Among other topics this paper describes a novel method for selection of an appropriate number of support nodes for each point of the meshless MLS approximation and ensuring that the approximation is contiguous. Improved stability of the new method in comparison with known methods was proven, especially in case of varying nodal density. The rest of the paper deals with the use of such MLS approximation in partial differential equation solvers.
COBISS.SI-ID: 21305383
The paper analyzes in detail the computational complexity of MLS approximation as well as the complexity of other steps of meshless solution of partial differential equations, where MLS is also used often. We have also proven that all the steps of the approximation can be efficiently parallelized using spatial decomposition, which is also applicable to landscape visualization.
COBISS.SI-ID: 21895463
At the Cracow Grid Workshop we had a presentation about landscape visualization. Although the presentation was about the preliminary results from the period before this post-doc project, the paper preparation for this conference is traditionally after the conference (the proceedings were published in summer 2007). The published paper thus includes the results of this project, namely distributed visualization data processing on grid systems in real time.
COBISS.SI-ID: 21004327
At the conference we showed a live demonstration of the open-source visualization application that was developed within this project. We have shown parallel use of data from multiple servers and automatic switch to an alternative server in case of server outage. The paper, which is published in the conference proceedings, also includes a comparison of various algorithms for scheduling the data requests among the servers.
COBISS.SI-ID: 21770023