Achtung: Die Navigationslinks auf dieser Seite funktionieren nicht, weil das nicht der Rahmen ist, der zu diesem Skript gehört. Wenn das Skript ordnungsgemäß dort installiert ist, wo es später laufen soll, funktionieren auch die Links in der dortigen Umgebung.
LINUX Cluster Project
- Name: Lehrstuhl für Biologische Bildgebung
- Address: Ismaninger Straße 22, 81675 München
- Project Proposal Date: 2017-01-23 17:17:16
Our research focuses on innovative optical and optoacoustic methods that enable novel visualization of living biological tissue. To this end, we develop and evaluate state-of-the-art image reconstruction methods for tomographic problems and raster-scanning systems. From simulations of these systems we continuously improve our hardware and optimize our algorithms. Simulations of optoacoustic signals mostly involves simulation of diffuse radiation transfer by means of Monte-Carlo based simulations and finite element methods. Moreover, the acoustic part deals with simulation of linear and non-linear acoustic wave equations. Reconstruction of multi-spectral images from optoacoustic signals involves solving inverse problems with various constraints and regularization terms. Since the acoustic signals are sampled at high bandwidth and the images are reconstructed at high-resolution (e.g. 1600x1600), matrices in the inverse problems become quite large often exceeding the memory of desktop hardware (>8Gb), even if their sparseness is leveraged. The same argument applies for simulations, where in addition the computing times are limiting or, in case of Monta-Carlo methods, leveraging parallel computations are useful. As a third application, for training of machine-learning models using Bayesian and Deep Learning methods access to the Linux cluster will be essential for us.