• echo state networks
• federated learning
• recurrent networks
• reservoir computing

Nowadays different types of data are transferred between people. These data are private to their owners and the users are not willing to share their local private data. There are three main challenges in order to create a machine learning model in this scenario. First, huge amounts of data are in the network which is not centralized. Second, these data can be in any form, including sequential or time-series data. Finally, the data are private.
In the thesis we introduce a federated reservoir computing framework that is able to train the model in a decentralized fashion, being able to process and predict sequential or time-series data. Moreover, no sensitive data is transferred in the network and users don’t know any knowledge about each other.