ABSTRACT
As a deep learning framework, MetaNN involves a large number of matrix operations, and correspondingly, the data types involved in MetaNN are matrix-based. This chapter discusses the type system and basic data types used in MetaNN. It explains the type system used in MetaNN from this perspective so as to bring readers a higher level of awareness of it. However, in the type system introduced by MetaNN, the conceptual affiliations between tags are not obvious and author deal with each type of tags separately. The chapter discusses the basic data types introduced by MetaNN in turn. Then the chapter provides a discussion of element-level reading and writing, which leads to the specific meaning of shallow copy and detection of write operations. Based on the discussion, it also provides an overview of the interfaces to be supported by existing categories of MetaNN.
