Federated learning is a machine learning technique that enables multiple actors to build a common, robust machine learning model without sharing data, thus addressing critical issues such as data privacy, data security, data access rights, and access to heterogeneous data. It is a decentralized approach to training machine learning models that doesnt require an exchange of data from client devices to global servers. Instead, the raw data on edge devices is used to train the model locally, increasing data privacy. The final model is formed in a shared manner by aggregating the local updates.
Here are some key features of federated learning:
- Multiple independent sessions: Federated learning trains an algorithm via multiple independent sessions, each using its dataset.
- Privacy: Federated learning enables machine learning models to be trained without anyone seeing or touching the data, offering a way to unlock information to feed new AI applications.
- Decentralized: Federated learning is a decentralized approach to building machine learning models where data from different sources is not collected and stored on one server, and then the model is trained on a single server too.
- Collaborative: Federated learning enables multiple people remotely to share their data to collaboratively train a single deep learning model, improving on it iteratively.
- Communication: Federated learning requires frequent communication between nodes during the learning process, which requires not only enough local computing power and memory but also high bandwidth connections to be able to exchange parameters of the machine learning model.
Federated learning has many applications in industries such as defense, telecommunications, Internet of Things, and pharmaceuticals. It can help individual institutions collaborate with others and makes it possible for the models to learn from more datasets securely. Federated learning also enables learning at the edge, meaning it brings model training to the data distributed on millions of devices.