http://swrc.ontoware.org/ontology#InProceedings
POD-Based Parallel Compression for Visualizing Large-Scale Dataset
en
ポスターセッション： システム・可視化・性能評価
Advanced Institute for Computational Science, RIKEN
Advanced Institute for Computational Science, RIKEN
Chongke Bi
Kenji Ono
Visualizing and analyzing large-scale dataset is an important task for scientific research in various fields. However, the visualization process is time-consuming, which is quite inconvenient for researchers and engineers to analyze the time-varying dataset. In this poster,we proposed an approach to generate a small-scale dataset from the original large-scale one. The key idea is to divide the large-scale dataset into small groups firstly, and then compress the dataset in each group by using proper orthogonal decomposition (POD) method in parallel. This process is recursively carried on until the obtained dataset cannot be compressed any more. Here, the parallel computing greatly decreases the computational cost of the eigen resolving problem in the POD algorithm. Furthermore the compressed dataset can be easily restored linearly.
Visualizing and analyzing large-scale dataset is an important task for scientific research in various fields. However, the visualization process is time-consuming, which is quite inconvenient for researchers and engineers to analyze the time-varying dataset. In this poster,we proposed an approach to generate a small-scale dataset from the original large-scale one. The key idea is to divide the large-scale dataset into small groups firstly, and then compress the dataset in each group by using proper orthogonal decomposition (POD) method in parallel. This process is recursively carried on until the obtained dataset cannot be compressed any more. Here, the parallel computing greatly decreases the computational cost of the eigen resolving problem in the POD algorithm. Furthermore the compressed dataset can be easily restored linearly.
ハイパフォーマンスコンピューティングと計算科学シンポジウム論文集
2013
83-83
2013-01-08