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

We propose a new classifier based on Dempster-Shafer (DS) theory and convolutional neural network (CNN) architecture for set-valued classification. In this classifier, called the evidential deep-learning pooling layers first extract high-dimensional features from input data. The are then converted into mass functions aggregated by Dempster’s rule in DS layer. Finally, an expected utility layer performs classification functions. end-to-end learning strategy jointly updating parameters. Additionally, approach selecting partial multi-class acts is proposed. Experiments image recognition, signal processing, semantic-relationship tasks demonstrate that proposed combination of deep CNN, layer, makes it possible to improve accuracy make cautious decisions assigning confusing patterns sets.