|
Activity Number:
|
362
- SPEED: Food, Environment, Biomedical Imaging and Physical System Visualization/Learning, Part 2
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 30, 2019 : 11:35 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #307780
|
|
|
Title:
|
Where Does Our Working Memory Take Place? a Multi-Level Sub-Graph Analysis of Brain Functional Connectivities
|
|
Author(s):
|
Maoran Xu* and Li Duan
|
|
Companies:
|
University of Florida and University of Florida
|
|
Keywords:
|
Brain network;
Graph Partition;
Matrix Factorization;
Bayesian Lasso
|
|
Abstract:
|
Working memory is the process of encoding, retention, and retrieval of information. EEG experiments have revealed distinct levels of brain connectivity in handling different loads of memory tasks. However, it remains mysterious which parts of the brain are responsible for such a difference. Challenges come from the high dimensionality of the brain networks and the difficulty to quantify the uncertainty of random variation. In this work, we develop a flexible mixed-effect Bayesian model combining with a multi-scale decomposition of the connectivity matrix which is named ``Treed-SVD''. This method provides binary tree structures that reduce the analysis of networks to a very scalable logistic regression. Compared to existing studies relying on full-scale or shrinkage-based models, our network model produces much richer and more interpretable results, together with useful tests at varying statistical powers. This work provides an easy-to-use toolbox for neuroscientists to explore and identify the potentially important sub-brain.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2019 program
|
