Robustness Aspects of Knowledge Discovery
Kalina, Jan
2013 - English The sensitivity of common knowledge discovery methods to the presence of outlying measurements in the observed data is discussed as their major drawback. Our work is devoted to robust methods for information extraction from data. First, we discuss neural networks for function approximation and their sensitivity to the presence of noise and outlying measurements in the data. We propose to fit neural networks in a robust way by means of a robust nonlinear regression. Secondly, we consider information extraction from categorical data, which commonly suffers from measurement errors. To improve its robustness properties, we propose a regularized version of the common test statistics, which may find applications e.g. in pattern discovery from categorical data. Keywords: machine learning; outliers; neural networks; robust estimation Available at various institutes of the ASCR Robustness Aspects of Knowledge Discovery

The sensitivity of common knowledge discovery methods to the presence of outlying measurements in the observed data is discussed as their major drawback. Our work is devoted to robust methods for ...

Kalina, Jan
Ústav informatiky, 2013

System for Selection of Relevant Information for Decision Support
Kalina, Jan; Seidl, L.; Zvára, K.; Grünfeldová, H.; Slovák, Dalibor; Zvárová, Jana
2013 - English Keywords: decision support system; web-service; information extraction; high-dimension; gene expressions Available at various institutes of the ASCR System for Selection of Relevant Information for Decision Support

Kalina, Jan; Seidl, L.; Zvára, K.; Grünfeldová, H.; Slovák, Dalibor; Zvárová, Jana
Ústav informatiky, 2013

Source localization for EEG patterns relevant to motor imagery BCI control
Bobrov, P.; Frolov, A.; Húsek, Dušan; Tintěra, J.
2013 - English This work concerns spatial localization of sources of EEG patterns the most specific for control of the motor imagery based BCI. In our previous work we have shown that performance of Bayesian BCI classifier can be drastically improved by extraction of the most relevant independent components of the EEG signal. This paper presents the results of spatial localization of electrical brain activity sources which activity is reflected by the extracted components. The localization was performed by solving the inverse problem in EEG source localization, using individual finite-element head models. The sources were located in central sulcus (Brodmann area 3a), in the superior regions of post- and precentral gyri, and supplementary motor cortex. Keywords: brain computer interface; inverse EEG problem; brain activity location; signal classification; independent component analysis Available at various institutes of the ASCR Source localization for EEG patterns relevant to motor imagery BCI control

This work concerns spatial localization of sources of EEG patterns the most specific for control of the motor imagery based BCI. In our previous work we have shown that performance of Bayesian BCI ...

Bobrov, P.; Frolov, A.; Húsek, Dušan; Tintěra, J.
Ústav informatiky, 2013

Statistical Expectation of High Energy Physics Data Sets Separation Algorithms
Hakl, František
2013 - English Article focuses on the application of the basic results of the statistical learning theory known as Probabilistic Approximately Correct learning in the evaluation and post-processing of unique physical data obtained from the detectors of particle accelerators. The aim of this article is not direct separation of the measured data but evaluation of the appropriateness of separation methods used. The main principles and results of the PAC learning theory are briefly summarized, the main characteristics of selected multivariable data separation algorithms are studied from the VC-dimension point of view. Finally, based on actual data sets obtained from Tevatron D$\emptyset$ experiment, some practical hints for separation method selection and numerical computation are derived. Keywords: Probably Approximately Correct Learning; Refutability; HEP data separation; Neural networks; Decision trees; VC-dimension Available at various institutes of the ASCR Statistical Expectation of High Energy Physics Data Sets Separation Algorithms

Article focuses on the application of the basic results of the statistical learning theory known as Probabilistic Approximately Correct learning in the evaluation and post-processing of unique ...

Hakl, František
Ústav informatiky, 2013

Gaussian Radial and Kernel Networks with Varying and Fixed Widths
Kůrková, Věra
2013 - English The role of widths of Gaussians in computational models which they generate is investigated. Suitability of Gaussian kernel models with fixed widths for regression is proven in terms of their universal approximation capability. Large sets of argminima of error functionals minimized during learning from data over Gaussian networks with varying widths are described. Dependence of stabilizers modelling generalization on widths of Gaussian kernels and the input dimension is estimated. Keywords: Gaussian radial and kernel networks; functionally equivalent networks; universal approximators; stabilizers defined by Gaussian kernels Available at various institutes of the ASCR Gaussian Radial and Kernel Networks with Varying and Fixed Widths

The role of widths of Gaussians in computational models which they generate is investigated. Suitability of Gaussian kernel models with fixed widths for regression is proven in terms of their ...

Kůrková, Věra
Ústav informatiky, 2013

Shrinkage Approach for Gene Expression Data Analysis
Haman, Jiří; Valenta, Zdeněk; Kalina, Jan
2013 - English Keywords: shrinkage estimation; covariance matrix; high dimensional data; gene expression Available at various institutes of the ASCR Shrinkage Approach for Gene Expression Data Analysis

Haman, Jiří; Valenta, Zdeněk; Kalina, Jan
Ústav informatiky, 2013

Education for Medical Decision Support at EuroMISE Centre
Martinková, Patrícia; Zvára Jr., Karel; Dostálová, T.; Zvárová, Jana
2013 - English Keywords: education; decision support; knowledge evaluation; e-learning Available at various institutes of the ASCR Education for Medical Decision Support at EuroMISE Centre

Martinková, Patrícia; Zvára Jr., Karel; Dostálová, T.; Zvárová, Jana
Ústav informatiky, 2013

Introduction to Survival Analysis
Valenta, Zdeněk
2013 - English Survival analysis is concerned with analyzing time-to-event data where the event of interest usually represents some type of “failure”. In clinical medicine, the event of interest may be e.g. death of a patient from well specified causes, autoimmune rejection of the graft by the transplant recipient or other type of graft failure in transplant studies. In certain situations, however, the true survival outcomes may not be observable, because we have observed a so called “censoring event” which prevented the event of interest from occurring. Such censoring event may represent, for instance, loss of a particular subject from follow-up, occurrence of administrative censoring, which typically takes place in clinical trials, or we may indeed observe other type of “failure”, e.g. death from fatal injuries rather than from cardiovascular causes which were of primary interest in a particular clinical trial. In this article we will stress the importance of a key assumption relating censoring process to survival outcomes and review principle univariate survival analysis methods for uncorrelated data. We will review popular models for analyzing univariate survival data, many of which enable us quantifying effect the prognostic variables independently exert on survival outcomes. Model examples will cover the classes of non-parametric, parametric and semi-parametric methods. We will also review underlying assumptions of individual models and stress the importance of using appropriate models in analyzing univariate time-to-event data. Keywords: survival analysis; time-to-event data; censoring process; hazard function; survival time Available at various institutes of the ASCR Introduction to Survival Analysis

Survival analysis is concerned with analyzing time-to-event data where the event of interest usually represents some type of “failure”. In clinical medicine, the event of interest may be e.g. death of ...

Valenta, Zdeněk
Ústav informatiky, 2013

Capabilities of Radial and Kernel Networks
Kůrková, Věra
2013 - English Originally, artificial neural networks were built from biologically inspired units called perceptrons. Later, other types of units became popular in neurocomputing due to their good mathematical properties. Among them, radial-basis-function (RBF) units and kernel units became most popular. The talk will discuss advantages and limitations of networks with these two types of computational units. Higher flexibility in choice of free parameters in RBF will be compared with benefits of geometrical properties of kernel models allowing applications of maximal margin classification algorithms, modelling of generalization in learning from data in terms of regularization, and characterization of optimal solutions of learning tasks. Critical influence of input dimension on behavior of these two types of networks will be described. General results will be illustrated by the paradigmatic examples of Gaussian kernel and radial networks. Keywords: artificial neural networks; radial-basis-function; kernel units; advantages and limitations of networks; Gaussian kernel and radial networks Available at various institutes of the ASCR Capabilities of Radial and Kernel Networks

Originally, artificial neural networks were built from biologically inspired units called perceptrons. Later, other types of units became popular in neurocomputing due to their good mathematical ...

Kůrková, Věra
Ústav informatiky, 2013

Stochastic Models for Low Level DNA Mixtures
Slovák, Dalibor; Zvárová, Jana
2013 - English Keywords: forensic DNA interpretation; low level samples; allele peak heights; dropout probability Available at various institutes of the ASCR Stochastic Models for Low Level DNA Mixtures

Slovák, Dalibor; Zvárová, Jana
Ústav informatiky, 2013