Date(s) - 18/02/2019
11:00 am - 12:00 pm
Studio Villa Bosch
By Prof. Dr. Carsten Rother, Head of Visual Learning Lab Heidelberg
In this talk I will introduce the Visual Learning Lab Heidelberg. We conduct basic research in the fields of machine learning and combinatorial optimization with application to image analysis, microscopy, astrophysics, medicine, and other fields in Natural- and Lifescience. In this talk I will present a collection of our research projects. One of the projects is to solve inverse problems with a new form of so-called invertible neural networks, which allows a bijective mapping between parameter-space and measurement-space. The beautify of this approach is that a full distribution of the parameter-space can be reconstructed without any assumptions about its shape. Another project shows how we can learn to generate synthetic labeled data, that can be used as a proxy for real-world annotated data, which is often lacking in practical scenarios. A final project demonstrates how to combine neural networks with classical sampling-based algorithms, in order to achieve improved task-specific performance.
For registration please contact Benedicta Frech: firstname.lastname@example.org