We should make sure that the documentation mentions the one-vs.-other strategy (wrong for CNN). Then it makes sense to store all pairs of k-NN instead of the last one from the loop.
We can then deprecate estimator_ and create a new estimators_ attribute.
closes #908
We should make sure that the documentation mentions the one-vs.-other strategy (wrong for CNN). Then it makes sense to store all pairs of k-NN instead of the last one from the loop. We can then deprecate
estimator_and create a newestimators_attribute.