There's no deadline for this? I'll start taking a look this weekend, but will need to get up to speed on documentation format, inductive clustering, Pipeline and BaseSearchCV methods, and determine which one of the datasets is best used for this example. I welcome the challenge, and appreciate any shortcuts/suggestions. OK to submit the PR in pieces to collect feedback instead of only doing so when the example is fully polished?

Please submit a PR as soon as possible, it is always good to get feedback first. There is certainly no time constraint on it, but it is good that you say you want to work on it (so we don't duplicate work).

Omg, 15 days for a reply? I have been sick, sorry :-/

@amueller are we still looking at adding an example of this to the documentation? I could work on this. Any pointers on the kind of clustering or classification or dataset to use?

Often a 2d dataset is good for illustration purposes. You can also try something like digits, which might lead to interesting visualizations.

@amueller check out if http://chiragnagpal.com/blog/inductive_learning.. its an ipython notebook with what im trying to do.
Let me know if im on track

plus i think the right terminology for what we are trying to achieve here is "transductive learning" rather than "inductive learning"