Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Eco-Aware Graph Neural Networks for Sustainable Recommendations

Antonio Purificato, Fabrizio Silvestri

TL;DR

This paper tackles the often-overlooked environmental footprint of Graph Neural Network-based recommender systems by conducting a reproducible, emissions-focused evaluation. It uses CodeCarbon to measure $CO_2$-eq emissions while comparing NGCF, LightGCN, SimGCL, and LightGCL across embedding sizes 32–256 on datasets ML-1M, Amazon Beauty, and DianPing. The results show LightGCN generally offers the best performance with favorable emissions, while larger embeddings raise emissions and some models incur higher compute costs, revealing clear performance–sustainability trade-offs. The study emphasizes the need for eco-aware design in RSs and provides practical guidance for balancing recommendation quality with environmental impact.

Abstract

Recommender systems play a crucial role in alleviating information overload by providing personalized recommendations tailored to users' preferences and interests. Recently, Graph Neural Networks (GNNs) have emerged as a promising approach for recommender systems, leveraging their ability to effectively capture complex relationships and dependencies between users and items by representing them as nodes in a graph structure. In this study, we investigate the environmental impact of GNN-based recommender systems, an aspect that has been largely overlooked in the literature. Specifically, we conduct a comprehensive analysis of the carbon emissions associated with training and deploying GNN models for recommendation tasks. We evaluate the energy consumption and carbon footprint of different GNN architectures and configurations, considering factors such as model complexity, training duration, hardware specifications and embedding size. By addressing the environmental impact of resource-intensive algorithms in recommender systems, this study contributes to the ongoing efforts towards sustainable and responsible artificial intelligence, promoting the development of eco-friendly recommendation technologies that balance performance and environmental considerations. Code is available at: https://github.com/antoniopurificato/gnn_recommendation_and_environment.

Eco-Aware Graph Neural Networks for Sustainable Recommendations

TL;DR

This paper tackles the often-overlooked environmental footprint of Graph Neural Network-based recommender systems by conducting a reproducible, emissions-focused evaluation. It uses CodeCarbon to measure -eq emissions while comparing NGCF, LightGCN, SimGCL, and LightGCL across embedding sizes 32–256 on datasets ML-1M, Amazon Beauty, and DianPing. The results show LightGCN generally offers the best performance with favorable emissions, while larger embeddings raise emissions and some models incur higher compute costs, revealing clear performance–sustainability trade-offs. The study emphasizes the need for eco-aware design in RSs and provides practical guidance for balancing recommendation quality with environmental impact.

Abstract

Recommender systems play a crucial role in alleviating information overload by providing personalized recommendations tailored to users' preferences and interests. Recently, Graph Neural Networks (GNNs) have emerged as a promising approach for recommender systems, leveraging their ability to effectively capture complex relationships and dependencies between users and items by representing them as nodes in a graph structure. In this study, we investigate the environmental impact of GNN-based recommender systems, an aspect that has been largely overlooked in the literature. Specifically, we conduct a comprehensive analysis of the carbon emissions associated with training and deploying GNN models for recommendation tasks. We evaluate the energy consumption and carbon footprint of different GNN architectures and configurations, considering factors such as model complexity, training duration, hardware specifications and embedding size. By addressing the environmental impact of resource-intensive algorithms in recommender systems, this study contributes to the ongoing efforts towards sustainable and responsible artificial intelligence, promoting the development of eco-friendly recommendation technologies that balance performance and environmental considerations. Code is available at: https://github.com/antoniopurificato/gnn_recommendation_and_environment.

Paper Structure

This paper contains 15 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Calculating CO2 Emissions
  5. Models
  6. Experimental Pipeline
  7. Experiments
  8. Datasets
  9. Metrics
  10. Reproducibility
  11. Hardware
  12. Results
  13. RQ1: Which model excels, and at what cost?
  14. Embedding Size and Environmental Impact
  15. Conclusions

Figures (1)

  • Figure 1: Difference in terms of environmental impact when changing the embedding size of the GNNs. On the y-axis it can be seen the CO$_2$-eq while on the x-axis the embedding size.