MLCA: a tool for Machine Learning Life Cycle Assessment - Archive ouverte HAL
Communication Dans Un Congrès Année : 2024

MLCA: a tool for Machine Learning Life Cycle Assessment

Résumé

The on-going environmental changes challenge the ever-increasing use of digital technologies. Tools such as Green Algorithms or Carbontracker provide support for estimating the environmental impact of calculations (e.g., training a machine learning model). However, these tools only account for the dynamic consumption induced by calculations and only document carbon footprint while other types of impacts, such as resource depletion, are not evaluated. To provide a more comprehensive assessment of machine learning impact, we propose a modeling of graphics cards manufacturing impacts and a multi-criteria estimation tool called MLCA that accounts for the production impacts of hardware used to perform calculations. We evaluate MLCA through three reproduction studies thereby showing the validity of the assessments as well as the contribution of evaluating diverse impact categories over different life cycle phases. We hope this tool will help better understand the environmental impacts of Machine learning as a whole.
Fichier principal
Vignette du fichier
ICT4S_Paper.pdf (1.22 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04643414 , version 1 (10-07-2024)

Identifiants

  • HAL Id : hal-04643414 , version 1

Citer

Clément Morand, Aurélie Névéol, Anne-Laure Ligozat. MLCA: a tool for Machine Learning Life Cycle Assessment. 2024 International Conference on ICT for Sustainability (ICT4S), Jun 2024, Stockholm, Sweden. ⟨hal-04643414⟩
746 Consultations
283 Téléchargements

Partager

More