基於分佈的時序集成科學資料縮減及不確定性可視化與分析

Distribution-based Time-varying Ensemble Scientific Data Reduction for Uncertainty Visualization and Analysis

林志融(Lin Zhi-Rong)

指導教授 ： 王科植

國立臺灣師範大學/理學院/資訊工程學系/碩士(2022年)

https://doi.org/10.6345/NTNU202201624

摘要

科學家經常使用計算機模擬模型來研究物理現象。為了研究物理現象中的不確定性會投過調整初始參數內部的隨機變量來產生多個結果。因此，每個網格點是以模擬運行的多個數據值表示，我們稱這種類型的數據集成數據集。為了深入了解物理現象，科學家需要視覺化和分析具有不確定性的聚合數據集。基於分佈數據表示是處理集成數據集和研究不確定性可視化的流行方法。但是，存儲一個時變集成數據集需要非常大量的儲存花費可以輕易地超過數硬體的儲存空間。因此我們提出了一種新的數據表示來緊湊地表示隨時間變化的科學數據，用於不確定性可視化和分析。我們的方法將時域中的數據解耦為兩種類型的分佈和存儲。分佈匯總數據時域中的值，另一個分佈描述了時域中數據值出現的概率。我們的方法可以再交低的儲存花費下提供具有不確定性分析並保存時間特徵。

關鍵字

大數據處理 ； 科學資料 ； 數據縮減

並列摘要

Scientists often study physical phenomena using computer simulation models. The same simulation can generate different datasets because of different input parameter configurations or the internal random variables. Therefore, each grid point is represented by multiple data values from simulation runs, and we call this type of data ensemble dataset. To gain insight into the physical phenomenon, scientists often have to visualize and analyze the ensemble datasets with uncertainty. Distribution-based data representation is a popular approach to handle the ensemble dataset and support uncertainty visualization. However, storing a timevarying ensemble dataset needs hundreds or even thousands of times storage size. Given the size of the time-varying ensemble dataset, it is natural to develop storage reduced data representation to facilitate the time-varying ensemble data exploration. We propose a novel data representation to compactly represent the time-varying scientific data for uncertainty visualization and analysis. Our approach decouples data on the temporal domain into two types of distributions and stores. One distribution summarizes the data values on the temporal domain, and the other distribution describes the occurrence probability of a data value on the temporal domain. Our approach can provide time-varying ensemble scientific data analysis with uncertainty quantification and detailed temporal feature evolution with less storage requirement.

並列關鍵字

Data reduction ； ensemble data ； probability distribution ； statistical modeling

參考文獻

Radhakrishna Achanta, Appu Shaji, Kevin Smith, Aurelien Lucchi, Pascal Fua, and Sabine Susstrunk. Slic superpixels compared to state-of-the-art superpixel methods. ¨ IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(11):2274–2282, 2012.

Ann S. Almgren, John B. Bell, Mike J. Lijewski, Zarija Lukic, and Ethan Van Andel. ´ Nyx: A massively parallel AMR code for computational cosmology. Astrophysical Journal, 765(1), 2013.

Tushar Athawale, Elham Sakhaee, and Alireza Entezari. Isosurface visualization of data with nonparametric models for uncertainty. IEEE Transactions on Visualization and Computer Graphics, 22(1):777–786, 2016.

Chun-Ming Chen, Ayan Biswas, and Han-Wei Shen. Uncertainty modeling and error reduction for pathline computation in time-varying flow fields. In 2015 IEEE Pacific Visualization Symposium (PacificVis), pages 215–222, 2015.

Chun-Ming Chen, Ayan Biswas, and Han-Wei Shen. Uncertainty modeling and error reduction for pathline computation in time-varying flow fields. In 2015 IEEE Pacific Visualization Symposium (PacificVis), pages 215–222. IEEE, 2015.