Pre-Conference Talk by TRAN Nhu Thuat | Multi-Representation Variational Autoencoder via Iterative Latent Attention and Implicit Differentiation

Variational Autoencoder (VAE) typically samples a single vector for representing user’s preferences, which may be insufficient to capture the user’s diverse interests. Existing solutions extend VAE by employing prototypes to group items into multiple clusters, each capturing one topic of user’s interests. Despite showing improvements, the current design could be more effective since prototypes are randomly initialized and shared across users, resulting in uninformative and non-personalized clusters. To fill the gap, firstly, we introduce iterative latent attention for personalized item grouping into VAE framework to infer multiple interests of users. Secondly, we propose to incorporate implicit differentiation to improve training of our iterative refinement model. Thirdly, we study the self-attention to refine cluster prototypes for item grouping, which is largely ignored by existing works.

This is a Pre-Conference talk for 32nd ACM International Conference on Information and Knowledge Management (CIKM 2023)