|
|
Loop Analysis for LLVM IR Translation Validation Framework
Kubík, Jakub ; Kofroň, Jan (advisor) ; Parízek, Pavel (referee)
Bugs in compilers can have severe consequences. Apart from traditional methods like testing, one of the ways of keeping compilers correct that gained traction only in recent years is translation validation, a technique ensuring the semantic correctness of optimizations in compilers. Alive2 is an open-source translation validation framework for LLVM that is currently widely used by LLVM developers. In order to make any static analysis tool usable, the frequency of false alarms must be kept to a minimum. Alive2 was designed to have zero false alarms and has been very successful in this endeavor except in the case of certain loops. Our aim in this thesis is to analyze Alive2's loop algorithms in an attempt to find the cause of these false alarms. This was motivated by personal communication with authors of Alive2 who presented the false alarm issue in loops as one of the more challenging and pressing issues in Alive2. We were successful in pinpointing the cause of false alarms and even providing a fix for the issue. Our solution is now a part of the Alive2 framework. Furthermore, we have identified other potential issues in Alive2 which we discuss in the thesis as well.
|
|
|
Neural Network for Autocomplete in the Browser
Kubík, Ján Jakub ; Zemčík, Pavel (referee) ; Kolář, Martin (advisor)
The goal of this thesis is to create and train a neural network and use it in a web browser for English text sequence prediction during writing of text by the user. The intention is to simplify the writing of frequent phrases. The problem is solved by employing a recurrent neural network that is able to predict output text based on the text input. Trained neural network is then used in a Google Chrome extension. By normalized ouput of the neural network, text choosing by sampling decoding algorithm and connecting, the extension is able to generate English word sequences, which are shown to the user as suggested text. The neural network is optimized by selecting the right loss function, and a suitable number of recurrent layers, neurons in the layers, and training epochs. The thesis contributes to enhancing the everyday user experience of writing on the Internet by using a neural network for English word sequence autocomplete in the browser.
|
|
|
Řízení likvidity ve vybraném podniku
Kubík, Jakub
Kubík, J. Cash management in the selected company. Bachelor thesis. Brno: Mendel University, 2023. The bachelor thesis focuses on the examination of financial analysis and li-quidity management in a selected company between 2017 and 2021. The per-formance of the company is evaluated through financial analysis and its corre-sponding ratios. After the analysis is completed, suggestions are offered to im-prove the company's well-being
|
|
|
An Empirical Evaluation of Selfish Mining Strategies on Various Blockchains
Kubík, Ján Jakub ; Košťál,, Kristián (referee) ; Homoliak, Ivan (advisor)
Táto práca si kladie za cieľ posilniť bezpečnosť blockchainu tým, že prehĺbi pochopenie útokov typu sobecké ťaženie v rámci rôznych konsenzuálnych protokolov. Predchádzajúce výskumy sa sústredili na jednotlivé protokoly s jedným útočníkom. Štúdie s viacerými útočníkmi boli len pre konsenzuálny protokol Nakamoto. Na preklenutie tejto medzery bol vytvorený diskrétny, statický, stochastický simulačný framework na analýzu sobecnkého ťaženia s viacerými útočníkmi naprieč rôznymi protokolmi vrátane Nakamoto, Subchain a Strongchain. S použitím tohoto frameworku boli zreprodukované hranice pre úspešný útok na Nakamoto a Strongchain. Ďalej bola zistená hranica pre úspešný útok na Strongchain pre dvoch útočníkov. Pre Subchain bolo nezávisle skúmané sobecké ťaženie na slabých a silných blokoch. Bolo zistené, že len pre silné bloky je sobecké ťaženie ziskové s mierne vyššími hraniciami ako v Nakamoto. Hlavným prínosom tejto práce je vytvorený framework s overenými hranicami a vývojom sobeckého ťaženia pre podporované protokoly. Taktiež tento framework je ľahko rozšíriteľný o nové konsenzuálne protokoly, čo umožňuje hodnotiť odolnosť rôznych Proof of Work protokolov voči útokom typu sobecké ťaženie.
|
|
|
Neural Network for Autocomplete in the Browser
Kubík, Ján Jakub ; Zemčík, Pavel (referee) ; Kolář, Martin (advisor)
The goal of this thesis is to create and train a neural network and use it in a web browser for English text sequence prediction during writing of text by the user. The intention is to simplify the writing of frequent phrases. The problem is solved by employing a recurrent neural network that is able to predict output text based on the text input. Trained neural network is then used in a Google Chrome extension. By normalized ouput of the neural network, text choosing by sampling decoding algorithm and connecting, the extension is able to generate English word sequences, which are shown to the user as suggested text. The neural network is optimized by selecting the right loss function, and a suitable number of recurrent layers, neurons in the layers, and training epochs. The thesis contributes to enhancing the everyday user experience of writing on the Internet by using a neural network for English word sequence autocomplete in the browser.
|
guest ::
RSS feed.