Recommender Systems

Full implementation is complex: it needs advanced linear algebra.

Types of Recommender Systems:

• Content based. Focus on the attributes of the items: the usual “related items”.
• Collaborative filter (CF). Uses “wisdom of the crowd” to recommend items: eg Amazon. CF is most used on content based systems. It can do feature learning by itself.
  The Movie land dataset of movies to study.  These methods can be:
  – Memory based CF: singular value decomposition.
  – Collaborative CF: computing cosine similarity.