Refine
Year of publication
Document Type
- Master's Thesis (127) (remove)
Has Fulltext
- yes (127)
Keywords
- Beschleunigerphysik (2)
- E-Learning (2)
- Erwachsenenbildung (2)
- Experiment (2)
- Ionenstrahl (2)
- Konstruktion (2)
- Lehren mit neuen Medien (2)
- Lehrerbildung (2)
- Lernen mit neuen Medien (2)
- Machine Learning (2)
Institute
- Physik (61)
- Gesellschaftswissenschaften (11)
- Informatik und Mathematik (10)
- Mathematik (7)
- Informatik (6)
- Geowissenschaften (4)
- Erziehungswissenschaften (3)
- Neuere Philologien (3)
- Sportwissenschaften (3)
- Psychologie (2)
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).
The thesis delves into two research questions, drawing from the 2022 Final Report of the Colombian Truth Commission. It critically examines historical power differentials originating from the colonial period, probing their role in legitimizing and perpetuating violence against ethnic groups, women, and individuals diverging from cisgender or heteronormative norms within Colombian society. Additionally, the research explores the pervasive influence exerted by the colonial legacy on the foundational structures of Colombian societal organization. Employing discourse analysis as its methodological approach, the thesis undertakes the task of deconstructing and reconstructing the Report, thereby elucidating emergent and contingent discursive meanings that situate coloniality within the realms of cognition, language, and affect. Emphasizing the presence of counter-hegemonic knowledge within the Report, the thesis integrates its findings into a robust theoretical and conceptual framework, facilitating a nuanced and systematic comprehension of the underlying causes of violence perpetrated against marginalized groups and the environment. These causes are intricately linked to the intertwined and hybridized power structures that have endured since the colonial era.
Power structures of Eurocentric origin, alongside mental constructs imposed by European invaders over centuries — such as anthropocentrism, racism, internal colonialism, heteropatriarchy, cisnormativity, and classism — were gradually naturalized and institutionalized within Colombian society. This process has been perpetuated through the state's reproduction of these patterns since the inception of Colombia as a nation-state. Consequently, hierarchical discursive constructions, posited as universal and self-evident, have marginalized certain groups and justified environmental degradation. While the internal armed conflict exacerbated these issues, it did not create them; rather, it intensified pre-existing violences, targeting individuals, communities, and their territories.
The thesis also underscores the portrayal of otherness within the modern-colonial world-system, rooted in principles of domination and subalternity, perpetuating colonial patterns of thought and action, reinforcing the hegemonic cosmovision. Notions of superiority and inferiority that predate the establishment of the Colombian nation-state have influenced social categories, subject positions, and identities, resulting in disproportionate, differential, and cumulative harm inflicted upon subalternized population groups. This contributes to a culture of 'justified' violence.
The research underscores the profound entrenchment of coloniality in the structures of Colombian society and the various inherent logics of violence within its conflicts. Coloniality, a comprehensive framework encompassing colonial patterns of thought and action, originating from the invasion and conquest of Abya Yala, continuously shapes the contemporary realities of societies in diverse (re-)configurations, leaving indelible imprints. The effects of these dynamics are manifold, ranging from the imperative of monogamous sexuality according to Judeo-Christian principles as a self-evident norm to the view of nature as a resource rather than a unit comprising both people and environment. In other words, colonial patterns are deeply embedded in all structures of society. A key recommendation emerging from this thesis is to underscore the imperative to recognize and question the persistence of colonial patterns in social and individual lives.
The research urges recognition and interrogation of these persistent colonial patterns in societal and individual structures, advocating for transformative paradigms that challenge conventional thought patterns and foster self-reflection among Colombians. The report, functioning as a political instrument, holds the potential to significantly contribute to the formation of subjectivities that break away from the epistemic schemes of modernity/coloniality. The research and its findings create a political space for questioning the universalist notion of the Eurocentric civilizing project, the scientific rationality of the universal subject, and the presumed neutrality of its forms of knowledge. This opens avenues for questioning, disputing, and transforming entrenched paradigms.
Ausgehend von Scheffers Beschreibung Existentieller Probleme untersucht die Masterarbeit die Problembehandlung solcher. Dabei versucht sich der Autor an einer vorläufigen Begriffsdefinition Existentieller Apparate und untersucht im folgenden, wie sich diese von der Arbeitsweise, ihrer Struktur und anderer Merkmale von herkömmlichen Apparaten unterscheiden. Die Diskussion erfolgt anhand der Problembearbeitungskapazitäten von der Flüchtlingshilfe auf Lesbos, Impfzentren und der Fridays-for-Future-Bewegung, die jeweils für Sich genommen speziell nur im Kontext jeweiliger Existentieller Probleme existieren.
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.
Die Arbeit befasst sich mit einer Vereinfachung des von Devroye (1999) geprägten Begriffs der random split trees und verallgemeinert diesen im Sinne von Janson (2019) auf unbeschränkten Verzweigungsgrad. Diese Verallgemeinerung deckt auch preferential attachment trees mit linearen Gewichten ab, wofür ein Beweis von Janson (2019) aufbereitet wird. Zusätzlich bleiben die von Devroye (1999) nachgewiesenen Eigenschaften über die Tiefe der hinzugefügten Knoten erhalten.