Timo Kaufmann
Timo Kaufmann
Slightly off-topic but I figure if someone has a similar problem he my end up here like I did so: If you just have a big monitor you want to...
> @timokau what do you think? I'm not sure if it would be quite that simple. For example the `_construct_models` function in `fate_network.py` uses `self.n_object-features_fit_` and is also called for...
So this implementation should always be overridden? Could we just remove it then?
Continuing the discussion in https://github.com/kiudee/cs-ranking/pull/118#discussion_r430021615 here for better visibility (since that PR is merged). CC @prithagupta @kiudee > While GenNestedModel, says every object to some extend is a part of...
Thanks for the context @prithagupta :)
Currently our `FETANetwork` base class requires the mandatory `n_objects` and `n_object_features` parameters. According to sklearns guidelines, those should be passed to `fit` instead. That effectively means that we should only...
I've adapted FETALinear's init. I'll have to do the same for our other cores and estimators and then take care of the other parts of the scikit-learn estimator interface.
This is turning out to be a lot more effort than expected. To make it somewhat manageable and reviewable I'll proceed as follows: - Gradually make all estimators pass the...
The first part is ready for review: https://github.com/kiudee/cs-ranking/pull/117
I have rebased this on top of #118. After #118, the biggest remaining blocker for the default-constructible is the `optimizer` parameter of our estimators. See #119 for that.