Package es.uam.eps.ir.relison.links.recommendation.reranking.global.swap

Class SwapReranker<U,​I>

java.lang.Object
es.uam.eps.ir.relison.links.recommendation.reranking.global.swap.SwapReranker<U,​I>
Type Parameters:
U - type of the users.
I - type of the items.
All Implemented Interfaces:
GlobalReranker<U,​I>
Direct Known Subclasses:
SwapGreedyReranker
public abstract class SwapReranker<U,​I>
extends java.lang.Object
implements GlobalReranker<U,​I>
Abstract implementation for a family of reranking strategies for optimizing a global parameter of the recommendations (e.g. Gini). These reranking strategies start from a system which has a certain value of the property to optimize. Then, the rerankers consider an initial scenario where all the user-item pairs among the top-k recommended items for each user are added to the system. Then, we run over the top-k recommended items. For each of them, we check whether it is better to keep this item in the ranking, or choose one of the remaining (the not recommended ones) instead. If we decide that it is better to change them, we swap them, and update the global property accordingly.

  • Field Summary

    Fields 
    Modifier and Type Field Description
    protected double globalvalue
    Current global value of the metric.
    private long seed
    Seed for the order of the users

  • Constructor Summary

    Constructors 
    Constructor Description
    SwapReranker()
    Constructor.
    SwapReranker​(long seed)
    Constructor.

  • Method Summary

    Modifier and Type Method Description
    protected abstract es.uam.eps.ir.ranksys.core.Recommendation<U,​I> rerankRecommendation​(es.uam.eps.ir.ranksys.core.Recommendation<U,​I> rec, int maxLength)
    Reranks the recommendation for a user.
    java.util.stream.Stream<es.uam.eps.ir.ranksys.core.Recommendation<U,​I>> rerankRecommendations​(java.util.stream.Stream<es.uam.eps.ir.ranksys.core.Recommendation<U,​I>> recommendation, int maxLength)
    Reranks a set of recommendations.
    protected abstract void update​(es.uam.eps.ir.ranksys.core.Recommendation<U,​I> reranked)
    Updates the reranking algorithm values, using a certain recommendation.

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

  • Field Details

    • seed

      private final long seed
      Seed for the order of the users

    • globalvalue

      protected double globalvalue
      Current global value of the metric.

  • Constructor Details

    • SwapReranker

      public SwapReranker()
      Constructor. Default seed, not randomly chosen.

    • SwapReranker

      public SwapReranker​(long seed)
      Constructor.
      Parameters:
      seed - random seed for establishing the order in which we run over the users.

  • Method Details

    • rerankRecommendations

      public java.util.stream.Stream<es.uam.eps.ir.ranksys.core.Recommendation<U,​I>> rerankRecommendations​(java.util.stream.Stream<es.uam.eps.ir.ranksys.core.Recommendation<U,​I>> recommendation, int maxLength)
      Description copied from interface: GlobalReranker
      Reranks a set of recommendations.
      Specified by:
      rerankRecommendations in interface GlobalReranker<U,​I>
      Parameters:
      recommendation - a stream containing all the recommendations to rerank.
      maxLength - maximum length of the definitive ranking for each user.
      Returns:
      a stream of recommendations containing the definitive rankings.

    • update

      protected abstract void update​(es.uam.eps.ir.ranksys.core.Recommendation<U,​I> reranked)
      Updates the reranking algorithm values, using a certain recommendation.
      Parameters:
      reranked - the recommendation.

    • rerankRecommendation

      protected abstract es.uam.eps.ir.ranksys.core.Recommendation<U,​I> rerankRecommendation​(es.uam.eps.ir.ranksys.core.Recommendation<U,​I> rec, int maxLength)
      Reranks the recommendation for a user.
      Parameters:
      rec - the original recommendation.
      maxLength - maximum length of the definitive ranking.
      Returns:
      the new recommendation.