Refine
Year of publication
Document Type
- Master's Thesis (42) (remove)
Language
- English (42) (remove)
Has Fulltext
- yes (42)
Is part of the Bibliography
- no (42) (remove)
Keywords
- WaterGAP (2)
- global water model (2)
- (n (1)
- AI Safety (1)
- ALICE (1)
- Activism (1)
- App ecosystem (1)
- Autonomous Driving (1)
- Bayesian Inference (1)
- Beryllium-7 (1)
Institute
Analysis of machine learning prediction quality for automated subgroups within the MIMIC III dataset
(2023)
The motivation for this master’s thesis is to explore the potential of predictive data analytics in the field of medicine. For this, the MIMIC-III dataset offers an extensive foundation for the construction of prediction models, including Random Forest, XGBOOST, and deep learning networks. These models were implemented to forecast the mortality of 2,655 stroke patients.
The first part of the thesis involved conducting a comprehensive data analysis of the filtered MIMIC-III dataset.
Subsequently, the effectiveness and fairness of the predictive models were evaluated. Although the performance levels of the developed models did not match those reported in related research, their potential became evident. The results obtained demonstrated promising capabilities and highlighted the effectiveness of the applied methodologies. Moreover, the feature relevance within the XGBOOST model was examined to increase model explainability.
Finally, relevant subgroups were identified to perform a comparative analysis of the prediction performance across these subgroups. While this approach can be regarded as a valuable methodology, it was not possible to investigate underlying reasons for potential unfairness across clusters. Inside the test data, not enough instances remained per subgroup for further fairness or feature relevance analysis.
In conclusion, the implementation of an alternative use case with a higher patient count is recommended.
The code for this analysis is made available via a GitHub repository and includes a frontend to visualize the results.
Goal-Conditioned Reinforcement Learning (GCRL) is a popular framework for training agents to solve multiple tasks in a single environment. It is cru- cial to train an agent on a diverse set of goals to ensure that it can learn to generalize to unseen downstream goals. Therefore, current algorithms try to learn to reach goals while simultaneously exploring the environment for new ones (Aubret et al., 2021; Mendonca et al., 2021). This creates a form of the prominent exploration-exploitation dilemma. To relieve the pres- sure of a single agent having to optimize for two competing objectives at once, this thesis proposes the novel algorithm family Goal-Conditioned Re- inforcement Learning with Prior Intrinsic Exploration (GC-π), which sep- arates exploration and goal learning into distinct phases. In the first ex- ploration phase, an intrinsically motivated agent explores the environment and collects a rich dataset of states and actions. This dataset is then used to learn a representation space, which acts as the distance metric for the goal- conditioned reward signal. In the final phase, a goal-conditioned policy is trained with the help of the representation space, and its training goals are randomly sampled from the dataset collected during the exploration phase. Multiple variations of these three phases have been extensively evaluated in the classic AntMaze MuJoCo environment (Nachum et al., 2018). The fi- nal results show that the proposed algorithms are able to fully explore the environment and solve all downstream goals while using every dimension of the state space for the goal space. This makes the approach more flexible compared to previous GCRL work, which only ever uses a small subset of the dimensions for the goals (S. Li et al., 2021a; Pong et al., 2020).
WaterGAP (Water - Global Assessment and Prognosis) is a tool for modeling global water use and water availability. It participates among other models in the ISIMIP initiative (The Inter-Sectoral Impact Model Intercomparison Project). As part of this initiative, the water temperature should be calculated by participating hydrological models because it plays a vital role in many chemical, physical and biological processes. Therefore, the subject of this master thesis is to implement the physically based surface water temperature computation after VAN BEEK ET AL. (2012) and WANDERS ET AL. (2019) into WaterGAP and compare the results to the statistical regression approach by PUNZET ET AL. (2012). The computation is validated with observed water temperature data obtained from the GEMStat water quality database. The results are good for arctic and temperate latitudes. Surface water temperatures for tropical rivers are overestimated, most likely due to the overestimation of precipitation temperatures, incoming radiation and groundwater temperatures. The comparison with the regression model by PUNZET ET AL. (2012) shows matching results. The regression model even matches with WaterGAP results for most of the simulations of the future under climate change conditions, where the regression model should stop working due to changing environmental parameters. Several assumptions had to be made in order to implement the water temperature calculation in Water-GAP. These include, e.g., discharge temperatures for power plant cooling water, precipitation and surface runoff temperatures. For model improvements, perhaps three different values for the different regions of the world should be used to cool down the precipitation and surface runoff. The model could also be improved by refining the ice formation calculation, especially for the conditions when the ice melts, breaks up and is transported downstream. Furthermore, the feedback to the river channel roughness could be implemented if ice has formed. The WaterGAP model upgraded with the water temperature calculation will help the ISIMIP initiative in the future.
The reanalysis products and derived products, ERA5 (Copernicus Climate Change Service, 2018) and W5E5 (WATCH Forcing Data (WFD) methodology applied to ERA5) (LANGE ET AL., 2021) have been recently published initiating a new phase of scientific research utilizing these datasets. ERA5 and W5E5 offer the possibility to reduce insecurities in model results through their improved quality compared to previous climate reanalyses (CUCCHI ET AL., 2020). The suitability of either climate forcing as input for the hydrological model WaterGAP and the influence of the models specific calibration routine has been evaluated with four model experiments. The model was validated by analysing the models ability to produce reasonable values for global water balance components and to reproduce observed discharge in 1427 basins as well as total water storage anomalies in 143 basins using well established efficiency metrics. Bias correction of W5E5 was found to lead to more global realistic mean precipitation and consequently discharge and AET values. In an uncalibrated model setup ERA5 results in better performances across all efficiency metrics. Model results produced with W5E5 as climate input were strongly improved through calibration ultimately leading to the best performances out of all four model experiments. However, model performances considerably improved through calibration with both climate forcings hence calibration was found to have the strongest effect on model performance. Furthermore, spatial differences in performance of either forcing were identified. Snow-dominated regions show an overall better performance with ERA5, while wetter and warmer regions are better represented with W5E5. Finally, it can be concluded that W5E5 should be preferred as climate input for impact modelling; however, depending on the spatial scale and region ERA5 should at least be considered, in particular for snow-dominated regions.
We give theorems about asymptotic normality of general additive functionals on patricia tries, derived from results on tries. These theorems are applied to show asymptotic normality of the distribution of random fringe trees in patricia tries. Formulas for asymptotic mean and variance are given. The proportion of fringe trees with 𝑘 keys is asymptotically, ignoring oscillations, given by (1−𝜌(𝑘))/(𝐻 +𝐽)𝑘(𝑘−1) with the source entropy 𝐻, an entropy-like constant 𝐽, that is 𝐻 in the binary case, and an exponentially decreasing function 𝜌(𝑘). Another application gives asymptotic normality of the independence number and the number of 𝑘-protected nodes.
AI-based computer vision systems play a crucial role in the environment perception for autonomous driving. Although the development of self-driving systems has been pursued for multiple decades, it is only recently that breakthroughs in Deep Neural Networks (DNNs) have led to their widespread application in perception pipelines, which are getting more and more sophisticated. However, with this rising trend comes the need for a systematic safety analysis to evaluate the DNN's behavior in difficult scenarios as well as to identify the various factors that cause misbehavior in such systems. This work aims to deliver a crucial contribution to the lacking literature on the systematic analysis of Performance Limiting Factors (PLFs) for DNNs by investigating the task of pedestrian detection in urban traffic from a monocular camera mounted on an autonomous vehicle. To investigate the common factors that lead to DNN misbehavior, six commonly used state-of-the-art object detection architectures and three detection tasks are studied using a new large-scale synthetic dataset and a smaller real-world dataset for pedestrian detection. The systematic analysis includes 17 factors from the literature and four novel factors that are introduced as part of this work. Each of the 21 factors is assessed based on its influence on the detection performance and whether it can be considered a Performance Limiting Factor (PLF). In order to support the evaluation of the detection performance, a novel and task-oriented Pedestrian Detection Safety Metric (PDSM) is introduced, which is specifically designed to aid in the identification of individual factors that contribute to DNN failure. This work further introduces a training approach for F1-Score maximization whose purpose is to ensure that the DNNs are assessed at their highest performance. Moreover, a new occlusion estimation model is introduced to replace the missing pedestrian occlusion annotations in the real-world dataset. Based on a qualitative analysis of the correlation graphs that visualize the correlation between the PLFs and the detection performance, this study identified 16 of the initial 21 factors as being PLFs for DNNs out of which the entropy, the occlusion ratio, the boundary edge strength, and the bounding box aspect ratio turned out to be most severely affecting the detection performance. The findings of this study highlight some of the most serious shortcomings of current DNNs and pave the way for future research to address these issues.
Statistical shape models learn to capture the most characteristic geometric variations of anatomical structures given samples from their population. Accordingly, shape models have become an essential tool for many medical applications and are used in, for example, shape generation, reconstruction, and classification tasks. However, established statistical shape models require precomputed dense correspondence between shapes, often lack robustness, and ignore the global surface topology. This thesis presents a novel neural flow-based shape model that does not require any precomputed correspondence. The proposed model relies on continuous flows of a neural ordinary differential equation to model shapes as deformations of a template. To increase the expressivity of the neural flow and disentangle global, low-frequency deformations from the generation of local, high- frequency details, we propose to apply a hierarchy of flows. We evaluate the performance of our model on two anatomical structures, liver, and distal femur. Our model outperforms state-of-the-art methods in providing an expressive and robust shape prior, as indicated by its generalization ability and specificity. More so, we demonstrate the effectiveness of our shape model on shape reconstruction tasks and find anatomically plausible solutions. Finally, we assess the quality of the emerging shape representation in an unsupervised setting and discriminate healthy from pathological shapes.
Electron identification with a likelihood method and measurements of di-electrons for the CBM-TRD
(2017)
In this work a likelihood method has been implemented and investigated as particle identification algorithm for the CBM-TRD.
The creation of the probability distributions for the likelihood method via V0-topologies seems to be feasible and the purity of the obtained samples is sufficient for the usage in the likelihood method.
The comparison between the ANN and the likelihood method shows no differences in the identification performance. The pion suppression factor reaches the same values for the same electron identification efficiencies and the yields of the resulting di-lepton signals are comparable. The signal-to-background ratios for both methods have the same values and show a value of about 10−2 in the invariant mass range of minv = 1.5 - 2.5 GeV/c2, which is expected to be sufficient to provide access to the thermal in-medium and QGP radiation.
The investigation of a detector system without a TRD shows no pion suppression for a momentum above p = 6 GeV/c. Therefore, the background contributions increase drastically and the signal-to-background ratio decreases at all invariant masses, but especially in the invariant mass range of minv = 1.5 - 2.5 GeV/c2.
The background contributions in the invariant mass range of minv = 1.5 - 2.5 GeV/c 2 are also influenced by the selected electron identification efficiency of the TRD, which significantly shifts the fraction of the eπ contributions relative to the total number of pairs.
Anisotropic collective flow of protons resulting from non-central heavy ion collisions is a unique hadronic observable providing information about the early stage of the nuclear collision. The analysis of collective flow in the energy regime between 1-2 AGeV enables the study of the phase diagram of hadronic matter at a high baryochemical potential µb, as well as the analysis of the equation of state at densities up to the threefold of the ground state density ρ0.
The algorithms of the standard event plane method and the scalar product method are used to analyse directed and elliptic flow of protons in a centrality range of 0-40 % most central events.
Prior to the analysis of experimental data, the respective influence of the reconstruction procedure on the algorithms is examined using Monte Carlo simulations based on the Ultra relativistic Quantum Molecular Dynamics (UrQMD) model.
Subsequently, experimental data measured in April 2012 with the High Acceptance DiElectron Spectrometer (HADES) is analysed using both methods. About 7.3 · 109 Au+Au events at a kinetic beam energy of 1.23 AGeV, equivalent to a centre of mass energy of √sNN = 2.42 GeV were recorded. A multi-differential analysis is feasible as the HADES detector provides a good transverse momentum and rapidity coverage.
Both algorithms result in identical values for directed and elliptic flow across all centrality classes within the observable phase space of protons. The calculated integrated value of v2 at mid rapidity is in good agreement with world data.
In April and May 2012 data on Au+Au collisions at beam energies of Ekin = 1.23A GeV were collected with the High Acceptance Di-Electron Spectrometer (HADES) at the GSI Helmholtzzentrum für Schwerionenforschung facility in Darmstadt, Germany. In this thesis, the production of deuterons in this collision system is investigated.
A total number of 2.1 × 109 Au+Au events is selected, containing the most central 0-40% of events. After particle identification, based on a mass determination via time-of-flight and momentum and on a measurement of the energy loss, the transverse mass spectra of the deuteron candidates are extracted for various rapidities and subsequently corrected for acceptance and efficiency.
The inverse slope parameter of a Boltzmann fit applied to the transverse mass spectra at midrapidity, which is referred to as the effective temperature, is extracted. For a static thermal source, this parameter corresponds to the kinetic freeze-out temperature Tkin and is therefore expected to be smaller or equal to the chemical freeze-out temperature Tchem. The extracted effective temperature of Tef f = (190 ± 10) MeV however exceeds the chemical freeze-out temperature that was obtained by a statistical model fit to different particle yields. The effective temperatures of various particle species, obtained in previous analyses, suggest a systematic rise with increasing particle mass, which is confirmed by the deuteron results.
An explanation can be the influence of a collective expansion with a radial expansion velocity βr. By fitting a Siemens-Rasmussen function to the transverse mass spectra, the global temperature of T = (100 ± 8) MeV and radial expansion velocity βr = 0.37 ± 0.01 are obtained. This temperature is still very high and only takes into account the production of deuteron nuclei.
The simultaneous fit of a blast-wave function to the transverse mass spectra of deuterons and other particles, as obtained by previous analyses, considers a velocity profile for the radial expansion velocity and takes into account the production of various particle species. The resulting global temperature Tkin = (68 ± 1) MeV and average transverse expansion velocity hβri = 0.341 ± 0.003 are within the expected range for the collision energy.
The Siemens-Rasmussen fits are also used to extrapolate the transverse mass spectra into unmeasured regions, to integrate them and obtain a rapidity-dependent count rate. This count rate exhibits a thermal shape for central events and shows increasing spectator contributions for more peripheral events.
The invariant yield spectra of the deuterons are compared to those of protons, as obtained by a previous analysis, in the context of a nucleon coalescence model. The hereby extracted nucleon coalescence factor B2 = (4.6 ± 0.1) × 10−3 agrees with the expected result for the beam energy that was studied.