Latent grouping models for user preference prediction

Abstract

We tackle the problem of new users or documents in collaborative filtering. Generalization over users by grouping them into user groups is beneficial when a rating is to be predicted for a relatively new document having only few observed ratings. Analogously, generalization over documents improves predictions in the case of new users. We show that if either users and documents or both are new, two-way generalization becomes necessary. We demonstrate the benefits of grouping of users, grouping of documents, and two-way grouping, with artificial data and in two case studies with real data. We have introduced a probabilistic latent grouping model for predicting the relevance of a document to a user. The model assumes a latent group structure for both users and items. We compare the model against a state-of-the-art method, the User Rating Profile model, where only the users have a latent group structure. We compute the posterior of both models by Gibbs sampling. The Two-Way Model predicts relevance more accurately when the target consists of both new documents and new users. The reason is that generalization over documents becomes beneficial for new documents and at the same time generalization over users is needed for new users.

A social recommendation framework based on multi-scale continuous conditional random fields

“…Typical examples include Pearson Correlation Coefficient (PCC) [27] and Vector Similarity (VS) [4]. Alternatively, model-based methods [7,13,32,33] are from a probabilistic perspective, which builds a probabilistic model to calculate the expectation of a user's rating on an item. A classical way is to utilize probabilistic latent class [13].…”

“…Traditionally, CF algorithms are used in these systems, assigning each user-item pair a score indicating the user's rating on the item, based on which a ranking list of items is generated to the user as suggestions. Classical CF methods are divided into memory-based methods [4,12,14,20,30,34] and model-based methods [7,13,32,33]. Recently, social relations have been considered in many applications, and in this paper, we address the issue of social recommendation.…”

Two-Way Grouping by One-Way Topic Models

“…This problem of unseen or almost unseen users and documents is generally referred to as the cold-start problem in recommender system literature, see for instance [23]. The Two-Way Model was proposed to tackle this problem of either new users or new documents [22,24].…”

“…In addition, there is a two-way grouping model, called Flexible Mixture Model (FMM; [21]). We have discussed the main differences between our Two-Way Model and these related models in [22].…”

Region-Based Correspondence Between 3D Shapes via Spatially Smooth Biclustering

“…Biclustering is a well-studied problem, with applications to gene expression data, recommender systems, market segmentation, and other areas [7,8,14,16,25,36]. However, maybe surprisingly, biclustering has not been used for shape correspondence, with the notable exception of [10].…”