Open miguelgfierro opened 5 years ago
Yes, there existing many machine learning training process, and each of them makes sense to some extent.
We can make the code style consistent. Can you point me to the "rest of the methods"?
I can take a look at first. Previously I mainly followed my own conventions to program.
one of the patterns that we follow is single responsibility: https://github.com/Microsoft/Recommenders/wiki/Coding-Guidelines#single-responsibility, there there is an example for splitting train and test procedures
An example is https://github.com/Microsoft/Recommenders/blob/master/reco_utils/recommender/sar/sar_singlenode.py Basically the fit(), recommend_k_items() and predict() methods are separate. It is similar to the scikit-learn convention.
Description
Question to @Leavingseason, in this method:
would it be ok if we remove the
test_file
and the code that evaluates in the test file and makevalid_file
optional?The reason is for consistency with the rest of the methods.
However, I wouldn't like to touch a lot the base class in case you are using the test intensively.
FYI @anargyri @yexing99