Green Recommender Systems – Minimizing Carbon Footprint for Sustainable Personalization

Did you know one recommender systems research paper emits ~3,300 kgCO₂e — the same as one person flying from New York to Melbourne? This talk unveils the urgent need for Green Recommender Systems and delivers actionable guidelines to achieve them. We quantify the carbon footprint of training and inference, comparing deep learning and traditional algorithms. The goal? High-performance recommender systems that don’t cost the Earth. Discover how to minimize the carbon footprint while maintaining performance through energy-aware design, efficient hardware, and transparent reporting. This is a call to action: by rethinking how we design and measure recommenders, we can pioneer sustainable AI that benefits both users and the planet.

Lukas Wegmeth is a Ph.D. Student of the Intelligent Systems Group at the University of Siegen. Before joining the ISG he completed his bachelor’s and master’s degree in Medical Computer Science at the University of Siegen. During his time as a graduate student, Lukas set his focus on the topic of Machine Learning and collaborated with different chairs of the University of Siegen to work on and release scientific research papers in the field. Lukas is currently analysing recommender systems from an energy efficiency context, measuring amongst other power consumptions.

Tobias is a joint Ph.D. candidate at the ISG – Intelligent Systems Group (University of Siegen) in Siegen, Germany and the ADReM – Adrem Data Lab (University of Antwerp) in Antwerp, Belgium, working on model selection and automation in recommender systems. Generally, his research interests revolve around Recommender Systems, specifically applying ideas from AutoML (Automated Machine Learning) to information retrieval and recommender systems.