Joel Nothman
Joel Nothman
Yes, the clone would refer to the same object.
A the sprint today we discussed four solutions for this, with a focus on transfer learning: 1. a freezing wrapper like #8374, #8372, #9464 2. labelling a base estimator with...
Maybe that isn't an important constraint, just as long as its behaviours are equivalent?
I mean that being an instance of the frozen type is not necessarily important as long as it has all attributes of the frozen type.
Frankly the downstream packages should be doing this integration testing by testing various use cases and checking with scikit-learn master. We should be encouraging downstream packages to include such integration...
I'd be happy to see, in the first instance, a script that runs the test suites of these downstream packages. We can look separately at how we best deploy that...
I think we should have an MRMR approach implemented, perhaps especially after we have dropped Randomized L1. PR welcome.
I'm not sure. SelectFromModel with lasso might be as good. perhaps we need to look at the computational complexity with lots of features also.
and sequential feature selection might be a more expensive variant again that would account for redundancy