used for prediction of nonlinear deterministic data. Industrial data is usually highly contaminated by noise, therefore it is necessary to establish the robustness and stability of prediction techniques in the presence of noise. This article investigates the noise sensitivity on Lorenz attractor data with the aim of establishing the quality and limitations of predictability for this nearest neighbours deterministic prediction model.

… in Medicine and Biology Magazine, IEEE, 1998

tive representation of the signal data, it is often necessary to find a simpler parametric repres... more tive representation of the signal data, it is often necessary to find a simpler parametric representation. One possibility for such a simpler system description is to use the so-called correlation dimension [9, 101. This parameter determines an order of the system, i.e., the number of dimensions needed to model the dynamics of the system. It does not have to be an integer value and its decimal portion expresses a measure of system complexity.

On selecting a constituent part of MU the" Overview of publishing activities" page will... more On selecting a constituent part of MU the" Overview of publishing activities" page will be displayed with information relevant to the selected constituent part. The" Overview of publishing activities" page is not available for non-activated items.

Hungarian Educational Research Journal

The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative eva... more The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative evaluation of research policies – based either on inputs (public investment, number of researchers), or outputs (publications, EU grants, number of patents). To give an example, we compare selected indicators across Visegrad countries in the period between 2006 and 2015. We conclude that both quantitative and qualitative perspectives as well as societal and political contexts should be taken into account when the performance of any R&D system and the impact of public investments into a public R&D sector are scrutinized.

ERGO

The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative eva... more The purpose of our contribution is to discuss shortcomings of purely descriptive quantitative evaluation of research policies – based either on inputs (public investment, number of researchers), or outputs (publications, number of patents). To give an example we compare selected indicators across Visegrad countries in the period between 2006 and 2015. We conclude that both quantitative and qualitative perspectives as well as societal and political context should be taken into account when the performance of any R&D system and the impact of public investments into a public R&D sector are scrutinized.

This paper discusses the application of the technique of principal component analysis (PCA) to no... more This paper discusses the application of the technique of principal component analysis (PCA) to nonlinear time series analysis. First, we briefly describe state space reconstruction, then we continue with an overview of PCA with the emphasis on state space reconstruction applications. Next, we summarise our results from the analysis of industrial data and present the periodicity problem observed in the data. The report is concluded with a suggestion of alternative approaches to building Takens' matrix for reconstruction techniques. N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p, N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p,

In this paper we outline an approach to the analysis of sequential manufacturing data from indust... more In this paper we outline an approach to the analysis of sequential manufacturing data from industry using techniques from nonlinear dynamics. The basic idea is to consider a factory as a dynamical system. A process in the factory generates data, which contains information about the state of the system. If it is possible to analyse this data in such a way that knowledge of the system is increased, control and decision making processes can be improved. This will result, if applied, in a basis of competitive advantage to the factory. First, we give details of the type of recorded data and the necessary preprocessing techniques. We follow this with a description of our analysis. Our approach consists of state space reconstruction, applications of principal component analysis, a technique used to reveal structures in data and nonlinear deterministic prediction algorithms. The paper concludes with suggestions for future work and recommendations for improved data management. Appendices at the end of the paper contain more detailed information.

scientific articles have been written on the subject of data analysis and data knowledge discover... more scientific articles have been written on the subject of data analysis and data knowledge discovery. Nevertheless, the neophyte who wants to enter this field for the first time can have difficulties because of the large variety of many different methods, approaches and software tools. This overview gives a brief summary of modern methods of data analysis together with a description of some problems that may emerge. References at the end of the paper will guide the reader through essential books and articles written on the particular subject. The text focuses on real practical needswe have a problem, which methods are suitable for its solution? This is not a typical academic article. N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p, Department of Mathematics, University of York, England N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p, Department of Mathematics, University of York, England N Ne et tw wo or rk ks s a an nd d N No on nl li in ne ea ar r D Dy yn na am mi ic cs s G Gr ro ou up p,

In this paper we present rst a modi ed form of the fuzzy time series predictor used by Pang et al... more In this paper we present rst a modi ed form of the fuzzy time series predictor used by Pang et al. in 10] to predict telecom tra c and then the nonlinear deterministic predictor used by Ott et al. in 9] to predict chaotic time series. After describing the close connection between the two approaches, we produce a new nonlinear adaptive time series predictor, combining the best features of both methods. We test the new method on time series originating from nature and industry. Finally, we discuss the potential applications of the predictor and its limitations.

In our previous paper , we introduced the general algorithm of a nonlinear adaptive time series p... more In our previous paper , we introduced the general algorithm of a nonlinear adaptive time series predictor with fuzzy weighting and presented some experimental prediction results. This non-parametric (model-free) predictor is based on the pattern matching principle and uses a fuzzy weighting scheme for evaluation of relevant patterns. In this paper, we describe the practical implementation of this algorithm to a highspeed telecommunication networks, particularly for the prediction of the volume of incoming traffic at an aggregation point. We describe the optimisation system for adaptive selection of an appropriate embedding dimension and number of nearest neighbours. Furthermore, we test the optimiser's performance on both real and simulated telecom traffic data. We compare the obtained results with a standard 'moving average' prediction and finally discuss the results of the optimiser.

In this paper, we present a brief review of network control mechanisms and quality of service par... more In this paper, we present a brief review of network control mechanisms and quality of service parameters with the aim to narrow the focus of the research down to bandwidth allocation (BA) by using prediction techniques. A simple flow model of a telecom server is introduced and one possible switch variation is described for a queuing trial. Three methods of bandwidth allocation were applied: static BA (SBA), proportional BA (PBA) using the current queue lengths in each buffer and dynamic BA (DBA) using the current queue lengths of each buffer together with a prediction of incoming traffic for each queue. The Fuzzy-Non-Linear Traffic predictor introduced in [4] and optimised in [5] was used. The simulations were carried out using real traffic data of LAN traffic observed on an Ethernet at the Bellcore Morristown Research and Engineering facility [7,8] and data recorded at the University of York [6]. The simulation results of DBA have shown decrease in Cell Loss Ratio (CLR) and increase in Link Utilisation (LU) compared to PBA and SBA results. Further tests in a telecom lab are necessary to confirm the stability and long-term improvement of LU and CLR for DBA.

Science and Innovation Network, Feb 2012

Scan article, Aug 1, 2012

NNDG, Department of Mathematics, …, 1999

A deterministic model based on state space reconstruction and nonlinear dynamical methods was use... more A deterministic model based on state space reconstruction and nonlinear dynamical methods was used for prediction of nonlinear deterministic data. Industrial data is usually highly contaminated by noise, therefore it is necessary to establish the robustness and stability of prediction techniques in the presence of noise. This article investigates the noise sensitivity on Lorenz attractor data with the aim of establishing the quality and limitations of predictability for this nearest neighbours deterministic prediction model.

… in Medicine and Biology Magazine, IEEE, 1998

tive representation of the signal data, it is often necessary to find a simpler parametric repres... more tive representation of the signal data, it is often necessary to find a simpler parametric representation. One possibility for such a simpler system description is to use the so-called correlation dimension [9, 101. This parameter determines an order of the system, i.e., the number of dimensions needed to model the dynamics of the system. It does not have to be an integer value and its decimal portion expresses a measure of system complexity.