The zero-sample study review published in ACM Trans Intell Syst Technol 10, 2 , Article 13 (January 2019)....
Summary : Most machine learning approach focuses on already seen its category in the training examples are classified. In fact, many applications need to be classified for instance, but not seen like these instances before. Zero sample learning (Zero-Shot Learning) is a powerful and promising learning paradigm in which categories covered by the training examples and think classification categories are disjoint. In this paper, zero-sample study conducted a comprehensive review. First, we outlined the zero-sample study. The data model used in the optimization, three zero learning learning samples provided. Secondly, different semantics described conventional zero spatial learning samples employed. Third, the existing zero-sample learning methods are classified, and a representative of each class presentation method. Fourth, the discussion of the different applications of zero-sample study. Finally, the study focuses on zero samples have promising future research directions.
