Neural Networks with the Wolfram Language

This chapter starts with the basic foundations of the neural network framework in the Wolfram Language. The chapter begins with the concepts of layers, how to use the commands for different layers, and the most common layers. You learn how to enter data into the layers by the net port and the different forms of equivalent expression of the layers. This topic is followed by how to distinguish different layers by their symbol. You see that layers can have multiple options that enable them to have various specifications by viewing the concept of a layer in the Wolfram Language scheme, comparing different layers with different purposes, and performing different computations. You also achieve this by looking at the various activation functions supported by the Wolfram Language and inspecting the plots of each function in addition to different syntax forms. Next, you learn about encoders and decoders and how these tools are used to construct a neural network model, depending on the task to be fulfilled. You then learn how these encoders and decoders are used to convert different data types to numeric arrays and how to convert the numeric arrays back to the initial data. You introduce the concept of a container, what it means for the created models, and what types exist. You see how to handle and build containers with different commands and graphically visualize the created model. You see how the Wolfram Neural Net Framework supports MXNet-related operations and how to export a network to the format of the MXNet operation.