Adversarial Deep Factorization for Recommender Systems

Recently, hacking from malicious users is a big challenge for major corporations and government organizations. Tensor factorization methods have been developed to learn predictive models in multi-criteria recommender systems by dealing with the three-dimensional (3D) user-item-criterion ratings. However, they suffer from the data sparsity and contamination issues in real applications. In order to overcome these problems, we propose a general architecture of adversarial deep factorization (ADF) by integrating deep representation learning and tensor factorization, where the side information is embedded to provide an effective compensation for tensor sparsity, and the adversarial learning is adopted to enhance the model robustness. Experimental results on three real-world datasets demonstrate that our ADF schemes outperform state-of-the-art methods on multi-criteria rating predictions. Specifically, the proposed model considers a brilliant combination with adversarial stacked denoising autoencoder (ASDAE), where the adversarial training is used to learn effective latent factors instead of being placed on the extrinsic rating inputs.