On Aug 21, 2009, at 5:02 AM, Sean Owen wrote:
One meta-issue here for the library is this: there's a 'standard' item-based recommender algorithm out there, and we want to have that. And we do. So I don't want to touch it -- perhaps add some options to modify its behavior. So we're talking about maybe inventing a variant algorithm... or three or four. That's good I guess, though not exactly the remit I had in mind for the CF part. I was imagining it would provide access to canonical approaches with perhaps some small variants tacked on, or at least hooks to modify parts of the logic. Basically I also need to have a think about how to include variants like this in a logical way.
+1. Theory and practice are two different things. :-) We run into this in Lucene often as well. The goal is to create an API that has high out of the box performance, but gives people the proverbial rope to go hang. Documentation can often help give people the insight into what they need to know to properly tune the libraries. "Standard" collections like Netflix and GroupLens can be used to show "out of the box" performance, while still being careful of overtuning. Over time, we will want the Wiki to evolve to capture best practices for Taste and Mahout with people contributing real world experience. The code, of course, can evolve too. It need not be perfect the first time.
