Refine
Document Type
- Master's Thesis (121) (remove)
Year of publication
Language
- English (121) (remove)
Keywords
- Maschinelles Lernen (24)
- Vektorquantisierung (8)
- Blockchain (7)
- Algorithmus (5)
- Bioinformatik (5)
- Neuronales Netz (5)
- Deep learning (4)
- Kryptologie (4)
- Virtuelle Währung (4)
- China (3)
This thesis comprehensively explores factors contributing to malaria-induced anemia and severe malarial anemia (SMA). The study utilizes a comprehensive dataset to investigate immunological interactions, genetic variations, and temporal dynamics. Findings highlight the complex interplay between immune markers, genetic traits, and cohort-specific influences. Notably, age, HIV status, and genetic variations emerge as crucial factors influencing anemia risk. The incorporation of Poisson regression models sheds light on the genetic underpinnings of SMA, emphasizing the need for personalized interventions. Overall, this research provides valuable insights into the multifaceted nature of malaria-induced complications, paving the way for further molecular investigations and targeted interventions.
As new sensors are added to VR headsets, more data can be collected. This introduces a new potential threat to user privacy. We focused on the feasibility of extracting personal information from eye-tracking. To achieve this, we designed a preliminary user study focusing on the pupil response to audio stimuli. We used a variation of machine learning models to test the collected data to determine the feasibility of obtaining information such as the age or gender of the participant. Several of the experiments show promise for obtaining this information. We were able to extract with reasonable certainty whether caffeine was consumed and the gender of the participant. This demonstrates the unknown threat that embedded sensors pose to users. A further studies are planned to verify the results.
Computationally solving eigenvalue problems is a central problem in numerical analysis and as such has been the subject of extensive study. In this thesis we present four different methods to compute eigenvalues, each with its own characteristics, strengths and weaknesses. After formally introducing the methods we use them in various numerical experiments to test speed of convergence, stability as well as performance when used to compute eigenfaces, denoise images and compute the eigenvector centrality measure of a graph.
Robust soft learning vector quantization (RSLVQ) is a probabilistic approach of Learning vector quantization (LVQ) algorithm. Basically, the RSLVQ approach describes its functionality with respect to Gaussian mixture model and its cost function is defined in terms of likelihood ratio. Our thesis work involves an approach of modifying standard RSLVQ with non-Gaussian density functions like logistic, lognormal, and Cauchy (referred as PLVQ). In this approach, we derive new update rules for prototypes using gradient of cost function with respect to non-Gaussian density functions. We also derive new learning rules for the model parameters like s and s, by differentiating the cost function with respect to parameters. The main goal of the thesis is to compare the performance results of PLVQ model with Gaussian-RSLVQ model. Therefore, the performance of these classification models have been tested on the Iris and Seeds dataset. To visualize the results of the classification models in an adequate way, the Principal component analysis (PCA) technique has been used.
Machine learning models for timeseries have always been a special topic of interest due to their unique data structure. Recently, the introduction of attention improved the capabilities of recurrent neural networks and transformers with respect to their learning tasks such as machine translation. However, these models are usually subsymbolic architectures, making their inner working hard to interpret without comprehensive tools. In contrast, interpretable models such learning vector quantization are more transparent in the ability to interpret their decision process. This thesis tries to merge attention as a machine learning function with learning vector quantization to better handle timeseries data. A design on such a model is proposed and tested with a dataset used in connection with the attention based transformers. Although the proposed model did not yield the expected results, this work outlines improvements for further research on this approach.
In this paper, we conduct experiments to optimize the learning rates for the Generalized Learning Vector Quantization (GLVQ) model. Our approach leverages insights from cog- nitive science rooted in the profound intricacies of human thinking. Recognizing that human-like thinking has propelled humankind to its current state, we explore the applica- bility of cognitive science principles in enhancing machine learning. Prior research has demonstrated promising results when applying learning rate methods inspired by cognitive science to Learning Vector Quantization (LVQ) models. In this study, we extend this approach to GLVQ models. Specifically, we examine five distinct cognitive science-inspired GLVQ variants: Conditional Probability (CP), Dual Factor Heuristic (DFH), Middle Symmetry (MS), Loose Symmetry (LS), and Loose Symme- try with Rarity (LSR). Our experiments involve a comprehensive analysis of the performance of these cogni- tive science-derived learning rate techniques across various datasets, aiming to identify optimal settings and variants of cognitive science GLVQ model training. Through this research, we seek to unlock new avenues for enhancing the learning process in machine learning models by drawing inspiration from the rich complexities of human cognition. Keywords: machine learning, GLVQ, cognitive science, cognitive bias, learning rate op- timization, optimizers, human-like learning, Conditional Probability (CP), Dual Factor Heuristic (DFH), Middle Symmetry (MS), Loose Symmetry (LS), Loose Symmetry with Rarity (LSR).
Adversarial robustness of a nearest prototype classifier assures safe deployment in sensitive use fields. Much research has been conducted on artificial neural networks regarding their robustness against adversarial attacks, whereas nearest prototype classifiers have not chalked similar successes. This thesis presents the learning dynamics and numerical stability regarding the Crammer-normalization and the Hein-normalization for adversarial robustness of nearest prototype classifiers. Results of conducted experiments are penned down and analyzed to ascertain the bounds given by Saralajew et al. and Hein et al. for adversarial robustness of nearest prototype classifiers.
Traditional user management on the Internet has historically required individuals to give up control over their identities. In contrast, decentralized solutions promise to empower users and foster decentralized interactions. Over the last few years, the development of decentralized accounts and tokens has significantly increased, aiming at broader user adoption and shared social economies.
This thesis delves into smart contract standards and social infrastructure for Ethereum-based blockchains to enable identity-based data exchange between abstracted blockchain accounts. In this regard, the standardization landscapes of account and social token developments were analyzed in-depth to form guidelines that allow users to retain complete control over their data and grant access selectively.
Based on the evaluations, a pioneering Solidity standard is presented, natively integrating consensual restrictive on-chain assets for abstracted blockchain accounts. Further, the architecture of a decentralized messaging service has been defined to outline how new token and account concepts can be intertwined with efficient and minimal data-sharing principles to ensure security and privacy, while merging traditional server environments with global ledgers.
Analysis of the Forensic Preparation of Biometric Facial Features for Digital User Authentication
(2023)
Biometrics has become a popular method of securing access to data as it eliminates the need for users to remember a password. Although exploiting the vulnerabilities of biometric systems increased with their usage, these could also be helpful during criminal casework.
This thesis aims to evaluate approaches to bypass electronic devices with forged faces to access data for law enforcement. Here, obtaining the necessary data in a timely manner is critical. However, unlocking the devices with a password can take several years with a brute force attack. Consequently, biometrics could be a quicker alternative for unlocking.
Various approaches were examined to bypass current face recognition technologies. The first approaches included printing the user's face on regular paper and aimed to unlock devices performing face recognition in the visible spectrum. Further approaches consisted of printing the user's infrared image and creating three-dimensional masks to bypass devices performing face recognition in the near-infrared. Additionally, the underlying software responsible for face recognition was reverse-engineered to get information about its operation mode.
The experiments demonstrate that forged faces can partly bypass face recognition and obtain secured data. Devices performing face recognition in the visible spectrum can be unlocked with a printed image of the user's face. Regarding devices with advanced near-infrared face recognition, only one could be bypassed with a three-dimensional face mask. In addition, its underlying software provided evidence about the demands of face recognition. Other devices under attack remained locked, and their software provided no clues.
Analysis of Continuous Learning Strategies at the Example of Replay-Based Text Classification
(2023)
Continuous learning is a research field that has significantly boosted in recent years due to highly complex machine and deep learning models. Whereas static models need to be retrained entirely from scratch when new data get available, continuous models progressively adapt to new data saving computational resources. In this context, this work analyzes parameters impacting replay-based continuous learning approaches at the example of a data-incremental text classification task using an MLP and LSTM. Generally, it was found that replay improves the results compared to naive approaches but achieves not the performance of a static model. Mainly, the performances increased with more replayed examples, and the number of training iterations has a significant influence as it can partly control the stability-plasticity-trade-off. In contrast, the impact of balancing the buffer and the strategy to select examples to store in the replay buffer were found to have a minor impact on the results in the present case.
RNA tertiary contact interactions between RNA tetraloops and their receptors stabilize the folding of ribosomal RNA and support the maturation of the ribosome. Here we use FRET assisted structure prediction to develop structural models of two ribosomal tertiary contacts, one consisting of a kissing loop and a GAAA tetraloop and one consisting of the tetraloop receptor (TLR) and a GAAA tetraloop. We build bound and unbound states of the ribosomal contacts de novo, label the RNA in silico and compute FRET histograms based on MD simulations and accessible contact volume (ACV) calculations. The predicted mean FRET efficiency from molecular dynamics (MD) simulations and ACV determination show agreement for the KL-TLGAAA construct. The KL construct revealed too high FRET efficiency and artificial dye behavior, which requires further investigation of the model. In the case of the TLR, the importance of the correct dye and construct parameters in the modeling was shown, which also leads to a renewed modeling. This hybrid approach of experiment and simulation will promote the elucidation of dynamic RNA tertiary contacts and accelerate the discovery of novel RNA interactions as potential future drug targets.
To investigate the effects of climate change on interactions within ecosystems, a microcosm experiment was conducted. The effects of temperature increase and predator diversity on Collembola communities and their decomposition rate were investigated. The predators used were mites and Chilopods, whose predation effects on several response variables were analysed. This data included Collembola abundance, biomass and body mass as well as basal respiration and microbial biomass carbon. These response variables were tested against the predictors in several models. Temperature showed high significance in interaction with mite abundance in almost all models. Furthermore, the results of the basal respiration and microbial biomass carbon support the suggestion of a trophic cascade within the animal interaction.
This scientific work reveals the potential for the development of the renewable energy market, due to many reasons. The reasons are the unstable political situation in the world, rising energy prices, environmental degradation and the growing demand of Ger man residents for government measures to reduce the negative impact on the environment. This work is related to business planning and development using strategies based on the above reasons. The purpose of the study is to develop methods for successfully regulating the market for renewable resources to solve the problem of environmental pollution through the promotion of environmentally friendly products. The work explores the driving forces and problems hindering the development of the market for renewable resources. The problems raised concerned all interested parties, from consumers and producers to the state body for regulating and stimulating the industry . An analysis was also made of the methods of environmentally oriented companies and the tools they use to strengthen their positions in the market. Based on the data obtained from the conducted research, a concept and business strategy for a new environmentally oriented generation” was created. The business consulting company “Sun’s idea of the new company is to involve all parties using marketing tools, creating a healthy competitive environment among commercial companies and benefiting not only the companies themselves but also the end user of the products and the German government.
The occurence of prostate cancer (PCa) has been consistently rising since three decades and remains the third leading cause of cancer-related deaths after lung and bowel cancer in Germany. Despite of new methods of early detection, such as prostate-specific antigen (PSA) testing, it persists to be the most common cancer in german men with over 63,400 new diagnoses in Germany every year and exhibits high prevalence in other countries of Northern andWestern Europe as well [64]. Men over the age of 70 are most commonly affected by the lethal disease, whereas an indisposition before 50 is rare. The malignant prostate tumor can be healed through operation or irradiation while the cancer hasn’t reached the stage of metastasis in which other therapeutic methods have to be employed [14] [15]. In the metastatic phase, the patient usually exhibits symptoms when the tumors size affects the urethra or the cancer spreads to other tissue, often the bones [16].
The high prevalence of this disease marks the importance of further research into prognosis and diagnosis methods, whereby identification of further biomarkers in PCa poses a major topic of scientific analysis. For this task, the effectiveness of high-throughput RNA sequencing of the transcriptome (RNA molecules of an organism or specific cell type) is frequently exploited [66]. RNA sequencing or RNA-Seq in short, offers the possibility of transcriptome assessment, enabling the identification of transcriptional aberrations in diseases as well as uncharacterized RNA species such as non-coding RNAs (ncRNAs) which remain undetected by conventional methods [49]. To alleviate interpretation of the sequenced reads they are assembled to reconstruct the transcriptome as close to the original state as possible, thus enabling rapid detection of relevant biomolecules in the data [49]. Transcriptomic studies often require highly accurate and complete gene annotations on the reference genome of the examined organism. However, most gene annotations and reference genomes are far from complete, containing a multitude of unidentified protein-coding and non-coding genes and transcripts. Therefore, refinement of reference genomes and annotations by inclusion of novel sequences, discovered in high quality transcriptome assemblies, is necessary [24].
Glycans play an important role in the intracellular interactions of pathogenic bacteria. Pathogenic bacteria possess binding proteins capable of recognizing certain sugar motifs on other cells, which are found in glycan structures. Artificial carbohydrate synthesis allows scientists to recreate those sugar motifs in a rational, precise, and pure form. However, due to the high specificity of sugar-binding proteins, known as lectins, to glycan structures, methods for identifying suitable binding agents need to be developed. To tackle this hurdle, the Fraunhofer Institute for Cell Therapy and Immunology (Fraunhofer IZI) and the Max-Planck Institute of Colloids and Interfaces (MPIKG) developed a binding assay for the high throughput testing of sugar motifs that are presented on modular scaffolds formed by the assembly of four DNA strands into simple, branched DNA nanostructures. The first generation of this assay was used in combination with bacteria that express a fluorescent protein as a proof-of-concept. Here, the assay was optimized to be used with bacteria not possessing a marker gene for a fluorescent protein by staining their genomic DNA with SYBR® Green. For the binding assay, DNA nanostructures were combined with artificially synthesized mannose polymers, typical targets for many lectins on the surface of bacteria, presenting them in a defined constellation to bind bacteria strongly due to multivalent cooperativity. The testing of multiple mannose polymers identified monomeric mannose with a 5’-carbon linker and 1,2-linked dimeric mannose with linker as the best binding candidates for E. coli, presumably due to binding with the FimH protein on the surface. Despite similarities between the FimH proteins of E. coli and K. pneumoniae, binding was only observed between E. coli and the different sugar molecules on DNA structures. Furthermore, the degree of free movement seemed to affect the binding of mannose polymers to targeted proteins, since when utilizing a more flexible DNA nanostructure, an increase in binding could be observed. An alternative to the simple DNA nanostructures described above is the use of larger, more complex DNA origami structures consisting of several hundred strands. DNA origami structures are capable of carrying dozens of modifications at the same time. The results for the DNA origami structure showed a successful functionalization with up to 71 1,2-linked dimeric mannose with linker molecules. These results point towards a solution for the high-throughput analysis of potential binding agents for pathogenic bacteria e.g. as an alternative treatment for antibiotic-resistant.
Cryptorchidism is the most common disorder of sex development in dogs. It describes a failure of one or both testes to descend into the scrotum in due time. It is a heritable multifactorial disease. In this work, selected dogs of a german sheep poodle breed were sequenced with nanopore sequencing and subsequently examined for genetic variations correlating with cryptorchidism. The relationships of the studied dogs were also analyzed and visually processed.
Assessment of COI and 16S for insect species identification ti determine the diet of city bats
(2023)
Despite the numerous benefits of urbanization to human living conditions, urbanization has also negatively affected humans, their environment, and other organisms that share urban habitats with humans. Undoubtedly adverse while some wild animals avoid living in urban areas, others are more tolerant or prefer life in urban habitats. There are more than 1,400 species of bats in the world.
Therefore, they have the potential to contribute significantly to the mammalian biodiversity in urban areas. Insectivorous bats species play a key role in agriculture by improving yields and reducing chemical pesticide costs. Using metabarcoding, it is possible to determine the prey consumed by these noctule mammals based on the DNA fragments in their fecal pellets. This study
aimed to evaluate COI and 16S metabarcodes for insect species identification to determine the diet of metropolitan bats. For this purpose, COI and 16S metabarcodes were extracted, amplified, and sequenced from 65 bat feces collected in the Berlin metropolitan areas. Following a taxonomic annotation, I found that 73% of all identified insects could only be detected using the COI method, while 15% could be recovered using the 16S approach. Just 12% of all detected insects were identified simultaneously by both markers. According to this result, COI is more suitable for the taxonomic identification of insects from bat feces. However, given the bias of COI primers, it is recommended to use both markers for a more precise estimation of species diversity. Additionally,based on the insect species identified, I noticed that urban bats fed mainly on Diptera, Coleoptera,and Lepidoptera. The bat species Nyctalus noctula was most abundant in the samples. His diet analysis revealed that 91% of the samples contained the insect species Chironomus plumosus. 14 pest insect species were also found in his diet.
Our current research aims to establish a complete ribonucleic acid (RNA) production line from plasmid design to purification of in vitro transcribed RNA and labeling of RNA. RNA is the central molecule within the central dogma of molecular biology and is involved in most essential processes within a cell[1]. In many cases, only compact three-dimensional structures of the respective RNA are able to fulfill their function. In this context, RNA tertiary contacts such as kissing loops and pseudoknots are essential to stabilize three-dimensional folding[2]. We will produce a tertiary contact consisting of a kissing loop and a GAAA tetraloop that occurs in eukaryotic ribosomal RNA[3,4]. The RNA sequence is integrated into a vector plasmid. Subsequently, the plasmid is amplified in E. coli. After following plasmid purification steps, the RNA sequence will be transcribed in vitro[5,6]. In order for the RNA be used for Förster resonance energy transfer (FRET) experiments at the single molecule level, fluorescent dyes must be coupled to the RNA molecule[7].
Recently a deep neural network architecture designed to work on graph- structured data have been capturing notice as well as getting implemented in various domains and application. However, learning representation (feature embedding) from graphical data picking pace in research and constructing graph(s) from dataset remains a challenge. The ability to map the data to lower dimensions further makes the task easier while providing comfort in applying many operations. Graph neural network (GNN) is one of the novel neural network models that is catching attention as it is outperforming in various applications like recommender systems, social networks, chemical synthesis, and many more. This thesis discusses a unique approach for a fundamental task on graphs; node classification. The feature embedding for a node is aggregated by applying a Recurrent neural network (RNN), then a GNN model is trained to classify a node with the help of aggregated features and Q learning supports in optimizing the shape of neural networks. This thesis starts with the working principles of the Feedforward neural network, recurrent units like simple RNN, Long short-term memory (LSTM), and Gated recurrent unit (GRU), followed by concepts of Reinforcement learning (RL) and the Q learning algorithm. An overview of the fundamentals of graphs, followed by the GNN architecture and workflow, is discussed subsequently. Some basic GNN models are discussed in brief later before it approaches the technical implementation details, the output of the model, and a comparison with a few other models such as GraphSage and Graph attention network (GAN).
In the past few years, social media has become the most popular communication software, replacing phone calls, text messages, television and even advertisements. Social media has become the most important channel for spreading opinions. As a result of this trend, many politicians have also started to operate social media (Wang, Tsai, & Chen 2019). This study was conducted in order to understand whether there was an intercandidate agenda-setting effect between the Facebook posts of legislative candidates and presidential candidates during the election period, and whether the legislative candidates' Facebook posts were influenced by the presidential candidates' Facebook posts. The target population of this study was the three presidential candidates in Taiwan's 2020 presidential election — Dr. Tsai Ing-Wen, Mr. Han Kuo-Yu, and Mr. James Soong — as well as the 36 legislative candidates in Taipei, Taichung, and Kaohsiung.
The study focused on Facebook posts from 1thNovember 2019 to 10th January 2020, 10 weeks before the voting day. Text-mining and cosine similarity were used to organize the posts and compare the similarity between posts. Finally, the similarity between posts was presented as a line graph.
The study revealed that there was an inter-candidate agenda-setting effect between legislative candidate posts and presidential candidate posts, and that Dr. Tsai Ing-Wen, who was also the incumbent president during the campaign, was the most influential Facebook poster during the entire election.
Future research is proposed on the inter-candidate agenda-setting effect only analyzing the similarity of posts among the candidates to discuss the influence of the candidates' Facebook agenda-setting during a specific election period.
This is the first study in which the Facebook posts of Taiwanese politicians are analyzed and the relationships were analyzed and the relationships were systematically compared, across multiple degrees, which opens up a whole new subject for future elections in Taiwan.