Bayesian low-rank matrix completion with dual-graph embedding: Prior analysis and tuning-free inference

Recently, there is a revival of interest in low-rank matrix completion-based unsupervised learning through the lens dual-graph regularization, which has significantly improved performance multidisciplinary machine tasks such as recommendation systems, genotype imputation and image inpainting. While regularization contributes major part success, computational costly hyper-parameter tunning usually involved. To circumvent drawback improve completion performance, we propose novel Bayesian algorithm that automatically learns hyper-parameters associated with at same time, guarantee low-rankness completion. Notably, prior devised to promote encode information simultaneously, more challenging than single-graph counterpart. A nontrivial conditional conjugacy between proposed priors likelihood function then explored an efficient derived under variational inference framework. Extensive experiments using synthetic real-world datasets demonstrate state-of-the-art for various data analysis tasks.