Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training

Wesley Brewer; Murali Meena Gopalakrishnan; Matthias Maiterth; Aditya Kashi; Jong Youl Choi; Pei Zhang; Stephen Nichols; Riccardo Balin; Miles Couchman; Stephen de Bruyn Kops; P. K. Yeung; Daniel Dotson; Rohini Uma-Vaideswaran; Sarp Oral; Feiyi Wang

doi:10.1145/3731599.3767340

Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training

Wesley Brewer
, Murali Meena Gopalakrishnan
, Matthias Maiterth
, Aditya Kashi
, Jong Youl Choi
, Pei Zhang
, Stephen Nichols
, Riccardo Balin
, Miles Couchman
, Stephen de Bruyn Kops
, P. K. Yeung
, Daniel Dotson
, Rohini Uma-Vaideswaran
, Sarp Oral
, Feiyi Wang

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

With the end of Moore’s law and Dennard scaling, efficient training increasingly requires rethinking data volume. Can we train better models with significantly less data via intelligent subsampling? To explore this, we develop SICKLE, a sparse intelligent curation framework for efficient learning, featuring a novel maximum entropy (MaxEnt) sampling approach, scalable training, and energy benchmarking. We compare MaxEnt with random and phase-space sampling on large direct numerical simulation (DNS) datasets of turbulence. Evaluating SICKLE at scale on Frontier, we show that subsampling as a preprocessing step can, in many cases, improve model accuracy and substantially lower energy consumption, with observed reductions of up to 38×.

Original language	English
Title of host publication	Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops
Publisher	Association for Computing Machinery, Inc
Pages	1-10
Number of pages	10
ISBN (Electronic)	9798400718717
DOIs	https://doi.org/10.1145/3731599.3767340
State	Published - Nov 15 2025
Event	2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops - St. Louis, United States Duration: Nov 16 2025 → Nov 21 2025

Publication series

Name	Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops

Conference

Conference	2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops
Country/Territory	United States
City	St. Louis
Period	11/16/25 → 11/21/25

Funding

This research was sponsored by and used resources of the Oak Ridge Leadership Computing Facility (OLCF), which is a DOE Office of Science User Facility at the Oak Ridge National Laboratory (ORNL) supported by the U.S. Department of Energy under Contract No. DEAC05-00OR22725. This research was also sponsored by the Artificial Intelligence Initiative as part of the Laboratory Directed Research and Development (LDRD) Program of ORNL. We would also like to thank a number of colleagues for helpful information and advice along the way, including: Malik Hassanaly for help with UIPS, as well as Shantenu Jha, Joel Bretheim, Andrew Shao, Trey White, and Bronson Messer for providing feedback on the manuscript. The authors acknowledge the use of ChatGPT (version 4.5, OpenAI) for language editing, LaTeX formatting assistance, and structural suggestions.

Keywords

energy efficiency
maximum entropy
phase-space selection
scientific foundation models
subsampling

Access to Document

10.1145/3731599.3767340

Cite this

Brewer, W., Gopalakrishnan, M. M., Maiterth, M., Kashi, A., Choi, J. Y., Zhang, P., Nichols, S., Balin, R., Couchman, M., de Bruyn Kops, S., Yeung, P. K., Dotson, D., Uma-Vaideswaran, R., Oral, S., & Wang, F. (2025). Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training. In Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops (pp. 1-10). (Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops). Association for Computing Machinery, Inc. https://doi.org/10.1145/3731599.3767340

Brewer, Wesley ; Gopalakrishnan, Murali Meena ; Maiterth, Matthias et al. / Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training. Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops. Association for Computing Machinery, Inc, 2025. pp. 1-10 (Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops).

@inproceedings{1bcbe1714ca940f1ab390145878e5a6f,

title = "Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training",

abstract = "With the end of Moore{\textquoteright}s law and Dennard scaling, efficient training increasingly requires rethinking data volume. Can we train better models with significantly less data via intelligent subsampling? To explore this, we develop SICKLE, a sparse intelligent curation framework for efficient learning, featuring a novel maximum entropy (MaxEnt) sampling approach, scalable training, and energy benchmarking. We compare MaxEnt with random and phase-space sampling on large direct numerical simulation (DNS) datasets of turbulence. Evaluating SICKLE at scale on Frontier, we show that subsampling as a preprocessing step can, in many cases, improve model accuracy and substantially lower energy consumption, with observed reductions of up to 38×.",

keywords = "energy efficiency, maximum entropy, phase-space selection, scientific foundation models, subsampling",

author = "Wesley Brewer and Gopalakrishnan, \{Murali Meena\} and Matthias Maiterth and Aditya Kashi and Choi, \{Jong Youl\} and Pei Zhang and Stephen Nichols and Riccardo Balin and Miles Couchman and \{de Bruyn Kops\}, Stephen and Yeung, \{P. K.\} and Daniel Dotson and Rohini Uma-Vaideswaran and Sarp Oral and Feiyi Wang",

note = "Publisher Copyright: {\textcopyright} 2025 Copyright held by the owner/author(s).; 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops ; Conference date: 16-11-2025 Through 21-11-2025",

year = "2025",

month = nov,

day = "15",

doi = "10.1145/3731599.3767340",

language = "English",

series = "Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops",

publisher = "Association for Computing Machinery, Inc",

pages = "1--10",

booktitle = "Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops",

}

Brewer, W , Gopalakrishnan, MM, Maiterth, M, Kashi, A , Choi, JY , Zhang, P , Nichols, S, Balin, R, Couchman, M, de Bruyn Kops, S, Yeung, PK, Dotson, D, Uma-Vaideswaran, R, Oral, S & Wang, F 2025, Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training. in Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops. Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops, Association for Computing Machinery, Inc, pp. 1-10, 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops, St. Louis, United States, 11/16/25. https://doi.org/10.1145/3731599.3767340

Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training. / Brewer, Wesley ; Gopalakrishnan, Murali Meena; Maiterth, Matthias et al.
Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops. Association for Computing Machinery, Inc, 2025. p. 1-10 (Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training

AU - Brewer, Wesley

AU - Gopalakrishnan, Murali Meena

AU - Maiterth, Matthias

AU - Kashi, Aditya

AU - Choi, Jong Youl

AU - Zhang, Pei

AU - Nichols, Stephen

AU - Balin, Riccardo

AU - Couchman, Miles

AU - de Bruyn Kops, Stephen

AU - Yeung, P. K.

AU - Dotson, Daniel

AU - Uma-Vaideswaran, Rohini

AU - Oral, Sarp

AU - Wang, Feiyi

PY - 2025/11/15

Y1 - 2025/11/15

N2 - With the end of Moore’s law and Dennard scaling, efficient training increasingly requires rethinking data volume. Can we train better models with significantly less data via intelligent subsampling? To explore this, we develop SICKLE, a sparse intelligent curation framework for efficient learning, featuring a novel maximum entropy (MaxEnt) sampling approach, scalable training, and energy benchmarking. We compare MaxEnt with random and phase-space sampling on large direct numerical simulation (DNS) datasets of turbulence. Evaluating SICKLE at scale on Frontier, we show that subsampling as a preprocessing step can, in many cases, improve model accuracy and substantially lower energy consumption, with observed reductions of up to 38×.

AB - With the end of Moore’s law and Dennard scaling, efficient training increasingly requires rethinking data volume. Can we train better models with significantly less data via intelligent subsampling? To explore this, we develop SICKLE, a sparse intelligent curation framework for efficient learning, featuring a novel maximum entropy (MaxEnt) sampling approach, scalable training, and energy benchmarking. We compare MaxEnt with random and phase-space sampling on large direct numerical simulation (DNS) datasets of turbulence. Evaluating SICKLE at scale on Frontier, we show that subsampling as a preprocessing step can, in many cases, improve model accuracy and substantially lower energy consumption, with observed reductions of up to 38×.

KW - energy efficiency

KW - maximum entropy

KW - phase-space selection

KW - scientific foundation models

KW - subsampling

UR - https://www.scopus.com/pages/publications/105023428964

U2 - 10.1145/3731599.3767340

DO - 10.1145/3731599.3767340

M3 - Conference contribution

AN - SCOPUS:105023428964

T3 - Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops

SP - 1

EP - 10

BT - Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops

PB - Association for Computing Machinery, Inc

T2 - 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops

Y2 - 16 November 2025 through 21 November 2025

ER -

Brewer W , Gopalakrishnan MM, Maiterth M, Kashi A , Choi JY , Zhang P et al. Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training. In Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops. Association for Computing Machinery, Inc. 2025. p. 1-10. (Proceedings of 2025 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, SC 2025 Workshops). doi: 10.1145/3731599.3767340

Intelligent Sampling of Extreme-Scale Turbulence Datasets for Accurate and Efficient Spatiotemporal Model Training

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this