Details
| Original language | English |
|---|---|
| Title of host publication | Proceedings - 2025 11th International Conference on ICT for Sustainability, ICT4S 2025 |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 100-111 |
| Number of pages | 12 |
| ISBN (electronic) | 9798331587178 |
| ISBN (print) | 979-8-3315-8718-5 |
| Publication status | Published - 9 Jun 2025 |
| Event | 11th International Conference on ICT for Sustainability, ICT4S 2025 - Dublin, Ireland Duration: 9 Jun 2025 → 13 Jun 2025 |
Abstract
The reliability of machine learning (ML) software systems is heavily influenced by changes in data over time. For that reason, ML systems require regular maintenance, typically based on model retraining. However, retraining requires significant computational demand, which makes it energy-intensive and raises concerns about its environmental impact. To understand which retraining techniques should be considered when designing sustainable ML applications, in this work, we study the energy consumption of common retraining techniques. Since the accuracy of ML systems is also essential, we compare retraining techniques in terms of both energy efficiency and accuracy. We showcase that retraining with only the most recent data compared to all available data reduces energy consumption by up to 25%, being a sustainable alternative to the status quo. Furthermore, our findings show that retraining a model only when there is evidence that updates are necessary, rather than on a fixed schedule, can reduce energy consumption by up to 40%, provided a reliable data change detector is in place. Our findings pave the way for better recommendations for ML practitioners, guiding them toward more energy-efficient retraining techniques when designing sustainable ML software systems.
Keywords
- Green AI, ML-enabled systems, sustainable AI monitoring, sustainable retraining
ASJC Scopus subject areas
- Computer Science(all)
- Artificial Intelligence
- Computer Science(all)
- Computer Networks and Communications
- Computer Science(all)
- Information Systems
- Decision Sciences(all)
- Information Systems and Management
- Energy(all)
- Renewable Energy, Sustainability and the Environment
- Engineering(all)
- Building and Construction
Sustainable Development Goals
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
Proceedings - 2025 11th International Conference on ICT for Sustainability, ICT4S 2025. Institute of Electrical and Electronics Engineers Inc., 2025. p. 100-111.
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Sustainable Machine Learning Retraining
T2 - 11th International Conference on ICT for Sustainability, ICT4S 2025
AU - Poenaru-Olaru, Lorena
AU - Sallou, June
AU - Miranda da Cruz, Luis
AU - Rellermeyer, Jan S.
AU - Van Deursen, Arie
N1 - Publisher Copyright: © 2025 IEEE.
PY - 2025/6/9
Y1 - 2025/6/9
N2 - The reliability of machine learning (ML) software systems is heavily influenced by changes in data over time. For that reason, ML systems require regular maintenance, typically based on model retraining. However, retraining requires significant computational demand, which makes it energy-intensive and raises concerns about its environmental impact. To understand which retraining techniques should be considered when designing sustainable ML applications, in this work, we study the energy consumption of common retraining techniques. Since the accuracy of ML systems is also essential, we compare retraining techniques in terms of both energy efficiency and accuracy. We showcase that retraining with only the most recent data compared to all available data reduces energy consumption by up to 25%, being a sustainable alternative to the status quo. Furthermore, our findings show that retraining a model only when there is evidence that updates are necessary, rather than on a fixed schedule, can reduce energy consumption by up to 40%, provided a reliable data change detector is in place. Our findings pave the way for better recommendations for ML practitioners, guiding them toward more energy-efficient retraining techniques when designing sustainable ML software systems.
AB - The reliability of machine learning (ML) software systems is heavily influenced by changes in data over time. For that reason, ML systems require regular maintenance, typically based on model retraining. However, retraining requires significant computational demand, which makes it energy-intensive and raises concerns about its environmental impact. To understand which retraining techniques should be considered when designing sustainable ML applications, in this work, we study the energy consumption of common retraining techniques. Since the accuracy of ML systems is also essential, we compare retraining techniques in terms of both energy efficiency and accuracy. We showcase that retraining with only the most recent data compared to all available data reduces energy consumption by up to 25%, being a sustainable alternative to the status quo. Furthermore, our findings show that retraining a model only when there is evidence that updates are necessary, rather than on a fixed schedule, can reduce energy consumption by up to 40%, provided a reliable data change detector is in place. Our findings pave the way for better recommendations for ML practitioners, guiding them toward more energy-efficient retraining techniques when designing sustainable ML software systems.
KW - Green AI
KW - ML-enabled systems
KW - sustainable AI monitoring
KW - sustainable retraining
UR - http://www.scopus.com/inward/record.url?scp=105018907894&partnerID=8YFLogxK
U2 - 10.1109/ICT4S68164.2025.00019
DO - 10.1109/ICT4S68164.2025.00019
M3 - Conference contribution
AN - SCOPUS:105018907894
SN - 979-8-3315-8718-5
SP - 100
EP - 111
BT - Proceedings - 2025 11th International Conference on ICT for Sustainability, ICT4S 2025
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 9 June 2025 through 13 June 2025
ER -