Bayesian Joint Spike-and-Slab Graphical Lasso.

Area of Science:

• Statistics
• Machine Learning
• Computational Biology

Background:

• Bayesian inference is crucial for statistical modeling.
• Gaussian graphical models represent conditional independence relationships.
• Existing methods like group and fused graphical lasso have limitations in model selection and bias.

Purpose of the Study:

• To develop a new class of priors for Bayesian inference with multiple Gaussian graphical models.
• To introduce Bayesian treatments for group graphical lasso and fused graphical lasso.
• To enable simultaneous self-adaptive shrinkage and model selection.

Main Methods:

• Extension of group and fused graphical lasso to a continuous spike-and-slab framework.
• Development of an Expectation-Maximization (EM) algorithm for posterior mode exploration.
• Application of Bayesian inference for multiple Gaussian graphical models.

Main Results:

• The proposed approach allows for efficient and automatic sparse model selection.
• The methods achieve substantially smaller bias compared to alternative regularization procedures.
• Demonstrated performance through simulations and real-world data analysis.

Conclusions:

• The novel Bayesian priors offer an effective solution for analyzing multiple Gaussian graphical models.
• The developed EM algorithm facilitates fast and dynamic posterior mode exploration.
• The proposed methods provide a robust and less biased alternative for sparse graphical model selection.