Abstract:Collaborative Filtering (CF) has been widely used in recommendation due to its brevity and effectiveness. However, rating sparsity or even cold-start is still an inevitable problem when facing CF. Although multimodal information can be easily found in today's website, most existing CF methods are not well designed to incorporate them appropriately. In this paper, we intend to explore the potential methods of fusing abundant auxiliary cross-media information of user and items to help enhance traditional CF. In general, we divide the auxiliary data into discrete partial features (tags, genres ...) and continuous global features (images, videos ...) for different purposes. As for discrete features, embedding is adopted to represent them. In order to merge different discrete components, we use attention mechanism to adaptively learn their respective importance in contributing to a specific rating prediction. While for the continuous features, we use feature extraction network such as CNN to exploit informative representation to help restrain item's embedding space, consequently the interpretability of factorized vectors is improved. To validate the performance of our proposed method, extensive evaluations on public MovieLens dataset are carried. Experimental results on MSE improvements demonstrate both the importance of the auxiliary information and effectiveness of the proposed approach.

Mutual Information Based Similarity Measure for Collaborative Filtering

A Combinative Similarity Computing Measure for Collaborative Filtering

Improving Recommendations With Collaborative Factors

Nonparametric statistics inspired similarity measure for accuracy in collaborative filtering

A New User Similarity Model to Improve the Accuracy of Collaborative Filtering

An Hybrid Similarity Function for Neighbor Selection in Collaborative Filtering

An improved collaborative filtering method based on similarity

An Integrated Method for Similarity Computation

Information Filtering Via Collaborative User Clustering Modeling.

Combining similarity measures in content-based image retrieval guided by mutual information

A New Similarity Computation Method in Collaborative Filtering Based Recommendation System.

Similarity Measure and Instance Selection for Collaborative Filtering.

Collaborative Filtering Recommendation Algorithm based on Improved Similarity

A New Similarity Model Based on Collaborative Filtering for New User Cold Start Recommendation

A collaborative filtering algorithm based on correlation coefficient

Deep Collaborative Filtering Incorporating Auxiliary Multi-Media Information.

A hybrid user similarity model for collaborative filtering

Item Similarity Learning Methods for Collaborative Filtering Recommender Systems

Collaborative Filtering Algorithm Combining the Feature of Controversy

A Collaborative Filtering Algorithm for Improving Similarity Based on Probability Weighted Association Rules

A Fast Collaborative Filtering Algorithm for Implicit Binary Data