Bayesian Fused Lasso regression for dynamic binary networks

Abstract

We propose a multinomial logistic regression model for link prediction in a time series of directed binary networks. To account for the dynamic nature of the data we employ a dynamic model for the model parameters that is strongly connected with the fused lasso penalty. In addition to promoting sparseness, this prior allows us to explore the presence of change points in the structure of the network. We introduce fast computational algorithms for estimation and prediction using both optimization and Bayesian approaches. The performance of the model is illustrated using simulated data and data from a financial trading network in the NYMEX natural gas futures market. Supplementary material containing the trading network data set and code to implement the algorithms is available online.

References in corpus (13)

• The structure and function of complex networks
• Pathwise coordinate optimization
• Mixed membership stochastic blockmodels
• The solution path of the generalized lasso
• An Evolutionary Algorithm Approach to Link Prediction in Dynamic Social Networks
• Properties and refinements of the fused lasso
• A state-space mixed membership blockmodel for dynamic network tomography
• Nonparametric Link Prediction in Dynamic Networks
• On change point detection using the fused lasso method
• Analysis of the Formation of the Structure of Social Networks using Latent Space Models for Ranked Dynamic Networks
• A coordinate-wise optimization algorithm for the Fused Lasso
• Modelling and prediction of financial trading networks: An application to the NYMEX natural gas futures market
• High-dimensional Fused Lasso Regression using Majorization-Minimization and Parallel Processing

Cited by in corpus (6)

• Bayesian Dynamic Fused LASSO
• Logistic regression with total variation regularization
• Modelling Heterogeneity Using Bayesian Structured Sparsity
• Dynamic Networks with Multi-scale Temporal Structure
• Bayesian Fusion Estimation via t-Shrinkage
• Gibbs Sampling using Anti-correlation Gaussian Data Augmentation, with Applications to L1-ball-type Models