DAD/[Do's and Dont's] in Machine Learning Engineers
I am someone with an intermediate expertise in the field of machine learning. I used to fall victim for my own stupidities such as wrong architecture choices, selection of wrong api's, bug in custom losses, misinterpreting a frameworks design principle, no idea how to traina a no bias model, how to do proper feature engineering, how to do hyperparameter optimization, lack of knowledge about the best practices, project management etc etc. I needed a way to improve my skills and in the mean time be helpfull to the fellow learners.
Contributor can add their experience related to a topic, for example how they were able to implement this custom loss in the framework X. They can create a new directory with the name of the topic they wish to add. Topic can be anything. Some are added as tags in the description. There is no particular constraint on it.
With the help of this whole community, I think this effort can reduce pitfalls in a project/research, increse the outreach of best practices, reduce failure rates, point out potential issues, collaboration of varied fields etc.
Contributor shares his experience via Pull Request. Other contributors will review it. Accepts it if valid, suggests modifications etc as usual. Corrects if the contributor misinterpreted it. If so add it to related topic as new DAD.