Refine
Has Fulltext
- yes (14) (remove)
Document Type
- Book (4)
- Master's Thesis (4)
- Bachelor Thesis (3)
- Report (3)
Institute
- Physikalische Technik, Informatik (14) (remove)
Language
- English (14) (remove)
The book is the second of four volumes on data science and artificial intelligence. This second volume covers data visualization tool and techniques as well as fundamentals of supervised machine learning: linear regression, artificial neural networks, support-vector machines, decision trees, ensemble methods and more.
The book 'Reinforcment Learning: An Introduction' by Sutton and Barto is the standard text book for introductory courses to reinforcement learning. Next to concrete algorithms and extensive examples the book contains several fundamental results related to Markov decision processes (MDPs) and Bellman equations in Chapters 3 and 4. Unfortunately some proofs are missing, some theorems lack precise formulation, and for some results the line of arguments is quite garbled.
In this note we provide all missing proofs, give precise formulations of theorems and untangle the line of arguments. Further, we avoid using random variables and their expected values. Since we (like Sutton/Barto) restrict our attention to finite MDPs all expected values can be made explicit avoiding overloaded notation and murky conclusions.
This article bridges the gap between introductory literature like Sutton/Barto and research literature containing exact formulations and proofs of relevant results, but being less accessible to beginners due to higher generality and complexity.
In this project we develop an intelligent water meter based on software solutions offered by the IOTA Foundation. The water meter allows the customer to map water usage in real-time and pay water on demand, as well as the water provider to map water usage on a greater scale, regulate water supply during low- demand phases and offers regulatory functions to prepare for drought or humid climate, and to incentivise sustainable water usage in high-demand fields like agriculture. This functionality is phrased into a research issue:
Invention of a working prototype to demonstrate the IOTA Streams and Wallet protocols for a meter with additional focus on economic efficiency and the technical preparation of scalability.
Utilizing the IOTA streams protocol, a next generation secure data connection is established between the water meter and a server-sided software application. On this connection, water consumption is mapped into a data bench, and informative data and commands are issued to the graphical interface of the meter.
The IOTA wallet library is leveraged to provide customer accounts corresponding to their meter. IOTA tokens can be send to the account, which grants access to water in a matter of seconds. Depending on the regulatory scenario, water flow can be stopped as soon as the account is exhausted (i.e. public well), or an overdraw can be established in order to guarantee fulfilment of basic human rights (i.e. private households).
Since pricing data can be calculated server-sided and water consumption is mapped in very narrow intervals of as low as 4 seconds, the price can be used as tool to regulate consumption.
The physical components include an electrical ball valve to shut down water flow automatically, a command line interface to provide informative data,, a Raspberry Pi running the client-sided software application, and a water meter with MBus-Interface, as well as a Controlling Board to connect the Raspberry Pi with both peripheral devices.
The finished prototype shows, that water consumption can be mapped on a highly secure level, in near real-time, from afar, flexible for most applications.
The book is the first of four volumes on data science and artificial intelligence. This first volume covers fundamentals of data science: an introduction to Python programming, software libraries for data management, techniques for working with big data. It contains many exercises and projects with real-world data.
LEGO robotics sets are a well established tool for teaching programming in undergraduate courses. Starting with the now outdated EV3 set LEGO provided a Python programming interface and (inofficial) Linux support. The current LEGO Spike Prime set still provides Python programming, but no direct support for Linux.
In this report we collect and extend information on controlling Spike Prime robots from Linux hosts. We cover access to a robot's Python interpreter and code transfer as well as bidirectional robot-to-host communication via USB and Bluetooth. Results may be extended to robot-to-robot communication via Bluetooth.
Natural Language Processing (NLP) is a wide area in computer science and software engineering that finds its application not only in such fields as machine learning or artificial intelligence, but also has very quickly become a very popular solution to various questions and issues raised in the academic field. Those issues include information classification, text analysis, performing a so-called “smart search”, information grouping, providing feedback to academic papers, and many more. The application designed and implemented for this thesis targets solving a similar problem, which is described in details below.
The faculty of Physical Engineering and Informatics at the Westsächsische Hochschule Zwickau uses a software tool called “Quest”. The “Quest” serves as an online platform that enables students to ask questions regarding the module material introduced by the teaching professors. The professors can use these questions in order to keep track of the topics rising the most amount of uncertainty, as well as to clarify those questions during the lectures. Since the modules repeat over time, similar or identical questions are frequently asked by students.
The current functionality of the “Quest” tool, however, does not provide grouping or clustering questions with a similar content into one category. The reason why this is important is because in the future phases of the improvement of the functionality of the “Quest”, professors want to be able to link similar questions with a learning material or online resources that the students can benefit from. This, however, is only possible when each incoming question is labeled with a certain category, and all previously inputted questions associated with this category become visible through the UI of the tool. That would allow students to review the questions referring to the same category that were previously asked by others, as well as to get access to the study materials or resources associated with that category.
The functionality described above would significantly improve the study process and therefore a solution to this problem is needed. One way to solve this issue would be to perform a manual classification of existing questions into associated categories, which given the large size of the question database would be quite time- and effort-consuming. Another option would be to build an application that would be able to analyze the existing dataset of questions and, with the help of the NLP techniques, classify new questions asked by students and identify the similar ones from an existing database.
The tool implemented for the above-mentioned purposes aims to serve as text analytics and classification application that would be able to find and categorize similar text and questions provided by students and/or professors at the Westsächsische Hochschule Zwickau.
This bachelor thesis demonstrates the process of building a software tool using certain NLP techniques, such as tokenization, model training, categorization, NER (Named Entity Recognition), and POS (Parts of Speech tagging), which are defined and described in Chapter 2: “Literature Overview”. The relevance of these techniques in terms of the classification tool built in the context of this thesis is discussed in the section “Selection of NLP tools and techniques” of the “Methodology” chapter.
The aim of this thesis is to describe the methods and software tools used to develop a restful API for text classification, as well as to compare and analyze the effectiveness of two common NLP classifiers – the Naïve Bayes and Logistic Regression (Entropy Model) algorithms.
This work aimed to apply the Floating Film Transfer Method (FTM), developed by Kaneto et.al., as a new way of coating planar and nonplanar substrates with photoresist. Focus laid on the creation of a workflow to coat the substrate and process it by UV-Lithography and Nanoimprint lithography. Conventional coating methods like spin-, spray- or dip coating are well established in today’s industry but are limited in their capabilities to coat curved and structured surfaces. FTM offers the possibility to overcome these limitations. Therefore, two negative resists AR N 4400 and AR N 4600 10 as well as two positive resists AR P 3110 and AZ MIR 701 were drop cast on deionized water. The resist spreads into a thin film that can be transferred to a planar or curved substrate. Profilometric and ellipsometric measurements were conducted to evaluate the topography of the resist. A non-uniform thickness distribution was found depending on the resist and parameters like solid content, water temperature, and the amount of surfactant. UV-Lithography and Nanoimprintl ithography were successfully performed with these films. Resolutions as low as 2,3±0,4 µm were achieved by UV-Lithography of AR P 3110. A periodic pattern with a pitch of 1,51 µm was transferred by thermal nanoimprint lithography to AR N 4400 10. As proof of concept Laser Direct Writing was performed to structure AZ MIR 701coated on a glass vial.
The Jupyter ecosystem with JupyterHub and JupyterLab as its most prominent members is the de-facto standard for teaching Python programming and also for research in machine learning and data science. Although the Jupyter project is well documented, there are lots of settings and situations requiring deep knowledge of the internal workings of Jupyter, Linux and related software tools. This report describes three problems and possible solutions arising when installing and configuring a Jupyter-based teaching environment. These three problems are the installation and setup of the autograding tool nbgrader, the interplay between JupyterHub and Linux PAM, and providing access to WebDAV resources for users of JupyterHub.
Development and Characterization of a Dispersion-Encoded Method for Low-Coherence Interferometry
(2022)
This Open Access book discusses an extension to low-coherence interferometry by dispersion-encoding. The approach is theoretically designed and implemented for applications such as surface profilometry, polymeric cross-linking estimation and the determination of thin-film layer thicknesses. During a characterization, it was shown that an axial measurement range of 79.91 µm with an axial resolution of 0.1 nm is achievable. Simultaneously, profiles of up to 1.5 mm in length were obtained in a scan-free manner. This marked a significant improvement in relation to the state-of-the-art in terms of dynamic range. Also, the axial and lateral measurement range were decoupled partially while functional parameters such as surface roughness were estimated. The characterization of the degree of polymeric cross-linking was performed as a function of the refractive index. It was acquired in a spatially-resolved manner with a resolution of 3.36 x 10-5. This was achieved by the development of a novel mathematical analysis approach.
This bachelor thesis is about the development and design of the suture-based HeartStitch MR device for functional restoration of the mitral valve when a mitral regurgitation is present. The mitral valve is one of the four existing heart valves and due to their load, the mitral valve is most commonly affected by valve defects. In case of a mitral regurgitation, the leaflets of the mitral valve are not closing tight enough, which causes a reflux of the oxygenated blood from the left ventricle into the left atrium. As consequence, the human body won't be sufficiently supplied with oxygen anymore, which results in a disproportionate cardiac output. Untreated, this can lead to respiratory dis-tress, cardiac arrhythmias and stroke or even to death. In this case, a heart surgery is essential to improve the patient
Es wird ein neuartiges, interferometrisches Verfahren zur Analyse dünner Filme beschrieben.
Ferroelectric doped hafnium oxide might be able to revive the ferroelectric memory as a candidate for future non-volatile memories. Recent publications have reported ferroelectricity in HfO2 films deposited via physical vapor deposition (PVD) and thermal atomic layer deposition (ALD). Within this thesis, such layers have been fabricated by a plasma enhanced ALD process for the first time. Using an O2 remote plasma to oxidize TEMAHf and TMAl, respectively, Al:HfO2 stacks with different compositions have been deposited. Sputter deposited TiN was used as bottom and top electrode to form metal-insulator-metal capacitors. Compared to the PVD or thermal ALD films reported in literature, the plasma enhanced ALD films exhibited one to two orders higher leakage current at 1 V. The root of this higher leakage are oxygen vacancies as well as carbon and nitrogen impurities in the range of several at%. Despite the consequential perturbing effects, the main trends reported earlier are discernable. Just these perturbing effects turn out to substantiate the current explanation of a ferroelectric phase as route source for the observed electrical behavior.
This work analyzes and evaluates different approaches to translate UML state machines into C++ code. The first part of this thesis covers the ground of transforming information of a source language to a target language. It addresses the basics of language theory and different approaches of language transformation. The second part examines the properties and formalisms of state machines to value their characteristics for further reuse in the development cycle. The third part disassembles the programming language C++ with all its quirks and oddities. The last part puts all mentioned pieces together. Beyond this approach the thesis tries to point out several concepts of language engineering to ease the use of software languages for the language user as well as the language engineer. It scrutinizes diverging solutions with the resulting consequences.