|
| |
|
|
Geometric algebras and neural networks
Zapletal, Jakub ; Procházková, Jana (referee) ; Vašík, Petr (advisor)
This thesis deals with the use of geometric algebras in the field of neural networks. First, Conformal Geometric Algebra (CGA) and Geometric Algebra for Conics (GAC) and their Python implementations are introduced. The functioning of neural networks is then described, including an explanatory example. Finally, both topics are connected by using the appropriate library in the Python language, and the possibilities of geometric algebras for different models of neural networks are demonstrated on several examples.
|
|
|
Reconstruction of a Damaged Facial Image
Pleško, Filip ; Orság, Filip (referee) ; Goldmann, Tomáš (advisor)
Generatívne súperiace neurónové siete (GAN) sú rýchlo rozvýjajúca sa technológia, čo sa týka generovania obrazu. V tejto práci sú použité pre rekonštrukciu obrazu tváre, ktorá bola zakrytá predmetom. Práca najskôr vysvetľuje nutnú teóriu a potom preskúmava existujúce riešenia pre daný problém. Nakoniec navrhuje niekoľko rôznych GAN modelov za účelom zistiť, ktorá kombinácia vrstiev najviac pomáha pri rekonštrukcii tváre. Riešenia ktoré prispeli najviac, sú potom spolu skombinované aby bol vytvorený finálny model. Tento model je potom taktiež otestovaný na úlohe rozpoznavania tváre, aby bolo možné zistiť, či rekonštrukcia poškodeného snímku tváre môže byť pre túto úlohu nápomocná.
|
|
|
Gini coefficient maximization in binary logistic regression
Říha, Samuel ; Hanzák, Tomáš (advisor) ; Hlávka, Zdeněk (referee)
This Bachelor thesis describes a binary logistic regression model. By means of the term loss function a parameter estimation for the model is derived. A "rich" set of "proper" loss functions - beta family of Fisher-consistent loss functions - is defined. In the second part of the thesis, four basic goodness-of-fit criteria - Gini coefficient, C-statistics, Kolmogorov-Smirnov statistics and coefficient of determination R2 are defined. Further on, a possibility of parameter estimation by maximizing the Gini coefficient is analysed. Several algorithms are designed for this purpose. They are compared with so far existing methods in one simulated data set and three real ones. 1
|
|
|
Impact Of Loss Function On Multi-Frame Super-Resolution
Mezina, Anzhelika
Nowadays, one of the most popular topics in image processing is super-resolution. Thisproblem is getting more actual even in security, since monitoring cameras are everywhere and inthe case of an incident, it is necessary to recognize a person from records. A lot of approaches exist,which are able to reconstruct image, and the most of them are based on deep learning. The main focusof this work is to analyze, which loss function for neural networks is more effective for real-worldimage reconstruction. For this experiment chosen architecture and dataset are used for multi-framesuper-resolution for _x0002_8 scaling.
|
|
| |
|
|
Two-stage backtesting of Value-at-Risk models
Matyáš, Jan ; Seidler, Jakub (advisor) ; Brechler, Josef (referee)
Bachelor Thesis Two-stage backtesting of Value-at-Risk models Jan Matyáš Abstract This paper deals with a comparative evaluation of various Value-at-Risk models in terms of their prediction accuracy. We use two-stage backtesting procedure to find the most robust methodology in several aspects. Backtesting framework comprises of testing properties of independence, unconditional coverage, and conditional coverage and successive stage, that uses loss function allowing us to compare the two selected models from the previous part. Four European indices are taken to represent both well developed countries (DAX, ATX) and developing countries (PX, WIG). Models are examined over the period from January 1997 to February 2014. The best performing model in our selection appears to be the historical method with a 99% confidence interval. The use of stable distribution or lower confidence interval do not produce satisfactory results. Powered by TCPDF (www.tcpdf.org)
|
|
|
Gini coefficient maximization in binary logistic regression
Říha, Samuel ; Hanzák, Tomáš (advisor) ; Hlávka, Zdeněk (referee)
This Bachelor thesis describes a binary logistic regression model. By means of the term loss function a parameter estimation for the model is derived. A "rich" set of "proper" loss functions - beta family of Fisher-consistent loss functions - is defined. In the second part of the thesis, four basic goodness-of-fit criteria - Gini coefficient, C-statistics, Kolmogorov-Smirnov statistics and coefficient of determination R2 are defined. Further on, a possibility of parameter estimation by maximizing the Gini coefficient is analysed. Several algorithms are designed for this purpose. They are compared with so far existing methods in one simulated data set and three real ones. 1
|
|
|
Inflation Targeting in the Czech Republic
Klukavý, Petr ; Koderová, Jitka (advisor) ; Langer, Miroslav (referee)
This diploma thesis is focused on the description of Inflation Targeting regime in the Czech Republic. The paper is divided into three parts. The first part deals with inflation and its targeting and with the economical circumstances that led to the launch of this monetary policy regime in the Czech Republic. The next part concerns the central bank reaction function, transmission channels, evaluation of the inflation target sets and the description of prognostic models that the central bank uses for forecasting. Then main stress is laid on the new structural dynamic model "g3". The last part describes my own inflation prognosis, which is based on the time series analysis.
|
guest ::
