Research map / Federated learning and privacy / Communication-efficient FL
A First Look into the Carbon Footprint of Federated Learning
Xinchi Qiu, Titouan Parcollet, Javier Fernandez-Marques, Pedro P. B. Gusmao, Yan Gao, Daniel J. Beutel, Taner Topal, Akhil Mathur, Nicholas D. Lane
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use FL with 1 local epoch instead of centralized V100 training | CIFAR10 non-IID · CIFAR10 | CO2e emissions: more than 10 times more carbon | Hurt |
| increase local epochs from 1 to 5 in FL (ablation) | FedAdam on CIFAR10 non-IID · CIFAR10 | communication rounds: 1800 -> 800 | Improved |
| increase local epochs from 1 to 5 in FL (ablation) | FedAvg on CIFAR10 non-IID · CIFAR10 | communication rounds: >2000 -> 580 | Improved |
| increase local epochs from 1 to 5 in FL (ablation) | FedAvg on ImageNet non-IID · ImageNet | CO2e emissions | Hurt |
| increase local epochs from 1 to 5 in FL (ablation) | FL on CIFAR10, FEMNIST, Speech Commands non-IID · non-IID FL | CO2e emissions | Improved |
| use FedAdam instead of FedAvg | FL on Speech Commands and FEMNIST · Speech Commands, FEMNIST | CO2e emissions: decrease of up to 75% and 70% | Improved |
| use FedAdam instead of FedAvg | FL on ImageNet · ImageNet | total energy (Wh): 2,901 -> 5,049 | Hurt |
| use FedAdam instead of FedAvg | FL on CIFAR10 non-IID · CIFAR10 | total energy (Wh): 30.3 -> 18.8 | Improved |
| use non-IID instead of IID partition | FedAvg 5 LE on CIFAR10 · CIFAR10 | communication rounds: 180 -> 580 | Hurt |
| use NX instead of TX2 edge device | FedAvg on ImageNet · ImageNet | total energy (Wh): 2,901 -> 2,494 | Improved |
| use NX instead of TX2 edge device | FedAdam on CIFAR10 · CIFAR10 | total energy (Wh): 18.8 -> 18.9 | No clear effect |
| train in France instead of other countries | centralized and FL · CO2e by country | CO2e emissions | Improved |
Closest papers by meaning
- Carbon Footprint Reduction for Sustainable Data Centers in Real-Time (AAAI 2024 · similarity 0.86)
- HiFi-Gas: Hierarchical Federated Learning Incentive Mechanism Enhanced Gas Usage Estimation (AAAI 2024 · similarity 0.84)
- Holistically Evaluating the Environmental Impact of Creating Language Models (ICLR 2025 · similarity 0.83)
- LLMCarbon: Modeling the End-to-End Carbon Footprint of Large Language Models (ICLR 2024 · similarity 0.82)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.82)
- FedCompetitors: Harmonious Collaboration in Federated Learning with Competing Participants (AAAI 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
