ABSTRACT
Layers encapsulate operations to provide user-friendly interfaces. Users can organize the deep learning network through “layers” in MetaNN to train and predict models. It can be stated that “layer” is a concept quite attached to end-users. Deep learning is a fast-growing field with emerging new technologies. In order to support more technologies, on the one hand, author should simplify the introduction of new layers to the entire framework; on the other hand, existing layers ought to be flexible enough to support different formulations and adjust details of their behaviors. These are all what MetaNN needs to consider when designing the concept of layers. This chapter discusses the basic layers, whose main function is to encapsulate operations and form the most basic logic of execution.
