Enable evaluation of baseline predictors about lenskit HOT 3 CLOSED

Michael Ekstrand [email protected] on 2011-05-25 19:25:44 said:

In [0ff122ec96d2bde6d42ee74aea1bb9dff45838a1]:

Add `BaselineRatingPredictor` to allow baseline evaluation (closes #28)

* Implement `BaselineRatingPredictor`, a `DynamicRatingPredictor` delegating to a baseline predictor.
* Document difference between `BaselineRatingPredictor` and `BaselinePredictor`
* Turn off the default `@Normalized` binding so recommenders don't demand normalizers unnecessarily

Also, makes `ScriptedRecipeBuilder` public so scripts actually work.

Note: This comment has been automatically migrated from Bitbucket
Created by grouplens on 2013-02-01T21:55:21.940549+00:00, last updated: None

from lenskit.

ekstrand on 2011-05-25 19:23:05 said:

This is getting done for 0.0.3.

Note: This comment has been automatically migrated from Bitbucket
Created by grouplens on 2013-02-01T21:55:21.569178+00:00, last updated: None

from lenskit.

ekstrand on 2011-05-25 16:16:32 said:

Two ways to do this:

1. Introduce BaselineRatingPredictor as a RatingPredictor that is backed by a BaselinePredictor.
2. Introduce AbstractBaselinePredictor, subclassed by all baselines, that also implements the RatingPredictor and DynamicRatingPredictor interfaces.

(1) has naming confusion (what's the difference between a BaselinePredictor and a BaselineRatingPredictor?). (2) introduces complications with session/built distinctions for baselines with precomputed values (global mean, item means) - basically, any baseline predictor will need to be noticeably more complicated.

Note: This comment has been automatically migrated from Bitbucket
Created by grouplens on 2013-02-01T21:55:21.181490+00:00, last updated: None

from lenskit.