Which of the following is NOT a feature of federated learning?

Federated learning is characterized by its ability to train machine learning models across decentralized devices while keeping data localized. This method involves local training on sensitive data such that individual data remains on the user's device, enhancing privacy and security.

The feature that is not aligned with federated learning is centralized data storage. In federated learning, data is not aggregated in one central repository; instead, each device trains a model using its local data. After training, only the model updates (not the data) are sent to a central server, where they are aggregated to improve a global model without accessing the raw data.

Local training on sensitive data and training across decentralized devices ensure that privacy is maintained, while the aggregation of model updates allows for collaborative improvements to the machine learning model without compromising user data. Therefore, the presence of centralized data storage does not fit within the principles guiding federated learning practices, making it the correct choice in this context.