当前位置: X-MOL 学术arXiv.cs.MS › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
ATC: an Advanced Tucker Compression library for multidimensional data
arXiv - CS - Mathematical Software Pub Date : 2021-07-03 , DOI: arxiv-2107.01384
Wouter Baert, Nick Vannieuwenhoven

We present ATC, a C++ library for advanced Tucker-based compression of multidimensional numerical data, based on the sequentially truncated higher-order singular value decomposition (ST-HOSVD) and bit plane truncation. Several techniques are proposed to improve compression rate, speed, memory usage and error control. First, a hybrid truncation scheme is described which combines Tucker rank truncation and TTHRESH quantization [Ballester-Ripoll et al., IEEE Trans. Visual. Comput. Graph., 2020]. We derive a novel expression to approximate the error of truncated Tucker decompositions in the case of core and factor perturbations. Furthermore, a Householder-reflector-based approach is proposed to compress the orthogonal Tucker factors. Certain key improvements to the quantization procedure are also discussed. Moreover, particular implementation aspects are described, such as ST-HOSVD procedure using only a single transposition. We also discuss several usability features of ATC, including the presence of multiple interfaces, extensive data type support and integrated downsampling of the decompressed data. Numerical results show that ATC maintains state-of-the-art Tucker compression rates, while providing average speed-ups of 2.6-3.6 and halving memory usage. Furthermore, our compressor provides precise error control, only deviating 1.4% from the requested error on average. Finally, ATC often achieves significantly higher compression than non-Tucker-based compressors in the high-error domain.

中文翻译:

ATC:用于多维数据的高级 Tucker 压缩库

我们提出了 ATC,这是一个 C++ 库,用于基于顺序截断的高阶奇异值分解 (ST-HOSVD) 和位平面截断对多维数值数据进行基于 Tucker 的高级压缩。提出了几种技术来提高压缩率、速度、内存使用和错误控制。首先,描述了一种混合截断方案,它结合了 Tucker 秩截断和 TTHRESH 量化 [Ballester-Ripoll 等人,IEEE Trans. 视觉的。计算。图。,2020]。我们推导出一个新的表达式来近似在核心和因子扰动的情况下截断的 Tucker 分解的误差。此外,提出了一种基于 Householder-reflector 的方法来压缩正交 Tucker 因子。还讨论了量化程序的某些关键改进。而且,描述了特定的实现方面,例如仅使用单个换位的 ST-HOSVD 程序。我们还讨论了 ATC 的几个可用性特性,包括多个接口的存在、广泛的数据类型支持和解压缩数据的集成下采样。数值结果表明,ATC 保持了最先进的 Tucker 压缩率,同时提供 2.6-3.6 的平均加速并将内存使用量减半。此外,我们的压缩机提供精确的误差控制,平均仅偏离请求误差的 1.4%。最后,在高误差域中,ATC 通常比非基于 Tucker 的压缩器实现更高的压缩。广泛的数据类型支持和解压缩数据的集成下采样。数值结果表明,ATC 保持了最先进的 Tucker 压缩率,同时提供 2.6-3.6 的平均加速并将内存使用量减半。此外,我们的压缩机提供精确的误差控制,平均仅偏离请求误差的 1.4%。最后,在高误差域中,ATC 通常比非基于 Tucker 的压缩器实现更高的压缩。广泛的数据类型支持和解压缩数据的集成下采样。数值结果表明,ATC 保持了最先进的 Tucker 压缩率,同时提供 2.6-3.6 的平均加速并将内存使用量减半。此外,我们的压缩机提供精确的误差控制,平均仅偏离请求误差的 1.4%。最后,在高误差域中,ATC 通常比非基于 Tucker 的压缩器实现更高的压缩。
更新日期:2021-07-06
down
wechat
bug