@Documented @DefaultDouble(value=0.0) @Parameter(value=java.lang.Double.class) @Qualifier @Target(value={METHOD,PARAMETER}) @Retention(value=RUNTIME) public @interface MeanDamping
Damping parameter for means in baseline predictors.
The smoothing enabled by this parameter is based on FunkSVD. The idea is to bias item or user means towards the global mean based on how many items or users are involved. So, if the global mean is \(\mu\) and smoothing \(\gamma\), the mean of some values is \[\frac{\gamma\mu + \sum r}{\gamma + n}\]