A Transfer Learning Approach for Network Modeling.

Area of Science:

• Network Science
• Machine Learning
• Neuroimaging

Background:

• Network models are crucial for understanding interacting systems across various domains.
• Identifying networks for multiple related tasks presents challenges, necessitating knowledge transfer.
• Transfer learning offers a promising solution to leverage information across similar tasks.

Purpose of the Study:

• To propose a novel transfer learning approach for identifying networks of related tasks.
• To enhance network learning robustness, especially with limited sample sizes.
• To apply the method for identifying brain connectivity networks in Alzheimer's disease (AD).

Main Methods:

• Developed a Bayesian hierarchical model to characterize task relatedness.
• Incorporated L1-regularization for robust network learning from limited data.
• Utilized an Expectation-Maximization (EM) algorithm for network parameter estimation.

Main Results:

• Simulation studies confirmed the proposed transfer learning approach's superiority over single-task learning.
• The method effectively identified shared information across related network learning tasks.
• Application to Alzheimer's disease (AD) fMRI data yielded findings consistent with existing literature.

Conclusions:

• The proposed Bayesian transfer learning framework with L1-regularization is effective for network identification.
• This approach offers significant advantages over isolated task learning, particularly in complex domains.
• The method shows promise for neuroimaging applications, such as analyzing brain connectivity in AD.