Perspective
Open Access
Energy Scalability in the Training of AI Models for Image Processing: The Role of Hyperparameters
1
Department of Mathematics and Computer Science, University of the Balearic Islands, 07122 Palma, Spain
* Corresponding Author:
David Cortes,
[email protected]
Volume 1, Issue 1
2026
Received: 22 December 2025
Accepted: 11 February 2026
Published: 10 March 2026
Article Information
Abstract
This perspective article argues that hyperparameters such as learning rate, batch size, numerical precision, and training workers are key determinants of energy scalability in CNN training. These parameters directly influence convergence dynamics, hardware utilization, and training duration, leading to substantially different energy profiles even when comparable accuracy is achieved. Moreover, hyperparameter search itself introduces a significant cumulative energy cost, often exceeding that of the final selected model. By analyzing the interaction between convergence behavior and energy consumption, this work highlights the need to treat energy as an explicit scalability metric and to integrate energy-aware considerations into hyperparameter optimization. Adopting this perspective enables more efficient, sustainable, and reproducible training practices for large-scale image processing models.
Keywords
hyperparameters
energy
scalability
training of AI
Data Availability Statement
Not applicable.
Funding
This work was supported without any funding.
Conflicts of Interest
The authors declare no conflicts of interest.
AI Use Statement
The authors declare that no generative AI was used in the preparation of this manuscript.
Ethical Approval and Consent to Participate
Not applicable.
References
- Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (pp. 3645-3650).
[CrossRef] [Google Scholar] - Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L. M., Rothchild, D., ... & Dean, J. (2021). Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350.
[Google Scholar] - LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444.
[CrossRef] [Google Scholar] - Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012, December). ImageNet classification with deep convolutional neural networks. In Proceedings of the 26th International Conference on Neural Information Processing Systems-Volume 1 (pp. 1097-1105).
[Google Scholar] - Cortes, D., Bermejo, B., & Juiz, C. (2024). The use of CNNs in VR/AR/MR/XR: a systematic literature review. Virtual Reality, 28(3), 154.
[CrossRef] [Google Scholar] - Tan, M., & Le, Q. (2021, July). Efficientnetv2: Smaller models and faster training. In International conference on machine learning (pp. 10096-10106). PMLR.
[Google Scholar] - Cortes, D., Juiz, C., & Bermejo, B. (2025). Estudio de la eficiencia en la escalabilidad de GPUs para el entrenamiento de Inteligencia Artificial. arXiv preprint arXiv:2509.03263.
[Google Scholar] - García-Martín, E., Rodrigues, C. F., Riley, G., & Grahn, H. (2019). Estimation of energy consumption in machine learning. Journal of Parallel and Distributed Computing, 134, 75-88.
[CrossRef] [Google Scholar] - Aquino-Brítez, S., García-Sánchez, P., Ortiz, A., & Aquino-Brítez, D. (2025). Towards an energy consumption index for deep learning models: A comparative analysis of architectures, GPUs, and measurement tools. Sensors, 25(3), 846.
[CrossRef] [Google Scholar] - Winsta, J. (2025). The hidden costs of ai: A review of energy, e-waste, and inequality in model development. arXiv preprint arXiv:2507.09611.
[Google Scholar] - Tripp, C. E., Perr-Sauer, J., Gafur, J., Nag, A., Purkayastha, A., Zisman, S., & Bensen, E. A. (2024). Measuring the energy consumption and efficiency of deep neural networks: An empirical analysis and design recommendations. arXiv preprint arXiv:2403.08151.
[Google Scholar] - Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2020). Green ai. Communications of the ACM, 63(12), 54-63.
[CrossRef] [Google Scholar] - Henderson, P., Hu, J., Romoff, J., Brunskill, E., Jurafsky, D., & Pineau, J. (2020). Towards the systematic reporting of the energy and carbon footprints of machine learning. Journal of machine learning research, 21(248), 1-43.
[Google Scholar] - Geissler, D., Zhou, B., Suh, S., & Lukowicz, P. (2024). Spend more to save more (sm2): An energy-aware implementation of successive halving for sustainable hyperparameter optimization. arXiv preprint arXiv:2412.08526.
[Google Scholar] - Geißler, D., Zhou, B., Liu, M., Suh, S., & Lukowicz, P. (2024, May). The power of training: How different neural network setups influence the energy demand. In International Conference on Architecture of Computing Systems (pp. 33-47). Cham: Springer Nature Switzerland.
[CrossRef] [Google Scholar] - Smith, S. L., Kindermans, P. J., Ying, C., & Le, Q. V. (2017). Don't decay the learning rate, increase the batch size. arXiv preprint arXiv:1711.00489.
[Google Scholar] - Li, L., & Talwalkar, A. (2020, August). Random search and reproducibility for neural architecture search. In Uncertainty in artificial intelligence (pp. 367-377). PMLR.
[Google Scholar] - Smith, L. N. (2017, March). Cyclical learning rates for training neural networks. In 2017 IEEE winter conference on applications of computer vision (WACV) (pp. 464-472). IEEE.
[CrossRef] [Google Scholar] - Bottou, L. (2010, September). Large-scale machine learning with stochastic gradient descent. In Proceedings of COMPSTAT'2010: 19th International Conference on Computational StatisticsParis France, August 22-27, 2010 Keynote, Invited and Contributed Papers (pp. 177-186). Heidelberg: Physica-Verlag HD.
[CrossRef] [Google Scholar] - Ortiz, M., Cristal, A., Ayguadé, E., & Casas, M. (2018). Low-precision floating-point schemes for neural network training. arXiv preprint arXiv:1804.05267.
[Google Scholar] - Narang, S., Diamos, G., Elsen, E., Micikevicius, P., Alben, J., Garcia, D., ... & Wu, H. (2017, October). Mixed precision training. In Int. Conf. on Learning Representation.
[Google Scholar] - You, J., Chung, J. W., & Chowdhury, M. (2023). Zeus: Understanding and optimizing \{GPU\ energy consumption of \{DNN\ training. In 20th USENIX Symposium on Networked Systems Design and Implementation (NSDI 23) (pp. 119-139).
[Google Scholar] - Keskar, N. S., Mudigere, D., Nocedal, J., Smelyanskiy, M., & Tang, P. T. P. (2016). On large-batch training for deep learning: Generalization gap and sharp minima. arXiv preprint arXiv:1609.04836.
[Google Scholar] - Frey, N. C., Zhao, D., Axelrod, S., Jones, M., Bestor, D., Gadepally, V., ... & Samsi, S. (2022, May). Energy-aware neural architecture selection and hyperparameter optimization. In 2022 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) (pp. 732-741). IEEE.
[CrossRef] [Google Scholar] - dos Santos, G. C., Araújo, A. L. D., Nakamura, T. C. R., Sousa-Neto, S. S., Giraldo-Roldán, D., Lopes, M. A., ... & Moraes, M. C. (2024, September). Assessment of the Influence of Batch Size on ResNet-50 Processing Applied to Histopathological Datasets. In Brazilian Congress on Biomedical Engineering (pp. 548-556). Cham: Springer Nature Switzerland.
[CrossRef] [Google Scholar]
Cite This Article
APA Style
Atiénzar, B., Cortes, D., Bermejo, B., & Juiz, C. (2026). Energy Scalability in the Training of AI Models for Image Processing: The Role of Hyperparameters. Journal of Systems Scalability, 1(1), 1–5. https://doi.org/10.62762/JSS.2025.960646
Export Citation
RIS Format
Compatible with EndNote, Zotero, Mendeley, and other reference managers
TY - JOUR AU - Atiénzar, Blanca AU - Cortes, David AU - Bermejo, Belen AU - Juiz, Carlos PY - 2026 DA - 2026/03/10 TI - Energy Scalability in the Training of AI Models for Image Processing: The Role of Hyperparameters JO - Journal of Systems Scalability T2 - Journal of Systems Scalability JF - Journal of Systems Scalability VL - 1 IS - 1 SP - 1 EP - 5 DO - 10.62762/JSS.2025.960646 UR - https://www.icck.org/article/abs/JSS.2025.960646 KW - hyperparameters KW - energy KW - scalability KW - training of AI AB - This perspective article argues that hyperparameters such as learning rate, batch size, numerical precision, and training workers are key determinants of energy scalability in CNN training. These parameters directly influence convergence dynamics, hardware utilization, and training duration, leading to substantially different energy profiles even when comparable accuracy is achieved. Moreover, hyperparameter search itself introduces a significant cumulative energy cost, often exceeding that of the final selected model. By analyzing the interaction between convergence behavior and energy consumption, this work highlights the need to treat energy as an explicit scalability metric and to integrate energy-aware considerations into hyperparameter optimization. Adopting this perspective enables more efficient, sustainable, and reproducible training practices for large-scale image processing models. SN - 3142-7855 PB - Institute of Central Computation and Knowledge LA - English ER -
BibTeX Format
Compatible with LaTeX, BibTeX, and other reference managers
@article{Atinzar2026Energy,
author = {Blanca Atiénzar and David Cortes and Belen Bermejo and Carlos Juiz},
title = {Energy Scalability in the Training of AI Models for Image Processing: The Role of Hyperparameters},
journal = {Journal of Systems Scalability},
year = {2026},
volume = {1},
number = {1},
pages = {1-5},
doi = {10.62762/JSS.2025.960646},
url = {https://www.icck.org/article/abs/JSS.2025.960646},
abstract = {This perspective article argues that hyperparameters such as learning rate, batch size, numerical precision, and training workers are key determinants of energy scalability in CNN training. These parameters directly influence convergence dynamics, hardware utilization, and training duration, leading to substantially different energy profiles even when comparable accuracy is achieved. Moreover, hyperparameter search itself introduces a significant cumulative energy cost, often exceeding that of the final selected model. By analyzing the interaction between convergence behavior and energy consumption, this work highlights the need to treat energy as an explicit scalability metric and to integrate energy-aware considerations into hyperparameter optimization. Adopting this perspective enables more efficient, sustainable, and reproducible training practices for large-scale image processing models.},
keywords = {hyperparameters, energy, scalability, training of AI},
issn = {3142-7855},
publisher = {Institute of Central Computation and Knowledge}
}
Article Metrics
Publisher's Note
ICCK stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and Permissions
Copyright © 2026 by the Author(s). Published by Institute of Central Computation and Knowledge. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
Scan to read this article