Towards a Federated Fuzzy Learning System
Anna Wilbik, Paul Grefen
Abstract
The abundant availability of data allows the construction of predictive systems that support decision makers in business and society. A problem arises if an organization does not have a large enough data set by itself to construct a system of adequate quality. In this case, data across organizations has to be used, which introduces risks of data sharing. To overcome these risks, federated learning is getting increasingly popular to enable automated learning in distributed networks of autonomous partners without sharing raw data. So far, only crisp systems have been used in this context. The use of a fuzzy inference system can bring advantages to deal with vagueness and uncertainty in predictive systems. Therefore, in this paper we explore the (hopefully) happy marriage of federated learning and fuzzy inference mechanisms. We show that it is indeed possible to build a fuzzy inference model in a federated learning setting, resulting in a Federated Fuzzy Learning System (F <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> LS). We also show that this combination brings advantages to decision making that cannot be achieved with either mechanism in isolation.