Privacy-preserving distributed artificial intelligence in connectionism-based models

Abstract

Federated learning is an emerging machine learning approach that allows the construction of a model between several participants who hold their own private data. In the initial proposal of federated learning the architecture was a centralized neural network, and the aggregation was done with federated averaging, meaning that a central server will orchestrate the federation using the most straightforward averaging strategy. In this thesis we discuss several advances in aggregation methods, encryption methods to ensure the privacy of the system, study of non-iid datasets, and federation of Fuzzy Cognitive Maps.