A parameterized activation function for learning fuzzy logic operations in deep neural networks

摘要

We present a deep learning architecture for learning fuzzy logic expressions.Our model uses an innovative, parameterized, differentiable activation functionthat can learn a number of logical operations by gradient descent. Thisactivation function allows a neural network to determine the relationshipsbetween its input variables and provides insight into the logical significanceof learned network parameters. We provide a theoretical basis for thisparameterization and demonstrate its effectiveness and utility by successfullyapplying our model to five classification problems from the UCI MachineLearning Repository.