Guidelines for Neural Network System Architecture Diagrams

Diagrams are often used to describe system architectures in Computer Science. Neural network systems are no exception. Without any framework or guidelines establishing good practices for their design, the current state is heterogeneous and evolving. Investigation of this field necessarily employs mixed methods. The research aims to improve diagrammatic representation of neural network systems by 1) deriving appropriate requirements and metrics 2) selecting appropriate representational guidelines and 3) create and implement an iterative framework for improving a set of diagram guidelines.