That's odd since, yes, the cause is almost always that too many estimates are NaN. Can you point a debugger at it and try to figure out where it goes wrong with a watchpoint?
AFAIK the only reason this happens is if none of the test preferences can be estimated. So might make sure you are not somehow setting up a test where 0 or very few tests are run. On Sat, Oct 9, 2010 at 6:50 AM, Lance Norskog <[email protected]> wrote: > Why is my recommender evaluation returning NaN? I'm using the > AverageAbsoluteDifferenceRecommenderEvaluator. > > The individual calls to the recommender return valid results, and > stepping through the evaluator shows normal operation with floats > being shuffled around as one would expect. I made my own evaluation > loop by running Recommender.estimatePreference() manually. This shows > some variation from the SlopeOne recommender, but no crazy numbers or > NaNs. > > This is on 100k samples from the GroupLens 100m dataset. > > -- > Lance Norskog > [email protected] >
