A Regularization Approach for Prediction of Edges and Node Features in Dynamic Graphs

Forecasting the behavior of systems with multiple responses has been a challenging problem in the context of many applications such as collaborative filtering, financial markets, or bioinformatics, where responses may be, respectively, movie ratings, stock prices, or activity of genes within a cell. Statistical modeling techniques have been widely applied for learning multivariate time series either in the multiple linear regression setting [3] or with autoregressive models [19]. More recently, kernel-based regularized methods have been developed for multitask learning [7, 2]. These approaches share in common the use of the correlation structure between input variables to enhance prediction of every single output. Frequently, the correlation structure is assumed to be given or is estimated separately.