GAISSA Team Publishes Research on ML Optimization for Sustainability in Computing

Jul 17, 2025

A paper by the GAISSA team has been published as an open-access article in the Springer journal Computing. The research provides a detailed evaluation of various ML optimization techniques, offering practical guidelines for creating greener AI systems by reducing energy consumption during model inference.

GAISSA Team Publishes Research on ML Optimization for Sustainability in Computing — Our paper, "Impact of ML optimization tactics on greener pre-trained ML models," published in the journal Computing.

We are excited to announce that our research paper, "Impact of ML optimization tactics on greener pre-trained ML models," has been published in the peer-reviewed journal Computing. The work is authored by GAISSA team members Alexandra González, Joel Castaño, Xavier Franch, and Silverio Martínez-Fernández.

This study presents a controlled experiment evaluating the impact of different PyTorch optimization techniques, such as dynamic quantization and pruning, on 42 pre-trained models from Hugging Face. The research offers valuable, data-driven guidelines for practitioners on how to effectively reduce the energy consumption and economic costs of ML models during inference, without significantly compromising accuracy.

This work is a key contribution to the GAISSA project's goal of providing software engineering tactics for greener AI.

Publication Details

Full Citation: González, A., Castaño, J., Franch, X. & Martínez-Fernández, S. Impact of ML optimization tactics on greener pre-trained ML models. Computing 107, 103 (2025).
DOI (Open Access): https://doi.org/10.1007/s00607-025-01437-8