Federated Learning can help Reduce Carbon Emissions, study reveals.

Federated learning can help in the reduction of carbon emissions. Machine learning is among the biggest contributors of carbon emissions due to their substantial amount of energy consumption.

Researchers drawn from universities of Oxford, Cambridge and Avignon sought to explore whether federated learning could help lower carbon emissions in comparison to traditional machine learning. Their research concluded that federated learning could help reduce carbon emissions, though not at substantial level.

Federated learning allows mobile phones to collaboratively train from a shared prediction model while storing data on device. It is also referred as collaborative learning. The technique largely relies on machine learning that trains algorithms across various servers or edge devices.

The technology permits multiple users to create a common, robust machine learning prototype without necessarily sharing data. The method addresses critical issues such as data security, data privacy, access to heterogeneous data and data access rights.

The term ‘federated learning’ was invented by Google in 2016.  Federated learning is practically about migrating computations to data. Federated learning has some positive impact on the environment because of the cooling requests of datacenters.  However, researcher argued that federated learning wasn’t a silver bullet due to a number of factors that might distract its efficiency.