HyLIM: Hybrid Linear Method for Recommender System

Abstract

In this research work, we propose Hybrid Linear Method (HyLIM) for ton-n recommender systems, which is a very simple and natural extension to SLIM (Sparse Linear Method) and its dense alternative, EASE (Embarrassingly Shallow Auto-Encoder). Showing its simple closed-form solution, we apply HyLIM to some real-world data (for which some side-information about both the users and the items is available), arguing that the side-information does matter (at least in this data) to the recommender's performance.