Paper Title
Evolution of Class Embedding for Unseen Class In Zero-Shot Learning

In this paper, we propose a novel method to improve the generation of the CE with the following two characteristics: (1) It can quantify the value of feature. and (2) In addition to utilizing the "positive" features, our method takes into account the impact of the "negative" features. We name our method as P-learning model. In this paper, we conduct experiments to replace the CE learned from the P-Learning model with the CE provided in the data set with the ZSL Model designed by others. Under the three different accuracy levels of Seen / Unseen / Generalized, the accuracy by our method is much higher than that reported in other papers. It shows that the CE learned by the P-Learning Model is very helpful and feasible in classification. Keywords - Zero-shot Learning, Class Embedding, Variational Auto-Encoder, Knowledge Representation.