How should we structure a program whose main function is to calculate something; fitting data, simulating data, numerical calculations, machine learning model building and deployment, etc. (On ML I have found references but not so much the rest).
What would be of practical use is guidance on how to even do something simple. For example, what do I want to be thinking about for a computation that will only run on one machine, maybe not even automated. In other words, a ground-up approach before diving into how to translate things to a large scale would be most instructive.
For context, I am a self-taught programmer and I can write programs for any of the topics I mention and get the job done, but I don't have a good idea of if I'm doing it well from design/architecture point of view. For a concrete example, I rarely use classes, and it has never been a roadblock, but at the same time I know that means there are plenty of tools that could make my work better that I am blind to.
What I do lean on is more or less the following: Write a config file that changes for each data set to be used as input. Throw all of that that information into a dictionary. Then call a wrapper script that is basically a pipeline of functions that take the dictionary as input and outputs the dictionary is appended fields. In other words, at the start of writing a program. All other choices about how the script is written is made after work has started.
Aucun commentaire:
Enregistrer un commentaire