A novel loss estimator is introduced called FOLE (Fuzzy Objective Loss estimator) that aims to train a deep neural network in the fuzzy space and simaltaneusly ensures that valid fuzzy rules are being formed.
A latent fuzzy rules estimator is associated in parallel to the neural architecture that provides interpretable rules and their corresponding weights that can help understand how the neural architcture performs.
12dash / dnnfesfole Goto Github PK
View Code? Open in Web Editor NEWFinal Year Project for NTU. Learns a fuzzy neural architecture with interpretable rule and novel loss estimator for building consistency amongst the rules
Home Page: https://hdl.handle.net/10356/153483