Supercomputing Frontiers Europe 2019

Virtual ICM Seminars in Computer and Computational Science

Virtual ICM Seminars
in Computer and Computational Science

Ben Whitney

MGARD: Hierarchical Compression of Scientific Data

February 18, 2021 | 16:00 CET

Ben Whitney

Postdoc in the Computer Science and Mathematics Division at Oak Ridge National Laboratory


Thu, FEB 18, 2021

16.00 CET

Open for all

Free of charge

Online

MGARD: Hierarchical Compression of Scientific Data

The increasing scale of scientific datasets poses challenges to computational scientists needing to store, transfer, and understand their data. Lossy compression can help alleviate these problems, but practitioners are wary of incurring errors that may affect downstream analyses. In this talk, we present MGARD (MultiGrid Adaptive Reduction of Data), an algorithm for scientific data compression with guaranteed error bounds and the ability to preserve quantities of scientific interest. We begin by motivating the hierarchical decomposition underlying the algorithm. Next, we discuss quantization techniques for compression with guaranteed error bounds. Finally, we present numerical examples showcasing MGARD’s ability to preserve quantities of scientific interest.

See more




 

Honorary Patron
OPI

Media Partners
DatanamiHPCwireEnterpriseTechitwizComputerworld