VM-Rec: A Variational Mapping Approach for Cold-start User Recommendation

TL;DR

This work proposes a Variational Mapping approach for cold-start user Recommendation (VM-Rec), mapping from few initial interactions to expressive embeddings for cold-start users, demonstrating superior performance compared to other popular cold-start methods.

Abstract

The cold-start problem is a common challenge for most recommender systems. The practical application of most cold-start methods is hindered by the deficiency in auxiliary content information for users. Moreover, most methods necessitate simultaneous updates to the extensive parameters of recommender models, leading to significant training costs, particularly in large-scale industrial scenarios. We observe that the model can generate expressive embeddings for warm users with relatively more interactions. Initially, these users were cold-start users, and after transitioning to warm users, they exhibit clustering patterns in their embeddings with consistent initial interactions. Based on this motivation, we propose a Variational Mapping approach for cold-start user Recommendation (VM-Rec), mapping from few initial interactions to expressive embeddings for cold-start users. Specifically, we encode the initial interactions into a latent representation, where each dimension disentangledly signifies the degree of association with each warm user. Subsequently, we utilize this latent representation as the parameters for the mapping function, mapping (decoding) it into an expressive embedding, which can be integrated into a pre-trained recommender model directly. Our method is evaluated on three datasets using the same base model, demonstrating superior performance compared to other popular cold-start methods.

Figures (3)

• Figure 1: Exploration of the distribution of embeddings using the BPR model on the Ali Display Ad Click dataset.
• Figure 2: Predicting an expressive embedding for cold-start users involves two main stages: 1. Parameter Generation Phase: Based on the encoding of initial interactions, a variational spike-and-slab distribution is generated, from which parameters $\mathbf{w}$ for personalized mapping functions are through the process of reparameterization or sampling (R/S). 2. Mapping Phase: The mapping phase effectively involves selecting specific pre-trained warm user embeddings $\Phi^U$ for weighted summation, and obtains the predicted embedding $\hat{\phi}_{u^c}$.
• Figure 3: The impact of varying base model, $\beta$ and proportion on performance.