Markus Krimmel

Generative Geometric Deep Learning aims to revolutionize the field of machine learning by leveraging the power of deep neural networks and geometric data. This PhD project will explore the development of novel generative models capable of capturing complex geometric patterns and structures inherent in various domains, such as computer vision, molecular biology and computational chemistry. By combining the principles of geometric deep learning and generative modeling, we seek to address the challenges associated with generating realistic and meaningful structured data, such as molecules, proteins or general graph-structured data. This includes tasks such as 3D shape generation, and molecular and protein design. Key research areas within this project will encompass: - Geometric deep learning: Investigating effective methods for representing geometric data, including point clouds, meshes, and graphs, to ensure that the deep neural networks can effectively learn and process their underlying patterns. Generative model development: Designing and implementing novel generative models, such as diffusion models and Bayesian networks, tailored for geometric data. This involves exploring techniques to capture the complex dependencies and topological structures present in geometric objects. - Applications and evaluation: Applying the developed generative models to a wide range of real-world applications, including computer vision, molecular biology, and drug discovery. Evaluating the performance of the models using appropriate benchmarks. The ultimate goal of this PhD project is to advance the state-of-the-art in generative modeling for geometric data, enabling new breakthroughs in various fields and opening up exciting possibilities for future research and applications.