Abstract

Fuzzy theory is an old paradigm that has historically been applied in many fields of science and engineering. It provides a simple framework to implement systems able to codify imprecise human knowledge in terms of implication rules. This allows for human knowledge to be easily transformed into a workable algorithm. Lately, this approach lost interest in the big data era when we have plenty of data but sometimes lack of knowledge, which is key to design fuzzy logic based systems. However, recently, the approach to build fuzzy logic systems is changing: instead of creating them from human knowledge, big amounts of data are used to train the system, that is: to find the implication rules that link inputs and outputs and tune their parameters. In this talk, we will review the basics of fuzzy logic and provide some insights about how fuzzy logic systems can be learned from data. Also, an application of a trained fuzzy logic system for spectral curve recovery will be presented.