2022 (modified: 16 Nov 2022)ICML 2022Readers: Everyone
Abstract:This paper presents a method to explain how the information of each input variable is gradually discarded during the forward propagation in a deep neural network (DNN), which provides new perspecti...