Dual Similarity Regularization for Recommendation

Abstract

Recently, social recommendation becomes a hot research direction, which leverages social relations among users to alleviate data sparsity and cold-start problems in recommender systems. The social recommendation methods usually employ simple similarity information of users as social regularization on users. Unfortunately, the widely used social regularization may suffer from several aspects: 1 the similarity information of users only stems from users' social relations; 2 it only has constraint on users; 3 it may not work well for users with low similarity. In order to overcome the shortcomings of social regularization, we propose a new dual similarity regularization to impose the constraint on users and items with high and low similarities simultaneously. With the dual similarity regularization, we design an optimization function to integrate the similarity information of users and items, and a gradient descend solution is derived to optimize the objective function. Experiments on two real datasets validate the effectiveness of the proposed solution.