Title: Machine Learning for Visualization of Scientific Data and Simulations
Name: Eric Kinner
Phone: +49 631 205 237
Scientific simulations are an important tool for researchers of various domains to explore, verify, and adjust their hypothesis. The scale of these simulations have increased over the last decade. More and bigger simulations enable researchers to study their subject at ever increasing scales. This trend, however, creates new challenges, which need to be solved to avoid hindering scientific discovery. Among these is the ever increasing demand for storage space. The handling of massive amounts of data in processing, visualization, and the analysis of results needs new solutions to adapt to a new scale. Machine learning has the potential to overcome some of these challenges and may be a vital part of scientific data processing and visualization in the future.
While machine learning has gained a lot of attention in the last years, there is a distinct lack of machine learning approaches applied to scientific data processing. The outstanding accuracies may also be applicable in engineering and other domains. The current state of the project focuses on the embedding of streamlines. Streamlines are used in the visualization of vectorfields and are commonly found in simulations containing some kind of (air)flow like in the design of cars, ships, or planes.
The embedding of streamlines using machine learning may yield different benefits. Depending on the dimensionality of the embedding it may be possible to compress streamlines in that way and reduce the storage space required. A more versatile result would be the classification of streamlines inside the embedding. Guided touring and automated selection of suitable streamlines for visualization, storage, etc. could be determined.