What is it about?

Many applications in science and industry create large amounts of numbers organized on a grid with two (a table), three or more dimensions. We developed compression software to reduce the size of such data using tensor-based mathematical methods.

Featured Image

Why is it important?

We invented new techniques to improve the underlying algorithms used before to compress grid-based data, leading to faster and sometimes more compact compression, compressing data by a factor of up to 100,000. Furthermore, we combined these techniques with an improved implementation to arrive at a software package that is easier to use in many ways than existing work.

Perspectives

While reducing the size of a file on your computer might not sound interesting, data compression is in fact a very difficult problem as it relies on fundamental notions of information theory, such as what "information" really is. After all, how much data can be compressed does not only depend on the data but also on design choices in the compressor: which data is considered common (low information) or exceptional (high information)? Therefore, in a world with ever-increasing production of data but limited storage and transfer capacities, I hope this work finds use in practical settings or is improved by others to create even more optimized compressors.

Wouter Baert
Associatie KU Leuven

Read the Original

This page is a summary of: Algorithm 1036: ATC, An Advanced Tucker Compression Library for Multidimensional Data, ACM Transactions on Mathematical Software, June 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3585514.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page