During the first week of October took place IEEE Vis 2017 in Phoenix, AZ.
Kitware supported this event and was involved in at least 3 presentations, 2 of them took place at LDAV (Large Scale Data Analysis and Visualization) symposium:
- Task-based Augmented Merge Trees with Fibonacci Heaps at LDAV, by Charles Gueunet (KW), Julien Tierny, Pierre Fortin and Julien Jomier (KW).
This paper presented a new algorithm to compute the augmented merge tree and the augmented contour tree on a shared memory workstation. The approach relies on local propagations managed by tasks and offers good performances both sequentially and in parallel.
- Parallel Multi-Level Ghost Cell Generation for Distributed Unstructured Grids at LDAV, by John M. Patchett, Boonthanome Nouanesengesy, Joachim Pouderoux (KW), James Ahrens, Hans Hagen.
This paper presented a efficent parallel algorithm to compute one or multiple levels of ghost cells on a parallel unstructured grid. The described algorithm was added to VTK and ParaView.
- The Topology Toolkit [Best Paper Honorable Mention] in IEEE Vis, by Julien Tierny, Guillaume Favelier, Jushua A. Levine, Charles Gueunet (KW), Michael Michaux.
This technical report presented a new open-source platform to compute topological abstraction based on VTK and Paraview. It will be detailed in a future article.
Otherwise, this year machine learning seemed to be a hot topic. Involved in several panels, the main subjects were the role of the human and the use of visualization to understand, train or debug the model.