DCNFIS: Deep Convolutional Neuro-Fuzzy Inference System

  • 2023-09-15 18:20:48
  • Mojtaba Yeganejou, Kimia Honari, Ryan Kluzinski, Scott Dick, Michael Lipsett, James Miller
  • 0


A key challenge in eXplainable Artificial Intelligence is the well-knowntradeoff between the transparency of an algorithm (i.e., how easily a human candirectly understand the algorithm, as opposed to receiving a post-hocexplanation), and its accuracy. We report on the design of a new deep networkthat achieves improved transparency without sacrificing accuracy. We design adeep convolutional neuro-fuzzy inference system (DCNFIS) by hybridizing fuzzylogic and deep learning models and show that DCNFIS performs as accurately asthree existing convolutional neural networks on four well-known datasets. Wefurthermore that DCNFIS outperforms state-of-the-art deep fuzzy systems. Wethen exploit the transparency of fuzzy logic by deriving explanations, in theform of saliency maps, from the fuzzy rules encoded in DCNFIS. We investigatethe properties of these explanations in greater depth using the Fashion-MNISTdataset.


Quick Read (beta)

loading the full paper ...