Statical causal discovery is an approach to infer the causal relationship between observed variables whose causalities are not revealed. LiNGAM (Linear Non-Gaussian Acyclic Model), an algorithm for causal discovery, can calculate the causal relationship uniquely if the independent components of variables are assumed to be non-Gaussian. However, use-cases of LiNGAM are limited because of its O(d3x) computational complexity, where dx is the number of variables. This paper shows two approaches to accelerate LiNGAM causal discovery: SIMD utilization for LiNGAM's mathematical matrixes operations and MPI parallelization. We evaluate the implementation with the supercomputer Fugaku. Using 96 nodes of Fugaku, our improved version can achieve 17,531 times faster than the original OSS implementation (completed in 17.7 hours).
Kazuhito MATSUDA
Fujitsu Limited
Kouji KURIHARA
Fujitsu Limited
Kentaro KAWAKAMI
Fujitsu Limited
Masafumi YAMAZAKI
Fujitsu Limited
Fuyuka YAMADA
Fujitsu Limited
Tsuguchika TABARU
Fujitsu Limited
Ken YOKOYAMA
Fujitsu Limited
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Kazuhito MATSUDA, Kouji KURIHARA, Kentaro KAWAKAMI, Masafumi YAMAZAKI, Fuyuka YAMADA, Tsuguchika TABARU, Ken YOKOYAMA, "Accelerating LiNGAM Causal Discovery with Massive Parallel Execution on Supercomputer Fugaku" in IEICE TRANSACTIONS on Information,
vol. E105-D, no. 12, pp. 2032-2039, December 2022, doi: 10.1587/transinf.2022PAP0007.
Abstract: Statical causal discovery is an approach to infer the causal relationship between observed variables whose causalities are not revealed. LiNGAM (Linear Non-Gaussian Acyclic Model), an algorithm for causal discovery, can calculate the causal relationship uniquely if the independent components of variables are assumed to be non-Gaussian. However, use-cases of LiNGAM are limited because of its O(d3x) computational complexity, where dx is the number of variables. This paper shows two approaches to accelerate LiNGAM causal discovery: SIMD utilization for LiNGAM's mathematical matrixes operations and MPI parallelization. We evaluate the implementation with the supercomputer Fugaku. Using 96 nodes of Fugaku, our improved version can achieve 17,531 times faster than the original OSS implementation (completed in 17.7 hours).
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2022PAP0007/_p
Copy
@ARTICLE{e105-d_12_2032,
author={Kazuhito MATSUDA, Kouji KURIHARA, Kentaro KAWAKAMI, Masafumi YAMAZAKI, Fuyuka YAMADA, Tsuguchika TABARU, Ken YOKOYAMA, },
journal={IEICE TRANSACTIONS on Information},
title={Accelerating LiNGAM Causal Discovery with Massive Parallel Execution on Supercomputer Fugaku},
year={2022},
volume={E105-D},
number={12},
pages={2032-2039},
abstract={Statical causal discovery is an approach to infer the causal relationship between observed variables whose causalities are not revealed. LiNGAM (Linear Non-Gaussian Acyclic Model), an algorithm for causal discovery, can calculate the causal relationship uniquely if the independent components of variables are assumed to be non-Gaussian. However, use-cases of LiNGAM are limited because of its O(d3x) computational complexity, where dx is the number of variables. This paper shows two approaches to accelerate LiNGAM causal discovery: SIMD utilization for LiNGAM's mathematical matrixes operations and MPI parallelization. We evaluate the implementation with the supercomputer Fugaku. Using 96 nodes of Fugaku, our improved version can achieve 17,531 times faster than the original OSS implementation (completed in 17.7 hours).},
keywords={},
doi={10.1587/transinf.2022PAP0007},
ISSN={1745-1361},
month={December},}
Copy
TY - JOUR
TI - Accelerating LiNGAM Causal Discovery with Massive Parallel Execution on Supercomputer Fugaku
T2 - IEICE TRANSACTIONS on Information
SP - 2032
EP - 2039
AU - Kazuhito MATSUDA
AU - Kouji KURIHARA
AU - Kentaro KAWAKAMI
AU - Masafumi YAMAZAKI
AU - Fuyuka YAMADA
AU - Tsuguchika TABARU
AU - Ken YOKOYAMA
PY - 2022
DO - 10.1587/transinf.2022PAP0007
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E105-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2022
AB - Statical causal discovery is an approach to infer the causal relationship between observed variables whose causalities are not revealed. LiNGAM (Linear Non-Gaussian Acyclic Model), an algorithm for causal discovery, can calculate the causal relationship uniquely if the independent components of variables are assumed to be non-Gaussian. However, use-cases of LiNGAM are limited because of its O(d3x) computational complexity, where dx is the number of variables. This paper shows two approaches to accelerate LiNGAM causal discovery: SIMD utilization for LiNGAM's mathematical matrixes operations and MPI parallelization. We evaluate the implementation with the supercomputer Fugaku. Using 96 nodes of Fugaku, our improved version can achieve 17,531 times faster than the original OSS implementation (completed in 17.7 hours).
ER -