Ultra-Fast Approximate Inference Using Variational Functional Mixed Models.

Area of Science:

• Computational Statistics
• Functional Data Analysis
• High-Dimensional Data Analysis

Background:

• Bayesian functional mixed models are effective for complex functional data.
• Computational challenges in posterior sampling limit their application to high-dimensional data.

Purpose of the Study:

• To introduce a novel computational framework for ultra-fast approximate inference in high-dimensional functional data.
• To address the computational limitations of existing Bayesian functional mixed models.

Main Methods:

• Utilizes a parsimonious basis representation for functional observations.
• Employs variational Bayes for approximating posterior distributions, avoiding computationally intensive Markov Chain Monte Carlo (MCMC) sampling.
• Implements a fast iterative algorithm for parameter estimation and a rapid multiple testing procedure in basis space.

Main Results:

• The proposed framework enables efficient compression and parallel computing.
• Demonstrates ultra-fast approximate inference capabilities.
• Successfully identifies significant local regions indicating group differences in simulation studies and real-world datasets (proteomics, brain imaging).

Conclusions:

• The new framework offers a computationally efficient solution for high-dimensional functional data analysis.
• It provides a viable alternative to traditional methods, enabling faster and more scalable inference.
• The approach is validated through simulations and applications in diverse scientific domains.