An evidential deep-learning classifier based on Dempster-Shafer theory and deep learning

The proposed classifier consists of several stages for feature representation, a DS layer to construct mass functions, and an expected utility layer to make set-valued assignments based on the mass functions.

本文提出的分类器包含多个特征表示阶段、一个用于构建基本信任分配函数（mass functions）的DS层，以及一个基于上述基本信任分配函数完成集合值分配的期望效用层。