|
|
Improving Robustness of Neural Networks against Adversarial Examples
Gaňo, Martin ; Matyáš, Jiří (referee) ; Češka, Milan (advisor)
Tato práce pojednává o kontradiktorních útocích na klasifikační modely neuronových sítí. Naším cílem je shrnout a demonstrovat kontradiktorní metody a ukázat, že představují vážný problém v strojovém učení. Důležitým přínosem této práce je implementace nástroje pro trénink robustního modelu na základě kontradiktorních příkladů. Náš přístup spočívá v minimalizaci maximalizace chybové funkce cílového modelu. Související práce a naše vlastní experimenty nás vedou k použití Projektovaného gradientního sestupu jako cílového útoku, proto trénujeme proti datům generovaným Projektovaným gradientním sestupem. Výsledkem použití nástroje je, že můžeme dosáhnout přesnosti více než 90% proti sofistikovaným nepřátelským útokům.
|
|
|
Modern methods of QRS detection
Fajkus, Jiří ; Kozumplík, Jiří (referee) ; Vítek, Martin (advisor)
This bachelor’s thesis deals with methods, which are used for processing of ECG signals, specifically for the detection of QRS complexes. The QRS detection is an essential part of every ECG analysis because it is a source of points with timing information, which are used for classification and measurement of other waves and intervals required for diagnostics. Some ways of QRS detection are described in this work and there is also a description of detector implemented in programming environment MATLAB, which is based on zero crossing counts. This QRS detector was then tested CSE database and signals from Holter examination.
|
|
|
Monte Carlo simulations of electron scattering in scanning transmission electron microscopy
Záchej, Samuel ; Hrubanová, Kamila (referee) ; Krzyžánek, Vladislav (advisor)
This thesis deals with an electron scattering in STEM microscopy on objects with dif-ferent shapes, such as cuboid, sphere and hollow capsule. Monte Carlo simulations are used for description of multiple electron scattering. Except the theoretical analysis of the electron scattering and simulation methods, the thesis contains design and realiza-tion of an algorithm simulating electron scattering in given objects. In addition, there is a design for robustness evaluation of the simulation, based on comparison between results and known signals for a given object. Reliability of the algorithm was verified by experimental measurements of the electron scattering on a carbon layer.
|
|
|
Robustnost fuzzy řízení
Hebelka, Marek
Hebelka M. Robustness of fuzzy control. Diploma thesis. Brno: Mendel University, 2023. The thesis deals with the assessment of the robustness of the fuzzy controller in the framework of removing the influence of disturbance variables in the form of a uniform unit step of 10, 20, 30 and 40% of the input signal and adjustment of time constants by 5, 10, 15 and 20% of the original value. The thesis describes the basics of fuzzy control, simulation tools, classic controllers, and plants. As part of the work, the settings of individual systems are subsequently described and the effects of the key parameters of the regulated systems are analyzed. The resulting work is an evaluation and assessment of the robustness of the regulators in the control systems evaluated.
|
|
|
Safe and Secure High-Risk AI: Evaluation of Robustness
Binterová, Eliška ; Špelda, Petr (advisor) ; Střítecký, Vít (referee)
The aim of the thesis is to examine Invariant Risk Minimization (IRM) as an existing method for achieving model robustness and assess whether it could potentially serve as means for conformity assessment in the emerging legislative framework of the European Artificial Intelligence Act. Research shows that many cases of erroneous performance in AI systems are caused by machine learning models lacking robustness to changes in data distributions and thus being unable to properly generalize to new environments. In order to achieve reliable performance, the models must exhibit a certain level of robustness to these changes. IRM is a relatively new method designed to achieve such outcomes. This is very much in alignment to the objectives of the EU AI Act that aims for trustworthy AI. The thesis thus examines the congruence of the IRM method and the requirements in the EU AI Act and asks whether IRM can serve as a universal method for ensuring safe and secure AI compliant with European legal requirements through the analysis of existing empirical and theoretical results.
|
|
|
Asset-Liability Management:Application of Stochastic Programmingwith Endogenous Randomness andContamination
Rusý, Tomáš ; Kopa, Miloš (advisor) ; Consigli, Giorgio (referee) ; Branda, Martin (referee)
Title: Asset-Liability Management: Application of Stochastic Programming with Endogenous Randomness and Contamination Author: RNDr. Tomáš Rusý Department: Department of Probability and Mathematical Statistics Supervisor: doc. RNDr. Ing. Miloš Kopa, PhD., Department of Probability and Mathematical Statistics Abstract: This thesis discusses a stochastic programming asset-liability management model that deals with decision-dependent randomness and a subsequent contamination analysis. The main model focuses on a pricing problem and the connected asset- liability management problem describing the typical life of a consumer loan. The endogeneity stems from the possibility of their customer rejecting the loan, the possibility of the customer defaulting on the loan and the possibility of prepay- ment which are all affected by the company's decision on interest rate of the loan. Another important factor, which plays a major role for liabilities, is the price of money in the market. There, we focus on the scenario generation procedure and develop a new calibration method for estimating the Hull-White model [Hull and White, 1990] under the real-world measure. We define the method for the gen- eral class of one-factor short-rate models and perform an extensive analysis to assess the estimation performance and...
|
|
|
Flexibility, Robustness and Discontinuities in Nonparametric Regression Approaches
Maciak, Matúš ; Hušková, Marie (advisor) ; Hlávka, Zdeněk (referee) ; Horová, Ivanka (referee)
Thesis title: Flexibility, Robustness and Discontinuity in Nonparametric Regression Approaches Author: Mgr. Matúš Maciak, M.Sc. Department: Department of Probability and Mathematical Statistics, Charles University in Prague Supervisor: Prof. RNDr. Marie Hušková, DrSc. huskova@karlin.mff.cuni.cz Abstract: In this thesis we focus on local polynomial estimation approaches of an unknown regression function while taking into account also some robust issues like a presence of outlying observa- tions or heavy-tailed distributions of random errors as well. We will discuss the most common method used for such settings, so called local polynomial M-smoothers and we will present the main statistical properties and asymptotic inference for this method. The M-smoothers method is especially suitable for such cases because of its natural robust flavour, which can nicely deal with outliers as well as heavy-tailed distributed random errors. Another important quality we will focus in this thesis on is a discontinuity issue where we allow for sudden changes (discontinuity points) in the unknown regression function or its derivatives respectively. We will propose a discontinuity model with different variability structures for both independent and dependent random errors while the discontinuity points will be treated in a...
|
|
|
Testing Structural Changes Using Ratio Type Statistics
Peštová, Barbora ; Hušková, Marie (advisor) ; Prášková, Zuzana (referee) ; Jarušková, Daniela (referee)
Testing Structural Changes Using Ratio Type Statistics Barbora Peštová Charles University in Prague, Faculty of Mathematics and Physics, Department of Probability and Mathematical Statistics, Czech Republic Abstract of the doctoral thesis We deal with sequences of observations that are naturally ordered in time and assume various underlying stochastic models. These models are parametric and some of the parameters are possibly subject to change at some unknown time point. The main goal of this thesis is to test whether such an unknown change has occurred or not. The core of the change point methods presented here is in ratio type statistics based on maxima of cumulative sums. Firstly, an overview of thesis' starting points is given. Then we focus on methods for detecting a gradual change in mean. Consequently, procedures for detection of an abrupt change in mean are generalized by considering a score function. We explore the possibility of applying the bootstrap methods for obtaining critical values, while disturbances of the change point model are considered as weakly dependent. Procedures for detection of changes in parameters of linear regression models are shown as well and a permutation version of the test is derived. Then, a related problem of testing a change in autoregression parameter is studied....
|
|
|
Improving Robustness of Speaker Recognition using Discriminative Techniques
Novotný, Ondřej ; Ferrer, Luciana (referee) ; Pollák, Petr (referee) ; Černocký, Jan (advisor)
Tato práce pojednává o využití diskriminativních technik v oblasti rozpoznávání mluvčích za účelem získání větší robustnosti těchto systémů vůči vlivům negativně ovlivňující jejich výkonnost. Mezi tyto vlivy řadíme šum, reverberaci nebo přenosový kanál. Práce je rozdělena do dvou hlavních částí. V první části se věnujeme teoretickému úvodu do problematiky rozpoznávání mluvčích. Popsány jsou jednotlivé kroky rozpoznávacího systému od extrakce akustických příznaků, extrakce vektorových reprezentací nahrávek, až po tvorbu finálního rozpoznávacího skóre. Zvláštní důraz je věnován technikám extrakce vektorové reprezentace nahrávky, kdy popisujeme dvě rozdílná paradigmata možného přístupu, i-vektory a x-vektory. Druhá část práce se již více věnuje diskriminativním technikám pro zvýšení robustnosti. Techniky jsou organizovány tak, aby odpovídaly postupnému průchodu nahrávky rozpoznávacím systémem. Nejdříve je věnována pozornost předzpracování signálu pomocí neuronové sítě pro odšumění a obohacení signálu řeči jako univerzální technice, která je nezávislá na následně použitém rozpoznávacím systému. Dále se zameřujeme na využití diskriminativního přístupu při extrakci příznaků a extrakci vektorových reprezentací nahrávek. Práce rovněž pokrývá přechod od generativního paradigmatu k plně diskriminativnímu přístupu v systémech pro rozpoznávání mluvčích. Veškeré techniky jsou následně vždy experimentálně ověřeny a zhodnocen jejich přínos. V práci je navrženo několik přístupů, které se osvědčily jak u generativního přístupu v podobě i-vektorů, tak i u diskriminativních x-vektorů, a díky nim bylo dosaženo významného zlepšení. Pro úplnost jsou, v oblasti problematiky robustnosti, do práce zařazeny i další techniky, jako je normalizace skóre, či více-scénářové trénování systémů. Závěrem se práce zabývá problematikou robustnosti diskriminativních systému z pohledu dat využitých při jejich trénování.
|
|
| |
guest ::
