National Repository of Grey Literature 41 records found  1 - 10nextend  jump to record: Search took 0.00 seconds. 
Gradient learning for networks of smoothly pulse neurons
Hošek, Lukáš ; Šíma, Jiří (advisor) ; Petříčková, Zuzana (referee)
Networks of spiking neurons present a biologically more plausible alternative to perceptron networks, having great potential for processing time series. However, as of now, no practically usable learning algorithm has been known. SpikeProp, based on a gradient descent method, and its modifications have a fundamental problem with dis-continuity of spike creation and deletion. A new nontrivial gradient learning algorithm for a model of smoothly spiking neurons is proposed as a possible way to solve this problem. The goal of this work is to implement and test this model and eventually propose further improvements.
Transcultural Midwifery Care Focusing on the Mongolian Women Living in the Czech Republic
Hromková, Lucie ; Hroníková, Linda (advisor) ; Vrublová, Yvetta (referee) ; Šima, Jiří (referee)
The dissertation aims at the issue of transcultural care in the practice of midwives with a specific focus on Mongolian women, who currently represent the fifth largest group in terms of the number of births of foreigners in the Czech Republic. The theoretical framework of the text consists of chapters on nursing with subsequent differentiation of the field with regard to specialized care in the field of gynecology-obstetrics, as well as the divided subdiscipline of transcultural nursing care. An integral part of the theoretical part of the work are the basic general information about Mongolia and the Mongols. The practical part of the work is represented by my own qualitative and quantitative research. Qualitative research was carried out in the field through in-depth interviews with Mongolian women living in the Czech Republic and having experience with their own childbirth in a Czech maternity hospital. The aim of the research was to uncover possible peculiarities and requirements that a midwife may encounter from Mongolian women in the provision of transcultural care related to pregnancy, childbirth and the postnatal period. Additional quantitative research was conducted in the form of a questionnaire survey among Prague midwives and was focused on finding out the awareness, previous experience...
Building modeling from airborne laser scanning point clouds of low density
Hofman, Petr ; Potůčková, Markéta (advisor) ; Šíma, Jiří (referee)
A Laser scanning is a relatively recent remote sensing method which nevertheless quickly gained a prominent position, especially in the area of building detection and 3D modeling. Methods for building detection and 3D modeling initially used model-driven approaches which compare a laser scanning point cloud to a set of predefined building models. A method for determining building roof types using such approaches was presented in the article of Hofman, Potůčková (2012). An important advantage of model-driven approaches is their relative robustness to various data deficiencies such as low point density or low spatial accuracy. However, output of such methods is limited to a predefined set of building models and does not allow for diversity of actual buildings. For this reason, approaches used almost exclusively nowadays are data-driven. These methods search in datasets for a set of primitives (mostly roof planes) that are subsequently used to form the final model. This approach benefits from universality of resulting models but requires generally high data quality, especially in respect to input point cloud densities. The study of Hofman, Potůčková (2017) presented a data-driven method that can reliably detect buildings even in a very sparse point cloud in spite of using data-driven approach. At a density of...
Building modeling from airborne laser scanning point clouds of low density
Hofman, Petr ; Potůčková, Markéta (advisor) ; Šíma, Jiří (referee)
A Laser scanning is a relatively recent remote sensing method which nevertheless quickly gained a prominent position, especially in the area of building detection and 3D modeling. Methods for building detection and 3D modeling initially used model-driven approaches which compare a laser scanning point cloud to a set of predefined building models. A method for determining building roof types using such approaches was presented in the article of Hofman, Potůčková (2012). An important advantage of model-driven approaches is their relative robustness to various data deficiencies such as low point density or low spatial accuracy. However, output of such methods is limited to a predefined set of building models and does not allow for diversity of actual buildings. For this reason, approaches used almost exclusively nowadays are data-driven. These methods search in datasets for a set of primitives (mostly roof planes) that are subsequently used to form the final model. This approach benefits from universality of resulting models but requires generally high data quality, especially in respect to input point cloud densities. The study of Hofman, Potůčková (2017) presented a data-driven method that can reliably detect buildings even in a very sparse point cloud in spite of using data-driven approach. At a density of...
The Computational Power of Neural Networks and Representations of Numbers in Non-Integer Bases.
Šíma, Jiří
We briefly survey the basic concepts and results concerning the computational power of neural networks which basically depends on the information content of weight parameters. In particular, recurrent neural networks with integer, rational, and arbitrary real weights are classified within the Chomsky and finer complexity hierarchies. Then we refine the analysis between integer and rational weights by investigating an intermediate model of integer-weight neural networks with an extra analog rational-weight neuron (1ANN). We show a representation theorem which characterizes the classification problems solvable by 1ANNs, by using so-called cut languages. Our analysis reveals an interesting link to an active research field on non-standard positional numeral systems with non-integer bases. Within this framework, we introduce a new concept of quasi-periodic numbers which is used to classify the computational power of 1ANNs within the Chomsky hierarchy.
The Computational Power of Neural Networks and Representations of Numbers in Non-Integer Bases
Šíma, Jiří
We briefly survey the basic concepts and results concerning the computational power of neural networks which basically depends on the information content of weight parameters. In particular, recurrent neural networks with integer, rational, and arbitrary real weights are classified within the Chomsky and finer complexity hierarchies. Then we refine the analysis between integer and rational weights by investigating an intermediate model of integer-weight neural networks with an extra analog rational-weight neuron (1ANN). We show a representation theorem which characterizes the classification problems solvable by 1ANNs, by using so-called cut languages. Our analysis reveals an interesting link to an active research field on non-standard positional numeral systems with non-integer bases. Within this framework, we introduce a new concept of quasi-periodic numbers which is used to classify the computational power of 1ANNs within the Chomsky hierarchy.
Night Cooling in Condition of Czech Republic
Šíma, Jiří ; Katunský,, Dušan (referee) ; Plášek, Josef (referee) ; Šikula, Ondřej (advisor)
The PhD thesis deals with the theoretical evaluation of the potential use of night cooling techniques in condition of the Czech Republic. Mainly aims to evaluate the potential of night cooling, which uses the physical phenomenon of radiation against the night sky. This technique is widely used mainly in the dry and hot regions where there are the most suitable climatic conditions for this cooling method. The work is based on the current state of knowledge in the area of passive and low-energy cooling systems. These knowledges are applied to the conditions of Czech Republic and by using simulations of the energy, behaviour of the buildings explores the benefits in terms of improving of the internal microclimate and from the perspective of economy operation of conventional cooling equipment. In order to establish the benefits of night cooling on the operation of the building in the dynamically changing conditions, the modular simulation program TRNSYS was used, which is suitable for the analysis of energy systems of buildings and their behaviour. The results of simulation results were converted to thermal comfort parameters of the PMV and the PPD.
Topographic mapping of rock formations with the use of airborne laser scanning data
Lysák, Jakub ; Potůčková, Markéta (advisor) ; Brázdil, Karel (referee) ; Šíma, Jiří (referee)
Abstract This thesis focuses on topographic mapping of rock formations with the use of new technologies in a comprehensive manner, from airborne laser scanning (ALS) data acquisition and processing in rocky terrains, followed by their processing to the content of topographic databases and their cartographic processing in maps. The introduction discusses issues of importance for practice, and the relation between topographic mapping of rocks and other fields of human activity. The ALS section describes products for topographic mapping of rocks derived from ALS data, and discusses the specifics of ALS data acquisition and processing in wooded rugged terrain. Existing solutions of this problem are explained and their limitations are identified. Author's own approaches to solving this task are presented as case studies, including three made a further three designed experiments with ALS data processing and evaluation of their results. Recommendation regarding mapping of sandstone landscapes in Czechia have been also addressed. The topographic section describes the current representation of rocks and related objects in the ZABAGED database (Czech national digital topographic database), explains the historical context, analyzes this data and identifies their shortcomings in relation to the ALS. Research...
MARKETING MIX OF FIT BOX KLADNO
Šíma, Jiří ; Omcirk, Vilém (advisor) ; Pecinová, Markéta (referee)
Title: MARKETING MIX OF FIT BOX KLADNO Aims: The aim of this thesis is to collect information on the operation of the marketing mix from the customers and managers point of view. Based on these findings new proposal and recommendations were suggested in order to improve existing marketing mix. Methods: The analysis of marketing mix was performed by marketing research through questionnaire, interview and observation. Results: According to the results customers are least satisfied with changing rooms and web sites. In the contrary, most customers are satisfied with lessons quality and instructors attitude. Keywords: marketing mix, fit box, questionnaire, interview
The Computational Power of Neural Networks and Representations of Numbers in Non-Integer Bases
Šíma, Jiří
We briefly survey the basic concepts and results concerning the computational power of neural networks which basically depends on the information content of weight parameters. In particular, recurrent neural networks with integer, rational, and arbitrary real weights are classified within the Chomsky and finer complexity hierarchies. Then we refine the analysis between integer and rational weights by investigating an intermediate model of integer-weight neural networks with an extra analog rational-weight neuron (1ANN). We show a representation theorem which characterizes the classification problems solvable by 1ANNs, by using so-called cut languages. Our analysis reveals an interesting link to an active research field on non-standard positional numeral systems with non-integer bases. Within this framework, we introduce a new concept of quasi-periodic numbers which is used to classify the computational power of 1ANNs within the Chomsky hierarchy.

National Repository of Grey Literature : 41 records found   1 - 10nextend  jump to record:
See also: similar author names
11 ŠÍMA, Jan
3 ŠÍMA, Jaroslav
29 ŠÍMA, Jiří
6 Šíma, Jakub
11 Šíma, Jan
3 Šíma, Jaroslav
2 Šíma, Jindřich
29 Šíma, Jiří
2 Šíma, Josef
Interested in being notified about new results for this query?
Subscribe to the RSS feed.