Multi-agent Architecture for Federated Learning

The concept of federated learning has become widespread in working with data, mainly due to the fact that it allows training on data directly nodes where they are stored. As a result, no transfer is required. After completed each node, only trained model transmitted central server for aggregation. Multi-agent systems behave similar way, because agents allow you train machine models local devices, while preserving confidential information. ability interact other makes possible generalize (aggregate) such and reuse them. This article presents architecture multi-agent learning. It highlights elements make up agent platform structure JADE platform. Describes lifecycle all used perform full cycle MAC\_FL environment. configurations placement proposed architectures analyzed described: centralized, decentralized hierarchical.