There is MLKRR.fit, but no predict method ?!

lcmd-epfl / MLKRR

Code for the Metric Learning for Kernel Ridge Regression algorithm

MIT License

8 stars 4 forks source link

Open UnixJunkie opened 3 months ago

UnixJunkie commented 3 months ago

I am dreaming, right?

UnixJunkie commented 3 months ago

sklearn.RandomForestRegressor.fit sklearn.RandomForestRegressor.predict

UnixJunkie commented 3 months ago

my god using this MLKRR package is such a pain...

UnixJunkie commented 3 months ago

training and test data must be kept separate

at training time, the test data might not be available; still, we might want to train then save a model. The test data not being there also enforces that it is not "leaking" in any way to the training data
at production usage time, only the test data might be available (we are reloading a trained model, the training data is no more around)