Conceptual
Login

Designing PyTorch Runner Using The Information Expert Principle in Data Science Projects

The Information Expert principle, part of the GRASP set of object-oriented design principles proposed by Craig Larman, states that responsibility for a piece of behavior should be assigned to the class or module that already holds the information needed to fulfill it — i.e., software structure should follow the structure of data flow. Applying this principle means identifying which part of a system possesses the relevant data and consolidating the operations on that data there, rather than scattering duplicated logic across unrelated locations. This is a software design/object-oriented design concept applicable broadly, including to data science and machine learning pipeline architecture, where it guides how responsibilities for tracking metrics, running iterative loops, and managing state should be distributed among classes.