Master's Thesis
Refine
Year of publication
Document Type
- Master's Thesis (90) (remove)
Language
- English (90) (remove)
Has Fulltext
- yes (90) (remove)
Keywords
Constituent parsing attempts to extract syntactic structure from a sentence. These parsing systems are helpful in many NLP applications such as grammar checking, question answering, and information extraction. This thesis work is about implementing a constituent parser for German language using neural networks. Over the past, recurrent neural networks have been used in building a parser and also many NLP applications. In this, self-attention neural network modules are used intensively to understand sentences effectively. With multilayered self-attention networks, constituent parsing achieves 93.68% F1 score. This is improved even further by using both character and word embeddings as a representation of the input. An F1 score of 94.10% was the best achieved by constituent parser using only the dataset provided. With the help of external datasets such as German Wikipedia, pre-trained ELMo models are used along with self-attention networks achieving 95.87% F1 score.
Since the invention of U-net architecture in 2015, convolutional networks based on its encoder-decoder approach significantly improved results in image analysis challenges. It has been proven that such architectures can also be successfully applied in different domains by winning numerous championships in recent years. Also, the transfer learning technique created an opportunity to push state-of-the-art benchmarks to a higher level. Using this approach is beneficial for the medical domain, as collecting datasets is generally a difficult and expensive process.
In this thesis, we address the task of semantic segmentation with Deep Learning and make three main contributions and release experimental results that have practical value for medical imaging.
First, we evaluate the performance of four neural network architectures on the dataset of the cervical spine MRI scans. Second, we use transfer learning from models trained on the Imagenet dataset and compare it to randomly initialized networks. Third, we evaluate models trained on the bias field corrected and raw MRI data. All code to reproduce results is publicly available online.
The industry standard Decision Model and Notation (DMN) has enabled a new way for the formalization of business rules since 2015. Here, rules are modeled in so-called decision tables, which are defined by input columns and output columns. Furthermore, decisions are arranged in a graph-like structure (DRD level), which creates dependencies between them. With a given input, the decisions now can be requested by appropriate systems. Thereby, activated rules produce output for future use. However, modeling mistakes produces erroneous models, which can occur in the decision tables as well as at the DRD level. According to the Design Science Research Methodology, this thesis introduces an implementation of a verification prototype for the detection and resolution of these errors while the modeling phase. Therefore, presented basics provide the needed theoretical foundation for the development of the tool. This thesis further presents the architecture of the tool and the implemented verification capabilities. Finally, the created prototype is evaluated.
The Material Point Method (MPM) has proven to be a very capable simulation method in computer graphics that is able to model materials that were previously very challenging to animate [1, 2]. Apart from simulating singular materials, the simulation of multiple materials that interact with each other introduces new challenges. This is the focus of this thesis. It will be shown that the self-collision capabilities of the MPM can naturally handle multiple materials interacting in the same scene on a collision basis, even if the materials use distinct constitutive models. This is then extended by porous interaction of materials as in[3], which also integrates easily with MPM.It will furthermore be shown that regular single-grid MPM can be viewed as a subset of this multi-grid approach, meaning that its behavior can also be achieved if multiple grids are used. The porous interaction is generalized to arbitrary materials and freely changeable material interaction terms, yielding a flexible, user-controllable framework that is independent of specific constitutive models. The framework is implemented on the GPU in a straightforward and simple way and takes advantage of the rasterization pipeline to resolve write-conflicts, resulting in a portable implementation with wide hardware support, unlike other approaches such as [4].
Blockchain in Healthcare
(2020)
The underlying characteristics of blockchain can facilitate data provenance, data integrity, data security, and data management. It has the potential to transform the healthcare sector. Since the introduction of Bitcoin in the fintech industry, the blcockhain technology has been gaining a lot of traction and its purpose is not just limited to finance. This thesis highlights the inner workings of blockchain technology and its application areas with possible existing solutions. Blockchain could lay the path for a new revolution in conventional healthcare systems. We presented how individual sectors within the healthcare industry could use blockchain and what solution persists. Also, we have presented our own concept to improve the existing paper-based prescription management system which is based on Hyperledger framework. The results of this work suggest that healthcare can benefit from blockchain technology bringing in the new ways patients can be treated.
On-screen interactive presentations have got immense popularity in the domain of attentive interfaces recently. These attentive screens adapt their behavior according to the user's visual attention. This thesis aims to introduce an application that would enable these attentive interfaces to change their behavior not just according to the gaze data but also facial features and expressions. The modern era requires new ways of communications and publications for advertisement. These ads need to be more specific according to people's interests, age, and gender. When advertising, it's important to get a reaction from the user but not every user is interested in providing feedback. In such a context more, advance techniques are required that would collect user's feedback effortlessly. The main problem this thesis intends to resolve is, to apply advanced techniques of gaze and face recognition to collect data about user's reactions towards different ads being played on interactive screens. We aim to create an application that enables attentive screens to detect a person's facial features, expressions, and eye gaze. With eye gaze data we can determine the interests and with facial features, age and gender can be specified. All this information will help in optimizing the advertisements.
In this thesis, the performance of the IceCube projects photon propagation
code (clsim) is optimized. The process of GPU code analysis and perfor-
mance optimization is described in detail. When run on the same hard-
ware, the new version achieves a speedup of about 3x over the original
implementation. Comparing the unmodified code on hardware currently
used by IceCube (NVIDIA GTX 1080) against the optimized version run on
a recent GPU (NVIDIA A100) a speedup of about 9.23x is observed. All
changes made to the code are shown and their performance impact as well
as the implications for simulation accuracy are discussed individually.
The approach taken for optimization is then generalized into a recipe.
Programmers can use it as a guide, when approaching large and complex
GPU programs. In addition, the per warp job-queue, a design pattern used
for load balancing among threads in a CUDA thread block, is discussed in
detail.
Advanced Auditing of Inconsistencies in Declarative Process Models using Clustering Algorithms
(2021)
To have a compliant business process of an organization, it is essential to ensure a onsistent process. The measure of checking if a process is consistent or not depends on the business rules of a process. If the process adheres to these business rules, then the process is compliant and efficient. For huge processes, this is quite a challenge. Having an inconsistency in a process can yield very quickly to a non-functional process, and that’s a severe problem for organizations. This thesis presents a novel auditing approach for handling inconsistencies from a post-execution perspective. The tool identifies the run-time inconsistencies and visualizes them in heatmaps. These plots aim to help modelers observe the most problematic constraints and help them make the right remodeling decisions. The modelers assisted with many variables can be set in the tool to see a different representation of heatmaps that help grasp all the perspectives of the problem. The heatmap sort and shows the run-time inconsistency patterns, so that modeler can decide which constraints are highly problematic and should address a re-model. The tool can be applied to real-life data sets in a reasonable run-time.
This thesis focuses on approximate inference in assumption-based argumentation frameworks. Argumentation provides a significant idea in the computerization of theoretical and practical reasoning in AI. And it has a close connection with AI, engaging in arguments to perform scientific reasoning. The fundamental approach in this field is abstract argumentation frameworks developed by Dung. Assumption-based argumentation can be regarded as an instance of abstract argumentation with structured arguments. When facing a large scale of data, a challenge of reasoning in assumption-based argumentation is how to construct arguments and resolve attacks over a given claim with minimal cost of computation and acceptable accuracy at the same time. This thesis proposes and investigates approximate methods that randomly select and construct samples of frameworks based on graphical dispute derivations to solve this problem. The presented approach aims to improve reasoning performance and get an acceptable trade-off between computational time and accuracy. The evaluation shows that for reasoning in assumption-based argumentation, in general, the running time is reduced with the cost of slightly low accuracy by randomly sampling and constructing inference rules for potential arguments over a query.
In this thesis the possibilities for real-time visualization of OpenVDB
files are investigated. The basics of OpenVDB, its possibilities, as well
as NanoVDB and its GPU port, were studied. A system was developed
using PNanoVDB, the graphics API port of OpenVDB. Techniques were
explored to improve and accelerate a single ray approach of ray tracing.
To prove real-time capability, two single scattering approaches were
also implemented. One of these was selected, further investigated and
optimized to achieve interactive real-time rendering.
It is important to give artists immediate feedback on their adjustments, as
well as the possibility to change all parameters to ensure a user friendly
creation process.
In addition to the optical rendering, corresponding benchmarks were
collected to compare different improvement approaches and to prove
their relevance. Attention was paid to the rendering times and memory
consumption on the GPU to ensure optimal use. A special focus, when
rendering OpenVDB files, was put on the integrability and extensibility of
the program to allow easy integration into an existing real-time renderer
like U-Render.